IITM Journal of Management and IT

Indexed in:
Google Scholar, EBSCO Discovery, indianjournals.com & CNKI Scholar
(China National Knowledge Infrastructure Scholar)

 Monitoring water quality by sensors in Wireless Sensor Networks-A Review

Rakesh Kumar Saini
 Multimode Summarized Text to Speech Conversion Application
Archit Sehgal , Gitika Khanna
 Aadhaar Smart Meter: A Real-Time Bill Generator
Ramneek Kalra, Vijay Rohilla
A Review on Optical Character Recognition
Archit Singhal, Bhoomika
 Prediction of Heart Attack Using Machine Learning
Akshit Bhardwaj, Ayush Kundra, Bhavya Gandhi, Sumit Kumar, Arvind Rehalia,.
Manoj Gupta
 Detection and Prevention Schemes in Mobile Ad hoc Networks
Jeelani, Subodh Kumar Sharma, Pankaj Kumar Varshney
 A Review on Histogram of Oriented Gradient
Apurva Jain, Deepanshu Singh
 Breast Cancer Risk Prediction
Pankaj Kumar Varshney, Hemant Kumar , Jasleen Kaur, Ishika Gera
 CYBER: Threats in Social Networking Websites and Physical System Security
Tripti Lamba Ashish Garg
 Comparative Analysis of Different Encryption Techniques In Mobile Ad-Hoc
Networks
(MANETS)
INSTITUTE OF INFORMATION TECHNOLOGY AND MANAGEMENT Apoorva Sharma, Gitika Kushwaha
NAAC & NBA Accredited, Approved by AICTE, Ministry of HRD, Govt. of India  Fuchsia OS - A Threat to Android
Category 'A' Institute, An ISO 9001: 2015 Certified Institute Taranjeet Singh , Rishabh Bhardwaj
Affiliated to Guru Gobind Singh Indraprastha University, Delhi
D-29, Institutional Area, Janakpuri, New Delhi - 110058
 Sentiment Analysis using Lexicon based Approach
Tel: 91-011-28525051, 28525882 Telefax: 28520239 Rebecca Williams, Nikita Jindal, Anurag Batra
E-mail: director@iitmipu.ac.in, journal@iitmipu.ac.in
Website: www.iitmjanakpuri.com, www.iitmipujournal.org ISSN (PRINT): 0976-8629
ISSN (ONLINE): 2349-9826
www.iitmipujournal.org
ISSN (Print): 0976-8629 www.iitmipujournal.org
ISSN (Online): 2349-9826

Indexed in:
Google Scholar, EBSCO Discovery, indianjournals.com & CNKI Scholar (China
National Knowledge Infrastructure Scholar)

IITM Journal of Management and IT

Volume 10 Issue 1 January - June 2019

CONTENTS

Research Paper & Articles

• Monitoring water quality by sensors in Wireless Sensor Networks-A Review 1

Rakesh Kumar Saini
• Multimode Summarized Text to Speech Conversion Application 6
Archit Sehgal , Gitika Khanna
• Aadhaar Smart Meter: A Real-Time Bill Generator 11
Ramneek Kalra, Vijay Rohilla
• A Review on Optical Character Recognition 15
Archit Singhal, Bhoomika
• Prediction of Heart Attack Using Machine Learning 20
Akshit Bhardwaj, Ayush Kundra, Bhavya Gandhi, Sumit Kumar,
Arvind Rehalia,. Manoj Gupta
• Detection and Prevention Schemes in Mobile Ad hoc Networks 25
Jeelani, Subodh Kumar Sharma, Pankaj Kumar Varshney
• A Review on Histogram of Oriented Gradient 34
Apurva Jain, Deepanshu Singh
• Breast Cancer Risk Prediction 37
Pankaj Kumar Varshney, Hemant Kumar , Jasleen Kaur, Ishika Gera
• CYBER: Threats in Social Networking Websites and Physical System Security 46
Tripti Lamba Ashish Garg
• Comparative Analysis of Different Encryption Techniques In Mobile Ad-Hoc
Networks (MANETS) 55
Apoorva Sharma, Gitika Kushwaha
• Fuchsia OS - A Threat to Android 65
Taranjeet Singh , Rishabh Bhardwaj
 Sentiment Analysis using Lexicon based Approach 68
Rebecca Williams, Nikita Jindal, Anurag Batra

Volume 10, Issue 1 • January-June, 2019

 Monitoring water quality by sensors in Wireless Sensor
Networks-A Review
Rakesh Kumar Saini
Department of Computer Science & Application, DIT University, Dehradun
Uttrakhand, India
rakeshcool2008@gmail.com
Abstract– Monitoring water quality is critical to
human health, hence employing wireless sensor
netwoks for such a task requires a system that is
robust, secure and has a reliable communication.
Water borne diseases have become a major challenge
to human health. Around 400 million cases of such
diseases are reported annually, causing 6–12 million
deaths world-wide. Access to safe drinking water is
important as a health and development issue at
national, regional and local level. The population in
rural India mainly dependent on the ground water as
a source of drinking water. The main problems at the
time to effectively implement sensors are that, on one
hand, there is a lack of standards for contamination
testing in drinking water on the other hand, there are
poor links between available sensor technologies and
water quality regulations. In this paper the
application of WSN in environmental monitoring,
with particular emphasis on water quality. Various
WSN based water quality monitoring methods
suggested by other authors are studied and analyzed,
taking into account their coverage, energy and
security concerns.
Keywords-- Water quality monitoring, Remote,
Wireless Sensor Network
I. INTRODUCTION
Recent advances in MEMS-based sensor technology,
low-power analog and digital electronics, and low-
power RF design have enabled the development of
relatively in- expensive and low-power wireless
micro sensors that are capable of detecting ambient
conditions such as temperature and sound. Sensors
are generally equipped with data processing and
communication capabilities [1][2]. The sensing
circuitry measures parameters from the environment
surrounding the sensor and transforms them into an
electric signal. Processing such a signal reveals some
properties about objects located and/or events
happening in the vicinity of the sensor. Each sensor
Volume 10, Issue 1 ∙January-June 2019
has an onboard radio that can be used to send the
collected data to interested parties. Such
technological development has encouraged
practitioners to envision aggregating the limited
capabilities of the individual sensor in a large scale
network that can operate unattended. Numerous civil
and military applications can be leveraged by
networked sensors. A network of sensors can be
employed to gather meteorological variables such as
temperature and pressure. One of the advantages of
wireless sensor networks (WSNs) is their ability to
operate unattended in harsh environments in which
contemporary human-in-the-loop monitoring
schemes are risky, inefficient and sometimes
infeasible. Therefore, sensors are expected to be
deployed randomly in the area of interest by a
relatively uncontrolled means, e.g. dropped by a
helicopter, and to collectively form a network in an
ad-hoc manner. Given the vast area to be covered, the
short life span of the battery-operated sensors and the
possibility of having damaged nodes during
deployment, large population of sensors are expected
in most WSNs applications.
In most part of the world, ground water is the only
and important supply for production of drinking
water, particularly in areas where water supply is
limited. Groundwater quality will directly affect
human health [3]. A sensor is the electronic device
that can detects and responds to the external stimulus
from the physical environment. The external stimulus
can be temperature, heat, moisture, pressure in the
environment. The output of the sensors is generally a
signal which can be converted into human readable
form. Sensors may be classified as analog sensors
and digital sensors. Analog sensor senses the external
parameters (wind speed, solar radiation, light
intensity etc.) and gives analog voltage as an output
[4]. A digital sensor is an electronic or
electrochemical sensor, where data is digitally
converted and transmitted. A base station is
1
responsible for capturing and providing access to all
measurement data from the nodes, and can
sometimes provide gateway services to allow the data
to be managed remotely. The importance of
maintaining good water quality highlights the
increasing need for advanced technologies to help
monitor water and manage water quality. Sensors in
Wireless sensor networks offer a promising
infrastructure to municipal water quality monitoring
and surveillance. In this paper we suggest some
parameters for drinking water quality and proposed
model and methods for water quality [5][6].
II. SENSORS
Sensors are electronic devices that can be used for
monitoring physical and environmental condition
such as temperature, vibration and sound. Sensors are
defined as the sophisticated devices which aids to
detect and respond to electrical or optical signals. It
converts the physical parameters like temperature,
blood pressure, humidity, speed into a signal that can
be measured electrically. Sensors can be classified
based on the following criteria and conditions:
a. Primary Input quantity
b. Transduction principle
c. Material and Technology
d. Property Applications
Figure 1 represents another category of sensors.
Figure 1. Classification of Sensors
2 IITM Journal of Management and IT
III. CHALLENGES OF MONITORING
WATER QUALITY
Access to safe drinking water is important as a health
and development issue at national, regional and local
level. The population in rural indie mainly dependent
on the ground water as a source of drinking water. As
India is a developing country and it has wide-spread
emerging technologies, there is a need for system for
timely help and to monitor water pollution on the
total state of the water system [7] [8]. Monitoring of
water quality is very important for good health.
Wireless sensor networks can help for control quality
of water by using some methods but there are some
challenges of monitoring water quality are:
(a) Sensors Costs and Specifications
(b) Low energy of sensors
(c) Security
(d) Connectivity
(e) Sensor location
(a) Sensors Costs and Specifications
Site specific installation cost projections need to be
developed. Typical cost components in the total cost
of a sensor installation at each potential location
include:
1. Land purchase
2. Construction of the vault in which the sensors
and connections to the distribution piping will be
located
3. Installation of the sensors and RTU
4. Supplying power to the site
5. Installing the communications equipment and
upgrading/installing equipment at the central
control room
6. Design and bidding the construction and
installation work
7. Access to site
(b) Low energy of sensors
In Wireless Sensor Networks, Energy is a scarcest
resource of sensor nodes and it determines the
lifetime of sensor nodes. These are battery powered
sensor nodes. These small batteries have limited
power and also may not easily rechargeable or
removable. Long communication distance between
sensors and a sink can greatly drain the energy of
sensors and reduce the lifetime of a network. In
Wireless Sensor Networks, energy of sensors is a strengths. Relative coordinates of neighboring nodes
major issue to be considered [9]. can be obtained by exchanging such information
between neighbors. To save energy, some location
(c) Security
based schemes demand that nodes should go to sleep
Underwater wireless sensor network is a novel type if there is no activity. More energy savings can be
of underwater networked system. Due to the obtained by having as many sleeping nodes in the
characteristics of Underwater Wireless Sensor network as possible.
Network and underwater channel, Underwater
IV. QUALITY RANGE OF SUGGESTED
Wireless Sensor Network are vulnerable to malicious
PARAMETERS FOR DRINKING WATER
attacks. The existing security solutions proposed for
WSN cannot be used directly in Underwater Wireless Parameters examined by US environmental
Sensor Network. Moreover, most of these solutions protection agency, resolved that both chemical and
are layer wise. In figure 2 all sensors sense the data biological waste has an adverse effect on many water
under water and send these data to the base station. monitoring parameters such as pH, ORP, Turbidity,
Different types of sensors can be used under the Nitrates, Dissolved Oxygen, Water temperature,
water for monitoring quality of water [10]. Fluoride, Chlorine, Oxygen. In order to detect the
water contamination or impurities, it is enough to
determine the changes in the suggested parameters
(Table 1).if there is any deviation when compared to
that of the drinking water standards recommended by
WHO or Central pollution control board, India, then
the water is not safe for drinking [13].
Table 1: Quality range of suggested parameters
Quali
Sr Paramete Measure
Units ty
. r d Cost
range
6.5-
1 pH pH Low
9.0
650-
2 ORP mV Low
800
Figure 2. Underwater Wireless Sensor Network
3 Turbidity NTU 0-5 Medium
(d) Connectivity 4 Nitrates mg/L <10 High
Dissolved
The most fundamental problem in underwater sensor 5 mg/L 0.2-2 Medium
Oxygen
network is network connectivity. The connectivity
problem reflects how well a sensor network is Water
tracked or monitored by sensors. An underwater 6 temperatur C <10 Low
wireless sensor networks is the emerging field that is e
having the challenges in each field such as the mg/lit <0.05
7 Fluoride Medium
deployment of nodes, routing, floating movement of re –0.2
sensors etc. For the connectivity of nodes use DFS mg/lit
8 Chlorine 0.2–1 Low
(Depth first search) algorithm and for coverage use re
distributed coverage algorithm [11] [12]. 9 Oxygen mg/l 1-2 High

(e) Sensor location

Sensor nodes are addressed by means of their
locations. The distance between neighboring nodes
can be estimated on the basis of incoming signal

Volume 10, Issue 1 ∙January-June 2019 3

V. PROPOSED MODEL FOR WATER Ganga Rejuvenation, Government of
QUALITY india.http://cgwb.gov.in/.
[2] U.S.Environmental Protection Agency,
―Drinking water standards and health
advisories‖.Tech.Rep.EPA 822-S-12-001,
2012.
[3] Water Resource Information System of
India.http:www.india-
wris.nrsc.gov.in/wrpinfo/index.php?title=Rive
r_Water_Quality_Monitoring.
[4] Szabo, J. and Hall, J., Detection of
Figure 3. Proposed model for monitor quality of
contamination in drinking water using
drinking water
fluorescence and light absorption based online
Proposed model (Figure 3) for monitor quality of sensors, EPA/600/R-12/672 (2012),
drinking water is very effective for monitor ground http://www.epa.gov/ord.
water. In this proposed model all sensors sense the
[5] Homeland Security Presidential
data and collect data then stored these data in local-
Directive/HSPD-9, United States of America.
controller. Once the local-controller receives the
data, it then transferred to the cloud for analyzing the [6] Hart, D. B., Klise, K. A., McKenna, S. A. and
data. Cloud storage work as a mediator between data Wilson, M. P., CANARY User‘s Manual,
transmission layer and database management layer. Version 4.2. EPA-600-R-08e040 (2009), U.S.
After analyzing data by cloud storage data Environmental Protection Agency, Office of
transferred to end-user. Domestic water supplied Research and Development, National
form Municipal Corporation or directly takesn from Homeland Security Research Centre,
the ground water are mainly used for drinking and Cincinnati, OH.
cooking purposes.
[7] Laine, J., Huovinen, E., Virtanen, M. J. et al.,
Traditional water supply management An extensive gastroenteritis outbreak after
involves<storing the pool of water at various drinking-water contamination by sewage
locations and distributing the same through water effluent, Finland, Epidemiol. Infect. (2010)
head tanks and domestic pipelines. 139 (7), pp. 1105–1113.
VI. CONCLUSIONS [8] Szabo, J. and Hall, J., Detection of
contamination in drinking water using
The main objective of this paper is to show the role
fluorescence and light absorption based online
of sensors for the betterment of water quality in order
sensors, EPA/600/R-12/672
to obtain a hygienic environment. In this paper we
(2012),http://www.epa.gov/ord.
study the challenges of monitor quality of drinking
water and parameters required for monitor drinking [9] Bergamaschi, B., Downing, B., Pellerin, B.
water. According to the study, drinking water and Saraceno, J. F., In Situ Sensors for
obtained from both groundwater and surface water Dissolved Organic Matter Fluorescence:
must satisfied the standards for safe drinking water. Bringing the Lab to the Field, USGS Optical
This paper gives a clear view about what is a sensor, Hydrology Group, CA Water Science Centre.
parameters to identify quality of water, and stages to
[10] Hart, D. B., Klise, K. A., McKenna, S. A. and
create online water quality management system.
Wilson, M. P., CANARY User‘s Manual,
REFERENCES Version 4.2. EPA-600-R-08e040 (2009), U.S.
Environmental Protection Agency, Office of
[1] Central Ground Water Board, Ministry of
Research and Development, National
water resources, River development and

4 IITM Journal of Management and IT

 Homeland Security Research Centre, [12] Homeland Security Presidential
Cincinnati, OH. Directive/HSPD-9, United States of America.
[11] Hart, D. B., Klise, K. A., McKenna, S. A. and [13] Storey, M. V., Van der Gaag, B. and Burns, B.
Wilson, M. P., CANARY User‘s Manual, P., Advances in online drinking water quality
Version 4.2. EPA-600-R-08e040 (2009), U.S. monitoring and early warning systems, Water
Environmental Protection Agency, Office of Research 45 (2011), pp. 741–74.
Research and Development, National
Homeland Security Research Centre,
Cincinnati, OH.

Volume 10, Issue 1 ∙January-June 2019 5

 Multimode Summarized Text to Speech Conversion
Application
Archit Sehgal 1, Gitika Khanna 2
1,2
Department of Computer Science, HMR Institute of Technology & Management
Hamidpur, Delhi-110036, India
Archit_150@yahoo.co.in, gitikakhanna392@gmail.com
Abstract - This paper draws focus towards construction of meaningful summary by selecting
summarizing the tremendous amount of data useful paraphrases from the text available. The
collected from various sources and presenting the summarized text is then transformed into speech using
output as speech. In recent years, huge data sets are Text-to-Speech Synthesizer (TTS). The whole
being generated every moment and it becomes approach is categorized into three phases which are
difficult to manage it. In order to extract relevant text extraction from input, formation of summary and
information, an innovative, efficient and real- time conversion of the same into speech.
cost beneficial technique is required that enables
II. LITERATURE REVIEW
users to hear the summarized content instead of
reading it. This kind of application is beneficial for Mrunmayee Patil[1] This paper tells us about an
visually impaired and people with disabilities. Text OCR system to recognize the characters from image.
Rank algorithm, a ranking based approach is Edge detection and Image segmentation plays a
proposed with a variation in similarity function to significant role in extraction of text from image. The
make summary based on the scores computed for algorithm which can be used to summarize the
each sentence. The summarized text is then spoken extracted text works similar to PageRank Algorithm
out using text-to-speech synthesizer (TTS). discussed in the paper for web search engines [10].
Modifications can be made to make the TextRank
Keywords - TextRank, PageRank, Lexemes, Image
algorithm more effective. Sunchit Sehgal[5] This
Segmentation, Character Recognition, Text-to-
paper represents a way to make the algorithm more
Speech (TTS).
efficient by taking the score of the title in account.
I. INTRODUCTION Marcia A. Bush [19] shows us the efforts put in the
research of recognition of documents and their
In our proposed work of collecting data from different
prediction models This has enabled us to analyze the
sources and converting it into summarized text, we
signal based processes taking vocabulary, font and the
develop a cost efficient and user friendly interface.
sentence formation sequence into account.
The input to the application can be an image, audio or
video. While converting the input into editable text, III. OVERVIEW OF IMAGE ANALYSIS
there are various techniques used such as image
Over the decades, many researchers have been looking
processing, image segmentation [1] and edge
for possible ways of retrieving data from images and
detection. The approach direct towards format
video content. In a research paper, a framework was
conversion, where audio, video or image data is
proposed that will decompose the scanned image into
converted into symbolic representations that fully
its constituent visual patterns and the parsed results
describe the content. In case of an image, the
will be converted into semantically meaningful text
segmented characters are obtained from preprocessing
report. A model was also introduced where the users
of images. It is then provided as input to the Optical
will send the image of their respective meter‘s display
Character Recognition (OCR) to obtain the converted
screen along with the kilo-watt information [2]. The
text. In order to manage the enormous amount of
information will then be processed to convert it into
information, the derived text is summarized using a
text.
graph based technique i.e. TextRank Algorithm.
Image analysis [3] is the extraction of information
The TextRank Algorithm has application in
from images using different image processing

6 IITM Journal of Management and IT

approaches such as image filtering, image
compression, image editing and manipulation, image
preprocessing, image segmentation, feature
extraction, object recognition. An Image can be
considered as a matrix of square pixels arranged in the
pattern of rows and columns. It can be considered as a
linear sequence of characters.
Fig 1. Conversion of Image to text using Optical
Character Recognition
Fig 1. Depicts various blocks responsible for detecting
text from the image. Once the image is scanned, it is
preprocessed to remove any noise and is further
divided into segments. Every segment has its own
unique feature which must be further extracted and
classified to specific groups.
Edge detection and image segmentation are important
aspects in image analysis. Edge detection
differentiates different regions of an image by
identifying the change in gray scale and texture. Image
Segmentation is another technique which divide and
decompose image for further processing. It categorizes
the pixels with similar gray scale values and organizes
it into higher level units so that the objects become
more meaningful. The proposed system will work in
various phases. The input image will undergo pre-
processing such as removing noise induced due to the
technique applied for thresholding and improving the
quality. The image will undergo image segmentation
to separate non-text part present in the next step.
Further, feature extraction is performed to extract
preliminary features and comparing the same which
are stored in the database. Sometimes, there are often
error in which characters might be blurred or broken.
They are processed in post-processing stage.
IV. AUDIO ANALYSIS
Conversion of real time speech to text requires special
techniques as it must be quick and precise to be
Volume 10, Issue 1 ∙January-June 2019
recognizable. In an automatic speech recognition
system, the size of vocabulary affects the performance
of the system. Amidst the initial process, the system
learns about pattern, different speech sounds which
embody the vocabulary of the application. If there is
any unknown pattern, it is identified using the cluster
of references. The whole approach can be categorized
as phases such as analysis, feature extraction,
modeling and testing. The analysis phase is used to
extract information about speaker identity using vocal
tract, behavior feature and excitation source. Since
every speech has different characteristics which can be
fetched in the feature extraction phase in order to deal
with the speech signals.
V. TEXTRANK ALGORITHM
With the tremendous growth in chunks of text data,
there is a need to effectively summarize it to be useful.
Automatic text summarization [4] is very demanding
and non-trivial task. There have been methods
proposed which uses word and phrase frequency to
extract salient sentences from the text. Overall, there
are two different approaches for text summarization:
extraction and abstraction. Extraction works by
selecting the sentences from original text whereas
abstraction aim at modifying the original text using
advanced natural language techniques in order to
generate a new brief summary. However, extractive
summarization yields better results as compared to
abstractive summarization because abstraction face
issues such as semantic representation and natural
language generation. Here, we focus on graph based
TextRank algorithm to perform extractive
summarization. TextRank [5] is an autonomous
machine learning algorithm and is an extension of the
PageRank algorithm.
VI. PROPOSED APPROACH
In our proposed approach to build this application,
input can be taken in different modes such as editable,
text, image, audio and video also. After the input is
taken, TextRank Algorithm can be used to convert it
into summarized text. The summarized text will be
taken as output and convert into speech. Input
processing from different modes has been discussed
above. The major concern is summarizing the content
efficiently and accurately. In further section we will
discuss about improving the graph based technique i.e.
TextRank Algorithm for accurate summarization.
7
For any summarizer, intermediate representation is
done to express the main aspect of the text. It uses two
PR(Vi) = (1 - d) + d * X Vj ∈ In(Vi) PR(Vj) |
Out(Vj) (1)
In order to build a connected graph, an edge is to be
added between the two vertices which represents the
similarity between them. The similarity depends on
the words common between the two sentences which
can be calculated using the similarity function. Let Si
and Sj be two sentences where a sentence is
represented by Ni words that forms it.
Similarity(S, Si,j) = |𝑾𝒌 |𝑾𝒌∈𝑺𝒊 & 𝑾𝒌 ∈𝑺𝒋|
𝐥𝐨𝐠(|𝑺𝒊|)+𝐥𝐨𝐠(|𝑺𝒋|) (2)
Score is accredited to each sentence depending upon
the type of representation approach. In topic
representation, the score depends on how well the
sentence describes the topic whereas in case of
indicator representation, a variety of machine
learning techniques can be used in aggregating the
results. In the final step, a methodology should be
used which selects the best combination of sentences
that maximizes the importance and minimize
redundancy.
In our proposed method, TextRank algorithm is used
to find the similarity between the sentences. This
method describes the document as a connected graph
where sentences represents the vertices and an edge
indicates how similar the two sentences are. It is
based on frequency of occurrence of words so any
specific language processing is not required.
Consider an undirected graph, say G = (V, E), where
V = set of vertices.
E = set of edges.However, the title can also play an
important role in
For a given vertex Vi, let In(Vi) represent the set of
vertices pointing towards the former vertices and
Out(Vi) represents the set of vertices pointing to the
next-inline vertices. The score can be calculated for
each vertex using the formula: adding distinguishing
information to elaborate meaning of the text. The
similarity function can be improvised by computing
the correlation between each individual sentence and
title of article as well.
So, the modified similarity function between two
8 IITM Journal of Management and IT
sentences and for each individual sentence is given
by:
|𝑾𝒌|𝑾𝒌 ∈ 𝑺𝒊&𝑊𝑘 ∈ 𝑆𝑗|
𝑺𝒊𝒎𝒊𝒍𝒂𝒓𝒊𝒕𝒚 𝒔𝒊 , 𝒔𝒋 =
|𝑺𝒊| + |𝑺𝒋|
𝟐
𝑺𝒊𝒎𝒊𝒍𝒂𝒓𝒊𝒕𝒚𝒕𝒊𝒕𝒍𝒆 𝒔𝒊 , 𝒔𝒕𝒊𝒕𝒍𝒆
|𝑾𝒌|𝑾𝒌 ∈ 𝑺𝒊&𝑊𝑘 ∈ 𝑆𝒕𝒊𝒕𝒍𝒆|
=
|𝑺𝒊| + |𝑺𝒕𝒊𝒕𝒍𝒆|
𝟐
Therefore, the cumulative score for any sentence say
S1 is given by:
𝑺𝒊𝒎𝒊𝒍𝒂𝒓𝒊𝒕𝒚𝒕𝒊𝒕𝒍𝒆 𝒔𝒊 , 𝒔𝒕𝒊𝒕𝒍𝒆
𝒋=𝒏
+ 𝑺𝒊𝒎𝒊𝒍𝒂𝒓𝒊𝒕𝒚 𝑺𝒂𝒏𝒕𝒆𝒏𝒄𝒆𝒔 𝒔𝒊 , 𝒔𝒋
𝒋=𝒊,𝒋=𝟐
− 𝑺𝒊𝒎𝒊𝒍𝒂𝒓𝒊𝒕𝒚𝑺𝒆𝒏𝒕𝒂𝒏𝒄𝒆𝒔 𝒔𝟏 , 𝒔𝟏
VII. IMPLEMENTATION AND EVALUATION
We have made this application using Apache
Cordova and this application is compatible with both
android and IOS. We divided the whole approach
into three modules:
 Module 1: Uploading of image and processing
of input text using OCR approach. The text can
directly be typed in the text field. It can also be
taken in the form of audio for which we used a
button to record voice and then processing it
using audio analysis.
 Module 2: In this module, text summarization
takes place once an event is fired. Sentences are
ranked and the best sentences are picked to
make up a summary and be shown in the output
window.
 Module 3: The text in the output window is
converted into speech when a button is clicked.
We have evaluated our application by taking 3 sample
articles and evaluating the summary using ROUGE
evaluation. ROUGE [6] is the most widely used
method to evaluate the summary automatically by
correlating it to human summaries. There are various
variations of ROUGE such as ROUGE-n, ROUGE-L
and ROUGE-SU. In ROUGE-n, a series of n-grams is
elicited from the human summaries used as reference
and the candidate summary. ROUGE-L used the
longest common subsequence (LCS) approach i.e. the
longer the LCS, more will be the similarity. The metric
ROUGE-SU makes use of bi-grams as well as uni-
grams.

Results of our evaluation are shown in a given table

below:
Rouge Task Average Average Average Number
Type Name Recall Precision FScore Referenced
Summaries
ROUGE1 Sample 1.0 0.29664 0.45732 1
1
ROUGE1 Sample 1.0 0.09125 0.16841 1 Fig2. Overview of conversion of text to speech
2
ROUGE1 Sample 1.0 0.33504 0.50192 1
3 Fig 2 shows how speech is generated from text.
Natural Language Processing analyze and synthesize
Table 1. Results of summary evaluation using
natural language and speech.
ROUGE 2.0 Evaluation Toolkit
IX. CONCLUSION & FUTURE SCOPE
VIII. TEXT TO SPEECH SYNTHESIS
The paper proposes an approach to generate an
The text-to-speech synthesis [7] is the self-regulating
optimized summary taking input from various mode
conversion of a text into speech by transcribing the
such as image, audio and editable text. We also talked
text into phonetic representation and then generates
about different summarization techniques such as
the speech waveform. A text-to-speech consists of a abstraction and extraction based. In order to generate
front-end and a back-end. The front-end performs two summary we proposed modification to graph based
major operations which are text normalization and algorithm i.e. TextRank algorithm. Besides the entire
assigning phonetic transcription to each word (text-to- paragraph, score of the paragraph title is also taken
phoneme). The back-end part generates the speech account. Three sample articles are computed using
waveform. The engine is divided into modules such as ROUGE evaluation toolkit and the results are depicted
Natural Language Processing (NLP) module, Digital in table 1. However there is a scope for video analysis.
Signal Processing (DSP) module, text analysis and Since the paper discusses about taking input from
application of pronunciation rules. This can be multiple modes, video can be amongst them. Also,
developed using Java programming language. improvements can be made to make the summary
There are various techniques to preform speech algorithm more efficient and accurate. This will in turn
synthesis like Concatenative synthesis, Articulatory ensure that the generated summary has its logical
synthesis, Formant synthesis, Domain Specific meaning.
synthesis, Unit selection synthesis, Diphone synthesis, REFERENCES
HMM based Synthesis etc. Concatenative synthesis
[1] Mrunmayee Patil, Ramesh Kagalkar, ―A Review
involves concatenation of short samples of speech on Conversion of Image to Text As Well As
recording. Articulatory synthesis makes use of Speech Using Speech Detection and Image
articulatory parameters like human vocal tract to Segmentation‖, International Journal Of
generate speech. Formant Synthesis is clear at high Science and Research.
speeds. It is rule based synthesis which synthesize [2] S. Shahnawaz Ahmed, Shah Muhammed Abid
speech using acoustic rules. Domain-specific Hussain and Md. Sayeed Salam, ―A Novel
synthesis uses a simple approach of concatenating pre- Substitute for the Meter Readersin a Resource
recorded words and phrases to complete a sentence. Constrained Electricity Utility‖ IEEE Trans.
On Smart Grid, vol. 4, no. 3, Sept. 2013.
Unit selection synthesis makes use of segmented
[3] K.Kalaivani, R.Praveena, V.Anjalipriya,
records stored in database to create speech.
R.Srimeena, ―Real Time Implementation of
Image Recognition and Text to Speech
Conversion‖, International Journal of
Advanced Engineering Research and
Technology, vol.2, Sept. 2014.
[4] Mehdi Allahyari, Seyedamin Pouriyeh, Mehdi

Volume 10, Issue 1 ∙January-June 2019 9

 Assefi, Saeid Safaei, Elizabeth D. Trippe, Juan
B. Gutierrez, Krys Kochut, ―Text
Summarization Techniques: A Brief Survey‖,
28 Jul. 2017.
[5] Sunchit Sehgal, Badal Kumar, Maheshwar
Sharma, Lakshay Rampal, Ankit Chaliya, ―A
Modification to graph based approach for
extraction based Automatic Text
Summarization‖, Institute of Electrical and
Electronics Engineer.
[6] Chirantana Mallick, Ajit Kumar Das,
Madhurima Dutta, Asit Kumar Das, Apurba
Sarkar, ―Graph-based Text Summarization
Using Modified TextRank‖, Aug. 2018.
[7] Itunuoluwa Isewon, Jelili Oyelade, Olufunke
Oladipupo, ―Design and Implementation of Text
to Speech Conversion for Visually Impaired
People‖, International Journal of Applied
Information System, vol.7 no.2, Apr. 2014.
[8] Kaladharan N, ―An English Text to Speech
Conversion‖, International Journal of
Advanced Research in Computer Science and
Software Engineering, vol.5, Oct. 2015.
[9] R. Aida-Zade, C. Aril, A.M.
Sharifova, ―The Main Principles of
Text-to-Speech Synthesis System‖,
International Journal of Computer and
Information Engineering, vol. 7 no. 3,
2013.
[10] S. Brin and L. Page, ―The anatomy of
a large-scale hypertextual Web
search engine‖, Computer Networks
and ISDN systems, 30(1-7), 1998.
[11] Horacio Saggion and Thierry
Poibeau. 2013. ―Automatic text
summarization: Past, present and
future‖. In Multi-source,
Multilingual Information Extraction
and Summarization. Springer, 3–21.
[12] Dou Shen, Jian-Tao Sun, Hua Li,
Qiang Yang, and Zheng Chen. 2007.
10
―Document Summarization Using
Conditional Random Fields‖. In
IJCAI, Vol. 7. 2862–2867.
[13] Sérgio Soares, Bruno Martins, and
Pavel Calado. 2011. Extracting
biographical sentences from textual
documents. In Proceedings of the
15th Portuguese Conference on
Artificial Intelligence (EPIA 2011),
Lisbon, Portugal. 718–30.
[14] Karen Spärck Jones. 2007.
―Automatic summarizing: The state of
the art. Information Processing &
Management‖ 43, 6 (2007), 1449–
1481.
[15] Josef Steinberger, Massimo Poesio,
Mijail A Kabadjov, and Karel Ježek.
2007. Two uses of anaphora
resolution in summarization.
Information Processing &
Management 43, 6 (2007), 1663–
1680.
[16] Mark Steyvers and Tom Griffiths.
2007. Probabilistic topic models.
Handbook of latent semantic analysis
427, 7 (2007), 424–440.
[17] Fabian M Suchanek, Gjergji Kasneci,
and Gerhard Weikum. 2007. Yago: a
core of semantic knowledge. In
Proceedings of the 16th international
conference on World Wide Web.
ACM, 697–706.
[18] Simone Teufel and Marc Moens.
2002. ―Summarizing scientific
articles: experiments with relevance
and rhetorical status‖. Computational
linguistics 28, 4 (2002), 409–445.
[19] Marcia A. Bush, ―Speech and Text-Image
Processing in Documents‖, Technical Report
P92-000150, Xerox PARC, Palo Alto, CA,
November, 1992
IITM Journal of Management and IT
 Aadhaar Smart Meter: A Real-Time Bill Generator
Ramaneek Kalra1*, Vijay Rohilla2*
1*
Computer Science Department, IEEE Member, HMRITM, New Delhi, India
2*
Assistant Professor EEE Department, HMRITM, New Delhi, India
kalraramneek@ieee.org, vijay1402rohilla@gmail.com
Abstract- In present scenario, humans are at
outskirts of changing technology i.e. really helping
oneself to explore new things going around and get
familiar in short time and giving much output. In
these days, Indian Government is at extent to convert
all money in digital banks/ wallets like BHIM App,
UPI Payment etc.
With that, the technology of smart meters is emerging
in every street of Indian Societies/Colonies. Perhaps
this change initially is little bit unfamiliar since
citizens are facing problems in getting details
correctly in their hands. So, there‘s comes a need of
developing and inventing such innovative way that
government should become completely transparent to
public and long queues for submitting electricity
bills.
For this kind of problem approach, we are here to
introduce a new way to interact with all
governmental services like one we discuss above. So,
here introducing ―Aadhaar Smart Meter: A Real-
Time Bill Generator‖ (ASM). This new way of
connecting government services brings Aadhaar
server to come in existence and customer
connectivity with Smart Meter via Electricity
Department‘s Server. The whole scenario is
discussed in coming parts.
Keywords: Database, Smart Meter, Aadhaar
Database, Cloud, Gateway.
I. INTRODUCTION
A Smart Meter is a Digital Meter for analyzing the
Power Factor, Units Used, Power Used to be
displayed on a LCD screen with connectivity to both
Electricity Department‘s Server as well to Customer
Database stored in Cloud Database with respect to
customer‘s provided details like: Aadhaar_Number,
Customer_Name, Customer_Address,
Customer_Region etc. But nowadays government has
implemented Smart Meters for just displaying Power
Volume 10, Issue 1 ∙January-June 2019
Factor, Units Consumed with little bit of information
of days from installation. Due to this less
transparency between government‘s Database and
Customer‘s need we are hereby proposing idea of
―Aadhaar Smart Meter‖ which is nothing but the live
example of details discussed in above section. This
idea is purely made upon the Database Connectivity
and SMS Service provided to Customer for particular
Smart Meter they owned with. The complete
exploration of ASM will be discussed in coming
Methodology. But, for now instant let‘s discuss
coming scenario which Indian Government is trying
to establish under these needs.
Actually, since all governmental services are now
tending to connect with Aadhaar Database (The huge
Biometric Database in world). But, due to some
insecurity issue public is not ready to share their
personal Aadhaar Credentials with Government
because many are thinking of leaking of details in
some wrong hands.
For that, we have solved this issue just by connecting
Server‘s database with simple mobile number and if
Aadhaar Database is provided then the mobile
number and respective details can be retrieved easily
and quickly irrespective of huge database
connectivity.
II. PROPOSED IDEA
Before digging in the actual methodology let‘s have a
look on the basic architecture of ASM which can be
used to manipulate the required electrical factors and
to calculate the bill as month wise and set to count
zero.
Actually, smart meter consists of the readings
available for providing real-time sensors, power
outage notification and power quality monitoring.
Due to availability of these extra features, the simple
Automatic meter Reading differs in greater extent.
With that, the following symbolic block diagram can
11
clear much things about ASM and how‘s the working
and functioning occurs.
Fig 1: Symbolic Block Diagram of Smart Meter
Here, Fig 1 makes clear of important parts that plays
crucial role in making smart meter liable and
beneficial for home connection. This consists of
following components as follows:
- Data Reader: This is the initial point where the
functioning of a simple meter reading starts
which helps to store and get the data like power
outage, power units consumed which it passes to
LCD Controller. Also, the Home Connection is
connected wirelessly to Electricity Department
Server which helps to fetch electricity account
no. stored as Elect_Acc_No (say) and
corresponding customer‘s account details which
indirectly fetched from connected Aadhaar
Database.
- LCD Controller: This is the second phase which
consists of LCD Screen connected with Data
Reader which provides the data like Power
Factor, Units Consumed to the outside world.
With that, the details at real time are getting
updated over Database maintained under
Electricity Department Server.
- Automatic SMS Sender Block: This is the block
used for sending the data manipulated in 15 days
and shown over database with calculated
proposed bill amount to the user‘s registered
mobile number so that for next 15 days,
electricity can be consumed safely to ensure
green environment and this facility will too help
one to alert the late bill submission. Plus, after
12
one month the user gets online payment link over
SMS Service and thus helps customer not to go
electricity bill payment office to stand in long
queue.
- With that, the real innovation comes when
one month expires and now it‘s turn to pay the
bill using payment portal via SMS service or
using PayTm Different electricity
department‘s portals. Plus, now how to
generate manual bill that is officially sent by
electricity department. This can also be
resolved using Customer‘s android Device or
if customer is not having then he/she can
request government officials to come and
generate bill using their official android
device.
- Since, nowadays the trend of using android
device is so much at high scale of popularity
that every citizen is active online and has
tremendous amount of data consumption. So,
one can use this amazing trend as a use of
government‘s service.
- The basic idea can be used to generate
manual bill and clear out the reading shown on
particular customer‘s meter and help to print
that bill at same time.
- So, from Fig 2 one can understand the flow
of information takes place when scanning of
Meter LCD takes place to generate the
required bill from Meter itself. The
Functioning of each component is as follows:
Fig 2: Depiction of Flow of Information
- Android Application: There will be an android
IITM Journal of Management and IT
 application that scans the meter LCD screen and
thus fetches the information from electricity
department server which initializes the value in
database, that this particular meter for particular
customer_id with corresponding
Customer_name taken from Aadhaar Database to
status as true. Then, after turning status as true of
scanning done the value of units consumed and
rupees calculated inside server of electricity
department.
- Mini Printer: There is one new feature in Smart
Meter that includes mini printer which is
attached inside the meter which is used to print
the required bill which is requested by android
app by electricity dept. server and the
corresponding request is sent back by server to
Mini Printer and thus do the task of printing bill
which is sent officially by server automatically.
Plus, the data calculated over manipulation with
server is temporally stored over android app
database which secures that if any transaction
failed in between the changes will be rolled back.
- Aadhaar Database: Here, in this scenario we
used Aadhaar database since after scanning if
any customer wants to pay the bill amount at the
same time can give request from android app to
Server and thus the amount can be deduced using
connected Bank Server which is connected to
one‘s Aadhaar Database and indirectly to
Electricity Department Server and after the
transaction is done the receipt of payment status
can be printed over mini printer.
Moreover, the applications over this proposed idea
can be implemented easily under government‘s
services which will definitely reduce the manual
paper work and money exchange.
III. CONCLUSIONS
Therefore, this research idea if implemented by the
Governmental Services can give the current Nation‘s
Use of manual Electricity Bill payment to automatic
payment and thus, eliminating the long queues.
Moreover, the advantages that this proposed idea
gives to particular Indian Citizen are as follows:
- Initialization of Paperless and Cashless
Technology which will promote Digital India
Movement initiated by Indian Government.
Volume 10, Issue 1 ∙January-June 2019
- Elimination of wrong data transmitted over
Citizen‘s Electricity Bill.
- Will increase the transparency over
Government‘s Functioning and Processing.
to
IV. FUTURE WORK
This research paper represents one of the applications
that is applied but as there is still a huge number of
applications to study which can be further be taken in
contrast of study and analysis.
As this paper only includes the architectural point of
view of Smart Meter.The same proposed idea can be
studied for more enhancements/applications as
follows:
- Can be used for large-scale level study in
industrial application.
- Can be implemented by deploying the idea to
real-world machine.
- Security measures can study and apply
thoroughly to this study.
REFERENCES
[1] W. Luan, J. Peng, M. Maras, J. Lo and B.
Harapnuk, "Smart Meter Data Analytics for
Distribution Network Connectivity
Verification," in IEEE Transactions on Smart
Grid, vol. 6, no. 4, pp. 1964-1971, July 2015.
[2] G. R. Barai, S. Krishnan and B. Venkatesh,
"Smart metering and functionalities of smart
meters in smart grid a review," 2015 IEEE
Electrical Power and Energy Conference
(EPEC), London, ON, 2015, pp. 138-145.
[3] V. G. Vilas, A. Pujara, S. M. Bakre and V.
Muralidhara, "Implementation of metering
practices in smart grid," 2015 International
Conference on Smart Technologies and
Management for Computing, Communication,
Controls, Energy and Materials (ICSTM),
Chennai, 2015, pp. 484-487.
[4] R. Rashedi and H. Feroze, "Optimization of
process security in smart meter reading," 2013
Smart Grid Conference (SGC), Tehran, 2013,
pp. 150-152.
[5] S. Elakshumi and A. Ponraj, "A server based
load analysis of smart meter systems," 2017
13
 International Conference on Nextgen 2017 IEEE International Conference on Smart
Electronic Technologies: Silicon to Software Grid and Smart Cities (ICSGSC), Singapore,
(ICNETS2), Chennai, 2017, pp. 141-144. 2017, pp. 243-247.
[6] J. Russell, "Smart metering: Working towards [9] P. Bansal and A. Singh, "Smart metering in
mass roll-out," IET smart grid framework: A review," 2016 Fourth
International Conference on Parallel,
[7] Conference on Power in Unity: a Whole
Distributed and Grid Computing (PDGC),
System Approach, London, 2013, pp. 1-7.
Waknaghat, 2016, pp. 174-176.
[8] A. M. Barua and P. K. Goswami, "Smart
metering deployment scenarios in India and
implementation using RF mesh network,"

14 IITM Journal of Management and IT

 A Review on Optical Character Recognition
Archit Singhal1,and Bhoomika2
1 Department of Computer Science, The NorthCap University,
Gurugram, Haryana 122017
singhal97.archit@gmail.com, wbhoomikaw@gmail.com
Abstract--- Nowadays, using a keyboard for entering
data is the most common way but sometime it
becomes more time consuming and need lots of
energy. So, a technique was invented named Optical
Character Recognition abbreviated as OCR that
transfigures printed as well as handwritten text into
machine encoded text by electronic means. OCR has
been a topic for research for more than half a century.
It electronically and mechanically converts the
scanned images which can be handwritten,
typewritten or printed text. In general, to figure out
the characters of page, OCR compares each scanned
letter pixel by pixel to a known database of fonts and
decides onto the closest match.
Index Terms - optical character recognition,
processed, pixel, scanned document, machine
encoded text.
I. INTRODUCTION
Optical Character Recognition is a simple way of
digitizing machine-encoded text that can be searched
through and processed by a machine. It is amongst
the greatest topic of research in the field of Artificial
Intelligence, Pattern Recognition, Machine Vision
and Signal Processing. Character Recognition
techniques associate a symbolic identity within the
image of character. It extracts the significant
information and directly enters it to the database
instead of using accustomed methods of manual data
insertion.
This technique was firstly introduced for two main
reasons i.e., expanding telegraphy and helping blinds
to get education. Emanuel Goldberg and Edmund
Fourier d‘Albe were first to work on this technique in
1914. They built a machine that firstly scan the
characters and later convert them into standard
telegraph code and another device named Optophone
that produced specific tone around specific letters or
characters. These machines were patented in 1931
and now they are acquired by IBM.
Volume 10, Issue 1 ∙January-June 2019
OCR in general is classified into two types: Off-line
and On-line. This technique of Off-line recognition is
used for automated conversion of text into codes of
letter which are usable by computer and applications
developed for text processing. But, it is more
difficult, as different people have different
handwriting font. Whereas, On-line recognition deals
with a continuous input of data stream that comes
from a transducer when the user types or writes.
Fig. 1. Types of Optical Character Recognition
II. LITERATURE REVIEW
Research Paper Statement: A technique named
Optical Character Recognition abbreviated as OCR
which is in its development stage has proven to be
much beneficial for transfiguring any kind of
handwritten material to digitized form.
This paper reviews the work done by various authors
in the field of exploring Optical Character
Recognition. Prior studies have identified various
steps involved from pre-processing the image to give
the final Digitized output. Also, the paper has
depicted various fields where this technology has
been efficiently implemented. But as it is in its
development stage, it also faces few challenges in
giving the best required output. Integrating the
concept and theories provided in paper to various
15
other fields with more advanced development will
show much better results surpassing 99%.
Additionally, material learned in paper can be applied
to benefit the community through a variety of
tangible services
III. PHASES OF OCR
The whole procedure of transfiguring the handwritten
as well as printed text into machine encoded text is
broadly divided into four simple phases:
Fig. 2. Phases of Optical Character Recognition
A. Pre Processing
In this phase, the image is scanned starting from the
top to the end & converted into gray level image,
which is then converted into digital binary image.
This process is sometimes termed as Digitization of
image or Binarization. We use various scanners for
this phase and last digital image then goes to the next
step.
B. Character Extraction
The pre-processed image of the previous step serves
as the input of the following step. In this step, each
single character of the image is recognized. Also, the
image is converted into the window size from the
normalized form in this step.
C. Segmentation
This is the most important step of the whole process
as it removes most of the noises from the images for
more understandable form. It segments different
characters into various zones i.e., upper, middle and
lower zone.
Segmenting is difficult in offline recognition because
of variability in paragraph, words of line and
16
character‘s shape (slant, skew, curve, etc).
Sometimes this difficulty arises due to overlapping of
one or more characters also.
D. Post Processing
In this phase, features of every character is enhanced
and extracted. In this phase, we can classify every
character in a unique way. Feature of individual
character is enhanced. Also, if there are some
unrecognized characters found, they are also given
some meaning. Extra templates can also be added in
this phase for providing a wide range of compatibility
checking in the systems database.
IV. APPLICATIONS OF OCR
Optical Character Recognition transfigures the
scanned documents into more than an image file;
rather, turn it into a readable as well as searchable
text-file that can be processed by computers.
OCR is a field with enormous application in number
of industries such as legal, healthcare, banking,
education, etc.
A. Banking
In banks, cheques are processed using OCR without
any kind of involvement by humans. The inserted
cheque in the device is searched & scanned for the
writing in various fields and the amount is transferred
to the following payer. This whole process reduces
the overall cheque process time.
B. Healthcare
The use of OCR technology has also been increased
in Healthcare industry to process paperwork. In the
healthcare industry, they deal with the huge amount
of forms like patient details, medical-history,
insurance forms, etc., so, in order to reduce energy
and time, this technology is used.
C. Legal Industry
Documents are scanned; information is extracted and
automatically entered into the database to save space.
The time consuming task that requires the need to
search for information through boxes is also
eliminated. This helps in locating any of the specific
text/document easily. It has also helped legal industry
to have easy, fast and readily available access to a
huge library of documents.
IITM Journal of Management and IT
D. Invoice Imaging
It is important to maintain a track of financial records
to avoid any piles of backlog payments. Among other
processes, OCR helps in simplifying collection and
analysis of large sets of data. It is also used to decrypt
the large amount of information stored in the Digital
code like Bar & QR codes.
E. Other Fields
OCR is extensively used in many other different
areas also, like:
CAPTCHA- to prevent hacking;
Digital Libraries- sharing of digital teaching material;
Optical Music Recognition- to extract information
from images;
Automatic Number Recognition- to identify vehicle
registration plates;
Handwriting Recognition; Education;
Maestro Recognition Server; Trapeze.
V. CHALLENGES OF OCR
The techniques of OCR require images of high
resolution which have basic structural property
differentiating text and background to get high
accuracy in character recognition. Image generation
plays an important role in determining the accuracy
and successful recognition. The image generated by
scanners gives high performance and accuracy while
images generated by cameras have numerous errors
due to surroundings & factors related to camera.
These errors are clarified as follows
A. Tilting
The image of documents obtained by scanners is
parallel and in line to plane of sensor which is not
observed in image taken by hand-held devices. The
text nearer to camera seems a little large while the
text distant appears smaller which causes perspective
distortion resulting in tilted pictures. The perspective
intolerant recognizer causes lower recognition rate
B. Scene Complexity
The image taken by portable devices generally
involves various no of artificial objects such as
building, symbols, cars etc. considering a regular
Volume 10, Issue 1 ∙January-June 2019
environment. The object detected makes the text
recognition in processed image very challenging as
the appearances and structure of these objects is
comparative to the text present around it. Text itself
is easily present in any form to encourage
decipherability making the scene of segregating text
from non-text very intricate
C. Conditions of uneven lighting
The major challenge for OCR is degradation of text
quality due to uneven lighting and shadows when
images are taken in a natural environment. This
results in poor detection, recognition and detection of
text. This case of shadows and uneven lightning
differentiates between images taken by the camera
and scanners. The lack of proper lighting makes
scanned images more preferred than images
processed by camera for their better text and
characteristics quality. But these problems of lighting
can be solved by using flash in camera which also
lead to some new challenges.
D. Skewness
In OCR technique, the POV for the image used as
input might change when the image is taken from
camera or any hand-held devices which is not
applicable for scanner image input. As a result,
change of point of view leads to skewing which
provides a great degree of poor results when image is
processed. To overcome this problem, many deskew
techniques are available such as RAST algorithm,
Methods of Fourier transformation, projectile profile
etc.
E. Aspect ratio
The image of documents obtained by scanners is
parallel and in line to plane of sensor which is not
observed in image taken by hand-held devices. The
text nearer to camera seems a little large while the
text distant appears smaller which causes perspective
distortion resulting in tilted pictures. The perspective
intolerant recognizer causes lower recognition rate
and accuracy. The new latest cell phones can easily
recognize if the portable device is tilted and then can
prohibit clients to click images. This all detection and
prohibition is done with the help of orientation
sensors which also allows camera to align the text in
plane of form resulting in greater degree of evenness
17
F. Wraping
One character on another can be another challenge
for OCR to be precise. This situation arises when
images are scanned using flatbed scanners which
procured text on picture of the twisted text.
For panacea, a technique called dewraping was
introduced by Ulges et al, which treat these texts the
same way as they are equally distant and parallel to
each other.
G. Multilingual Environments
Latin language contains a large number of symbols,
character classes as it is composed of many other
languages like, Japanese, Korean and Chinese.
Arabic languages have characters with different
writing shapes. Hindi language contains syllables
which are made up of combining different shapes.
Therefore, multilingual become a primary problem in
OCR.
H. Fonts
Using different styles and fonts for different
characters can make them overlap with each other
and thus making OCR difficult. It is difficult to
perform precisely accurate recognition due to
various within subclasses variations and forming
pattern sub-spaces.
VI. SCOPE OF OCR
Nowadays, a diverse collection of OCR systems are
available but still we face many problems therein.
The collections of OCR systems were earlier
categorized into two groups. The first group includes
machines that are specially designed to recognize
specific set of problems which are mostly hard-wired
so become little expensive and also decreases
throughput rates. The second kind of group includes
all software based techniques which involve
computer or low cost scanner. Due to advancements
in recent technologies, the second group of OCR
systems is much more cost effective with high
throughput; however, there are few limitations in
these systems regarding speed and reading set of
characters. They read the data line-by-line and
transfer it to the OCR software systems. OCR
systems are now categorized into five different
groups based on character sets, namely, fixed-font,
multi-font, omni-font, constraint handwriting and
18
scripts.
Imperfections and irregularities in OCR systems are
mainly due to problems occurred during scanning
phase which usually result in inappropriate text or
character. These irregularities often result in the
misinterpretation among text and graphics or among
text and noise. Perfectly scanned character can also
cause imperfections due to characters with the same
shapes and features, which makes the system difficult
to exactly recognize the character. With this we can
conclude that precision of OCR totally depend on the
quality of input it takes.
Although we have seen a lot of improvement and
advancement in OCR in recent years, from reading
only a limited set of characters to reading characters
with different fonts and styles and further reading
handwritten text. In the coming years, seeing
advancement in technology, one can predict that
OCR can have much more potential and recognition
in following years.
VII. CONCLUSION
This paper tells about a field in Artificial Intelligence
i.e., Optical Character Recognition; its types, its
whole process and its applications in different areas.
Optical Character Recognition or OCR has made
scanned documents to become more than an image
file, rather, turning them into a fully searchable,
readable as well as editable text file that can be
processed by computers.
The research in this area has been going from more
than half of the century and the aftermath have been
striking with successful recognition rates surpass
99% with notable advancement accomplishing for
cursive handwritten character recognition. Further,
the research in this area aims for more improvement
and scope.
REFERENCES
[1] G.Vamvakas, B.Gatos, N. Stamatopoulos, and
S.J.Perantonis: A Complete Optical Character
Recognition Methodology for Historical
Documents 2007.
[2] Karez Abdulwahhab Hamad, Mehmet Kaya: A
Detailed Analysis of Optical Character
Recognition Technology. International Journal
of Applied Mathematics, Electronics and
IITM Journal of Management and IT
 Computers Advanced Technology and Science [12]
ISSN: 2147-8228.
[3] Combination of Document Image Binarization
Techniques 2011. [13]
[4] International Conference on Document
Analysis and Recognition 2015.
[14]
[5] D-Lib Magazine: How Good Can It Get?
Analysing and Improvising of OCR Accuracy
in Large Scale Historic Newspaper
Digitisation Programs.
[6] Raghuraj Singh1 , C. S. Yadav2 , Prabhat
Verma3 , Vibhash Yadav4: Optical Character
Recognition (OCR) for Printed Devnagari
[15]
Script Using Artificial Neural Network
[7] B. B. Chaudhary and U. Pal: OCR Error
Detection and Correction of an Inflectional
Indian Language Script, Pattern Recognition,
IEEE Proceeding of 13th International
[16]
Conference on Image Processing 2002.
[8] Agia Paraskevi, Athens: Institute of
Informatics and Telecommunications,
National Center for Scientific Research:
Demokritos, GR-153 10.
[17]
[9] Optical Character Recognition System Using
BP Algorithm: Department of Industrial
Systems and Information Engineering, Korea
[18]
University, Sungbuk-gu Anam-dong 5 Ga 1,
Seoul 136-701, South Korea.
[10] Dholakia, K., A Survey on Handwritten
[19]
Character Recognition Techniques for various
Indian
[11] Languages, International Journal of Computer
Applications, 115(1), pp 17–21, 2015.
Volume 10, Issue 1 ∙January-June 2019
Yu, F. T. S., Jutamulia, S. (Editors): Optical
Pattern Recognition, Cambridge University
Press, 1998.
Mantas, J.: An Overview of Character
Recognition Methodologies, Pattern
Recognition, 19(6), pp 425–430, 1986.
Pradeep J, Srinivasan E, Himavathi S.:
Diagonal based feature extraction for
handwritten character recognition system
using neural network. InElectronics Computer
Technology (ICECT), 2011 3rd International
Conference on 2011 Apr 8 (Vol. 4, pp. 364-
368). IEEE.
Bishnu A, Bhattacharya BB, Kundu MK,
Murthy CA, Acharya T.: A pipeline
architecture for computing the Euler number
of a binary image. Journal of Systems
Architecture. 2005 Aug 31;51(8):470-87.
Verma R, Ali DJ. A-Survey of Feature
Extraction and Classification Techniques in
OCR Systems. International Journal of
Computer Applications & Information
Technology. 2012 Nov;1(3).
Md. Anwar Hossain, Optical Character
Recognition based on Template
Matching(2018)
C. Vasantha Lakshmi1 and C. Patvardhan ―An
optical character recognition system for
printed
Telugu text , Pattern Analysis &
Applications‖, Category, Theoretical
Advances, Volume 7, Number 2 / July, 2004
Pages 190-204
19
 Prediction of Heart Attack Using Machine Learning
Akshit Bhardwaj1, Ayush Kundra2, Bhavya Gandhi3,
Sumit Kumar4, Arvind Rehalia5, Manoj Gupta6
1,2,3,4,5,6
Department of Instrumentation & Control Engineering
Bharati Vidyapeeth’s College of Engineering Delhi-110063
Abstract- Cardiovascular diseases are one of the
biggest reasons for death of millions of people
around the world only second to cancer. A heart
attack occurs when a blood clot blocks the blood flow
to a part of the heart. In case this blood clot cuts off
the blood flow entirely, the part of the heart muscle
begins to die as a result. Going by the statistics, a
heart problem can gradually start between the age of
40-50 for people with unhealthy diet and bad lifestyle
choices. So, an early prognosis can really make a
huge difference in their lives by motivating them
towards a healthy and active life. By changing their
lifestyle and diet this risk can be controlled. This
Project intends to pinpoint the most relevant/risk
factors of heart disease as well as predict the overall
risk using machine learning. The machine learning
model predicts the likelihood of patients getting a
heart disease trained on dataset of other individuals.
As the result, the probability of getting a heart
disease based on current lifestyle and diet is
calculated. The model was trained with Framingham
heart study dataset.
Keywords:-Heart Disease, Machine Learning, logistic
regression, Cross-validation
I. INTRODUCTION
Machine Learning is one of the most rapidly evolving
fields of AI which is used in many areas of life,
primarily in the healthcare field. It has a great value
in the healthcare field since it is an intelligent tool to
analyse data, and the medical field is rich with data.
In the past few years, numerous amounts of data were
collected and stored because of the digital revolution.
Monitoring and other data collection devices are
available in modern hospitals and are being used
every day, and abundant amounts of data are being
gathered. It is very hard or even impossible for
humans to derive useful information from these
massive amounts of data that is why machine
learning is widely used nowadays to analyse these
20
data and diagnose problems in the healthcare field.
A simplified explanation of what the machine
learning algorithms would do is, it will learn from
previously diagnosed cases of patients. A heart
Problem must be diagnosed quickly, efficiently and
correctly in order to save lives. Due to this
Researchers are interested in predicting risk of heart
disease and they created different heart risk
prediction systems using various machine learning
techniques. The presence of missing and outlier data
in the training set often hampers the performance of a
model and leads to inaccurate predictions. So, it is
critical to treat missing and outlier values before
making a prediction.
1. Missing: For continuous variable, we can find
the missing values using isnull() function. Mean
of the data can also help identify. We can also
write an algorithm to predict the missing
variables.
2. Outlier: We can use a scatter plot to identify and
as per need, delete the data, perform
transformation, binning, Imputation or any other
method. Diagnosis of heart disease using K-fold
cross validation method will be used to evaluate
the data and the result would be more accurate.
80% data of the patients will be used for training
and 20% for testing. Parameter tuning is also
necessary if accuracy is not close to 80 %.
Logistic regression is the suitable regression
analysis to perform when the dependent variable
„y‟ is either 0 or 1. Like all regression models,
the logistic regression is a type of predictive
analysis. Logistic regression is used to explain
the relationship between dependent variable
usually „y‟ and various nominal, ordinal,
interval or ratio-level independent variables
(array of x features).
3. Features have higher odds of explaining the
variance in the dataset. Thus, giving improved
IITM Journal of Management and IT
 model accuracy. The dependent variable or target
variable should be binary/dichotomous in nature.
4. There should be no missing value/outlier in the
data, which can be assessed by or converting the
continuous values to standardized scores.
5. There shouldn‟t be a high correlation among the
predictors. This can be interpreted by a
correlation matrix among the predictors. The
regression analysis is the task of estimating the
log of odds ratio of an event.
6. Statistical tools easily allow us to perform the
analysis for better results. Adding independent
variables to a logistic regression model will
always increase the amount of variance which
would reduce the accuracy.
II. LITERATURE SURVEY
In the past few years, a lot of projects related to a
heart disease risk prediction have been developed.
Work carried out by various researchers in the field
of medical diagnosis using machine learning analysis
has been discussed in this section of the paper.
Das et al [1] worked on Deep Learning technology to
find odds ratio or the prediction values from various
different analytical models and with K-nearest
neighbours got 89.00% classification accuracy on the
Cleveland dataset for heart study.
Anbarasi et al [2] used 3 binary classifiers such as
Naive Bayes, K-means clustering and Random forest
for heart attack risk prediction using 13 features and
then applied feature engineering for algorithm tuning
and got great prediction results. They discovered that
Random forest outperforms the other two binary
Classifiers with an accuracy of 99.2% for binary
classification. The accuracy of K-means was 88.3%
and Naive Bayes was about 96.5%.
Zhang et al [3] suggested an effective heart attack
prediction model using Support Vector Machine
(SVM) algorithm. In this, Principal Component
Analysis was applied to retrieve the imperative
features and different kernel functions. The highest
accuracy was found with Radial Basis Function. To
get the optimum parameter values, Grid search in
SVM was brought to use and optimum values were
found. The maximum classification accuracy touched
about 88.64%.
Volume 10, Issue 1 ∙January-June 2019
Vadicherla and Sonawane et al [4] proposed a
minimal optimization technique of SVM for coronary
heart disease prediction. This technique helps in
training of SVM by looking for the optimal values
during training period. This shows minimal
optimization technique provided good results even on
a big dataset and execution time was also reduced
significantly.
Elshazly et al [5] presented a classifier called Genetic
algorithm SVM method Bio-Medical diagnosis in
which 18 features were reduced to 6 features via
dimensionality reduction. Different kernel functions
were put up for use and performance was compared
in terms of measures like accuracy, precision, recall,
area under curve and f1 score. The results showed
that linear SVM classifier managed an 83.10%
accuracy with 82.60% true positive rate, 84.90%
AUC and 82.70% f1 score.
III. PROPOSED SYSTEM
The machine learning technique used for the
prediction of heart attack is Logistic Regression. The
Dataset used for analysis and training is taken from
Framingham Heart Study. It is a long-term, ongoing
cardiovascular cohort study of residents of the city of
Framingham, Massachusetts. The study began in
1948 with 5,209 adult subjects from Framingham and
is now on its third generation of participants. This
Dataset can be found at
www.framinghamheartstudy.org
This research intends to pinpoint the relevant/risk
factors of heart disease as well as predict the overall
risk using logistic regression. Mathematically, we can
say that the logistic regression uses a Sigmoid
function. Logistic regression values are categorical
unlike linear regressions which are continuous. The
logistic function is a sigmoid function, which takes
any real value between 0 and 1. Mathematically,
S(y) = 1 / 1 + e-y
Or p (probability) = 1 / 1 + e - (β0 + β1x) Consider
„y‟ as a linear function in a regression analysis,
y+= β0 + β1x
Putting y in s(y) sigmoid function, it becomes a
logistic function after solving, logit(p) = log(p/(1−p))
=
β0+β1∗ Sexmale+β2∗ age+β3∗ cigsPeryear+β4∗ tot
21
Chol+β5∗ BP+β6∗ heartrate+β7∗ BMI
Here, β0 = Regression Constant p/1-p = odds ratio of
the event βk = coefficient of x (predictors) where k =
1,2...
Fig 3.1: Block Diagram of the Model
Diagnosis of heart disease using K-fold cross
validation method will be used to evaluate the data
and the result would be more accurate. 80% data of
the patients will be used for training and 20% for
testing. This shows us better accuracy score and takes
the least amount of execution time.
IV. IMPLEMENTATION
a. Feature Engineering
testing datasets using cross-validation.
d. Algorithm Tuning
The aim of parameter tuning is to find the best value
for each parameter to improve the accuracy of the
ML model. To tune them, we must have a good
knowledge about their impact on the output. We can
repeat this process for other algorithms.
e. Results and Analysis
The machine learning model and implementation of a
heart disease risk predictor for patients with risk of
future heart disease using a logistic regression
algorithm was successful. Accuracy was calculated as
the ratio of total number of correct predictions to the
total number of predicted outputs. And written as
(TP+FN)/(TP+TN+FP+FN) Where, TP = True
positive, TN = True negative, FN = False negative,
FP = False positive.
Cost function f(x) is the sum of the squares of the
difference between the actual value and the predicted
value and iterations are the number of times the code
will be executed to obtain the lowest value of f(x).
Here, the Global minimum was calculated using
gradient descent.

Feature Engineering aids in extracting information

from the current data. Information is extracted in
form of new features. These features might have a
higher chance of explaining the (variance) in the data.
Thus, giving better model accuracy.
b. Feature transformation
There are various cases where feature transformation
is required
• Changing the variable from original scale to a
scale between 0 and 1. This is known as
normalization.
• Some algorithms work well with normal
distribution. So, we have to remove skewness of
Fig 5.1: cost function
the variables. There are techniques such as
square root or log transformation to remove We also used a Receiver Operating Characteristic
skewness. curve which tells us about how good the model can
c. Feature Selection differentiate between TN and TP.
Feature Selection is a process of looking out for the If the area under the ROC curve is more than 0.8 or it
best attributes which better define the relationship of is bent more towards the left, the model will give
an independent variable with target or dependent better results and predictions would be precise.
variable. The data is sliced into x and y training-
Accuracy was 87% during testing as calculated by

22 IITM Journal of Management and IT

the expression for the model on predicted and test
values.
Details of 5 individuals and their corresponding
probabilities. Threshold is 0.2, that is if probability
risk > 0.2 would mean there is chance of a heart
attack in future
Table 5.1: Medical data of five individuals
In Table 5.1 we can see the parameters use in the
dataset for the analysis. Here several parameters are
considered which include controllable and non-
controllable factors. The parameters are Age, Sex i.e.
male or female, Cigarettes count per month, Total
cholesterol, Systolic and diastolic blood pressure,
Body Mass Index and Heart rate of the individual.
Table 5.2: Probabilities of the individuals
In Table 5.2 we can see the result of the system. Here
the probability of the individual getting a heart attack
is calculated based on parameters discussed in table
5.1. Here the percentage of NOT getting heart attack
is depicted in first column followed by the percentage
of getting heart attack in second column.
V. CONCLUSION
Men seem to be more susceptible to heart disease
than women. Every 1 in 4 men [8] are likely to have
heart disease whereas in case of women every 1 out
of 5 women is likely to have heart disease [9].
Increase in Age, number of cigarettes and systolic
Volume 10, Issue 1 ∙January-June 2019
Blood Pressure also show increasing odds of having
heart disease. Interestingly total cholesterol shows no
significant change in the odds of CHD. This could be
due to the presence of 'good cholesterol‟ in the total
cholesterol reading. The early prognosis of
cardiovascular diseases would aid in making better
decisions on lifestyle changes in high risk patients
and in turn reduce any future heart problems.
VI. FUTURE SCOPE OF THE PROJECT
At some point in future, the machine learning model
will make use of a larger training dataset, possibly
more than a million different data points maintained
in electronic health record system. Although it would
be a huge leap in terms of computational power and
software sophistication but a system that will work
on artificial intelligence might allow the medical
practitioner to decide the best suited treatment for the
concerned patient as soon as possible [6]. A software
API can be developed to enable health websites and
apps to provide access to the patients free of cost.
The probability prediction would be performed with
zero or virtually no delay in processing [7].
REFERENCES
[1] Das, Turkoglu, and Sengur, ―Efficient
diagnosis of heart disease via machine learning
models‖, Expert systems with applications,
2009.
[2] Vanisree and Jyothi, ―Decision Support model
for Heart Disease prognosis based on early
signs of 8–51 patients using binary
classification‖, International Journal of
Computer Applications, 2011.
[3] Y. Zhang, ―Studies on application of
Support Vector Machines in coronary heart
disease prediction model‖, Electromagnetic
Field Problems and Applications, Sixth
International Conference (ICEF), IEEE 2012.
[4] Vadicherla and Sonawane, ―Decision Support
for coronary Heart Disease analysis Based on
[5] Minimal Optimization technique‖,
International Journal of Engineering Sciences
and Emerging Technologies, 2013.
[6] H. Elshazly, Hassanien and Elkorany, ―Lymph
diseases prediction based on support vector
machine algorithm‖, Computer Engineering &
23
 Systems 9th International Conference (ICCES), [9] Xu, JQ, Murphy, SL., Kochanek, KD, Bastian,
2014. BA. Deaths: Final data for 2013. National Vital
Statistics Report. 2016.
[7] Bhupender Kumar & Yogesh Paul, "Medical
Applications of Machine Learning CDC. Million Hearts™: strategies to reduce the
Algorithms", UIET, Kurukshetra University, prevalence of leading cardiovascular disease risk
2016. factors. United States, 2011. MMWR 2011.
[8] ]Ram Avatar & Vineet Kumar, "Deep Learning
in healthcare", UIET, Kurukshetra University,
2018.

24 IITM Journal of Management and IT

 Detection and Prevention Schemes in Mobile Ad hoc
Networks
Jeelani1, Subodh Kumar Sharma2, Pankaj Kumar Varshney3
1
Scholar, Mangalayatan University, Aligarh, India
2
Associate Professor, Mangalayatan University Aligarh, India
3
Associate Professor, Department of Computer Science, IITM Janakpuri, New Delhi, India
jeelani.jee@gmail.com, Subodh.sharma@mangalayatan.edu.in
pankaj.surir@gmail.com,

Abstract- Wireless Sensor Network (WSN) has wide
range of application areas such as health care,
military and industry for real time event detection.
The sensing capability of a Wireless Sensor Network
(WSN) requires sensor node as a network of it. But
these nodes are constrained in terms of size, energy,
memory, processing power. These nodes sense
environmental data perform limited processing and
communicate over short distances. As the
applications of wireless sensor networks are
continuously growing also the need for security
mechanisms is increasing day by day. It is very
essential to save WSNs from malevolent attacks in
unfriendly situations. Such systems require security
design because of different restrictions of assets and
the noticeable attributes of a remote sensor arrange
which is a impressive test. This article is a broad
survey about issues of WSNs security, which
inspected as of late by analysts and a superior
comprehension of future bearings for WSN security.
Keywords- Mobile Ad hoc Network, Wireless Sensor
Network, Denial of service.
I. INTRODUCTION
Wireless sensor networks as a part of MANET
consist of large number of tiny sensor nodes that
continuously monitors environmental conditions.
Wireless Sensor Networks are a collection of
thousands of sensor nodes that are self-organized and
are capable of wireless communication. But these
nodes are constrained in terms of size, energy,
memory, processing power [23]. These nodes sense
environmental data perform limited processing and
communicate over short distances. As the
applications of wireless sensor networks are
continuously growing also the need for security
mechanisms is increasing day by day. Wireless
Sensor Networks may interact with sensitive data or
usually these networks operate in hostile, unattended
environments, it is necessary to address these security
concerns. Security challenges of sensor networks are
different from traditional networks due to many
constraints of these networks. Moreover when we
look at the applications of WSNs, there are many
applications areas, e.g., battlefield awareness, traffic
monitoring system etc. In which security of
information remains as an important issue. Providing
security to a WSN is a nontrivial problem. Security
mechanisms which are applicable to wired or other
ad-hoc networks are not suitable for WSN. There are
many reasons behind it and we discuss those in the
subsequent sections. Though there are varieties of
challenges in sensor networks, here we focus on
different security issues and possible remedies of
those.
A. Security Requirements in Wireless Sensor
Networks
The main security requirements that each WSN has
to fulfill are as follows.
Confidentiality: Secrecy of message transmitted
between nodes should be maintained properly. For
that important segments of message should be
encrypted. In some cases even the two end points are
also hidden. In some dynamic systems where nodes
keep on joining and leaving the network, forward and
backward secrecy needs to be maintained. Forward
Secrecy means that nodes leaving the network may
not be able to access future transmissions on the
network after leaving the network and Backward
Secrecy means that new nodes may not be able to
access past transmissions before their joining the

Volume 10, Issue 1 ∙January-June 2019 25

network. These phenomenons are needed to maintain jamming, tampering, exhaustion, flooding and so on
confidentiality of data in wireless sensor networks [3].
[23].
DENIAL-OF-SERVICE DEFENSE IN WSNs [22]
Authenticity: For preparation the security of
communicating node‘s identities, authenticity is vital.
Any node must verify even if an accepted message
comes from a true sender. In the absence of
authentication, attackers without difficulty are able to
extend wrong data into the wireless sensor networks.
Generally, for authentication the origin of a message,
an annexed message authentication code possibly
employed [22].
Integrity: Integrity should be prepared to assure that
attackers cannot change the transmitted messages.
Attackers are able to establish interference packets to
modify their polarities. In addition before forwarding
them a malicious routing node can alter significant
data in packets. To find random errors throughout
Hello Flood attack: HELLO flood attack is one of the
packet transmissions as a cyclic redundancy
active attacks that flood the HELLO packets in the
checksum (CRC) employed for detecting them,
network. In wireless sensor network attacker transmit
similarly keyed checksum, for example a MAC use to
the packets from source node to destination
secure packets against changes [22].
publicizing the packets as cluster head. All sensor
Availability: WSN services should always be nodes will select these packets and send join packet
available in spite of all the resource depletion attacks into it, thinking that the attacker is their neighbor and
that may occur on the system. So our network should the entire network will be in confusion. In wireless
be resistant to such attacks [23]. sensor network the sensor nodes are deploy in normal
orchestrated region. And the data is transmitted from
Non-Repudiation: Neither the sender nor the receiver
source node to destination through intermediate
should be able to deny that the message is sent by
nodes. Sensor nodes does not distinguish that the
him. For that message can be digitally signed by both
enemy node is not their neighbour nodes. So as a
the sender and the receive [23].
result network is spoofed by the attacker [3].
B. Attacks on Wireless Sensor Networks
Attacks on Information during transmission: The
Since wireless sensor networks operate in unsafe most dangerous attack in WSN are on information
environment these are vulnerable to several types of that is being transmitted between nodes because that
attacks. information is susceptible to eavesdropping,
Denial of service attack: Denial-of-Service attack is injection, modification.
the serious attack as it consumes the network Traffic analysis attack can also be performed because
resources like energy, bandwidth and power. Denial attacker may be able to get to know about the layout
of service attack floods access amount of unnecessary of the network and can damage the busiest portions
packets in the network and affects the overall of the network to perform greatest damage [23].
performance of the network. If there is only single
Replicating a Node Attack: The attacker may insert a
attacker in the network then this is DoS attack and if
new node into the sensor network which can be a
there are multiple attackers then this is known as
clone to an pre-existing node. This new cloned node
Distributed Denial of Service (DDoS) attack. Denial
can transmit useful information to the attacker. This
of service attack is multilayer attack. In WSN there
node replication attack is most dangerous when the
are numerous DOS attacks on different layers like
cloned node is some base station. So base stations

26 IITM Journal of Management and IT

needs to be deployed in secure locations [23].
C. Routing Attack
The attacks that affect the routing protocol of
wireless sensor network are as follows:
Selective Forwarding: In Selective Forwarding attack
the malicious node may drop certain packets and
transmit the rest. If it drops all the packets then it is a
Black Hole attack. But if it forwards selective packets
then is selective forwarding attack. The effectiveness
of the attack depends on how close is the malicious
node to the base station because then maximum
traffic will go through it [23].
Sinkhole Attacks: Sinkhole Attack is to attract
maximum traffic through malicious node which is
placed somewhere near the base station. If the sensor
network has one main base station then this attack
can be dangerous [23].
Sybil attack: In Sybil attack one node presents
multiple identities in the network that may mislead
nodes in the network. Sybil attacks can be used
against topology maintenance and routing algorithms
[23].
Wormhole Attack: In Wormhole attack just like
Sinkhole attack the attacker sitting closer to base
station may tunnel the traffic to a low-latency link
thus disrupting the traffic [23].
II. LITREATURE REVIEW
The present literature review is based on the research
work entitled ―Security attacks detection and
prevention schemes in wireless sensor networks‖. For
reviewing of literature, the researcher has gathered
more of articles as a secondary source of data from
which, selected material or articles related to the
researcher topic in order to acquire depth knowledge
on the related topic and completed in the past days.
After reviewing of previous research articles, the
researcher summarized the reviewed literature and to
end up with, a research gap has been introduced.
Guechari, M et al.[1] presented an experimental study
on dynamic approach for detecting Denial of Service
(DoS) attacks in cluster-based sensor networks. That
method is based on the election of controller nodes
called cNodes which observe and report DoS attack
activities. Each cluster contains cNodes and normal
Volume 10, Issue 1 ∙January-June 2019
sensor nodes. The role of a cNode is to analyze traffic
and to send back a warning to the cluster head if any
abnormal traffic is detected.The election of these
cNodes is dynamic, and done periodically and based
on a Multiplicative Linear-Congruential Generator
(MLCG).
Rolla P. and Kuar M.[2] concluded an experimental
study on time allocation based request forwarding
window technique to detect and prevent DoS (Denial
of Service) attack. DoS attacks are flooding access
amount of packets in the network that consumes the
energy of the network. In Profile based Protection
Scheme (PPS), the behavior of all the nodes deployed
in the network are observed.
Patil S. and Choudhari S.[3] analyzed that Denial-of-
Service (DoS) attack is most popular attack in sensor
network. Attack prevention techniques such as fuzzy
Q-learning algorithm, Dynamic Source Routing
(UDSR), Secure Auction-based Routing (SAR), have
been used by author against DoS attacks. They
proposed cooperative immune system is an
enhancement to the existing immune system, CO-
FAIS, which will improve the accuracy of the
system. In the stands they have reduced the false
alarm rate.
Naik S. and Shekokar N.[4] designed and
experimental program on defending the mechanism
against the denial of sleep attack. This solution is an
effective method for preventing this attack as all the
nodes sending the synchronization messages will be
validated before those messages are accepted and
rejected if the node is not validated. The attacker
node cannot replay the sleep synchronization signal
again as its sleep schedule will not be accepted
without authentication.
Chaudhary S. and Thanvi P.[5] concluded an
experimental study on modified variant of Ad-hoc On
Demand Distance Vector (AODV) protocol to
analyze the effect of Dos attack on system
performance and later apply the prevention scheme to
analyze the change in network performance.
Researcher have used scenario in an experimental
which are Topology scenario for 80 nodes,
simulation in progress, average End-to-End Delay,
Throughput analysis, Packet ratio and Packet drop
analysis. For successful attack detection various
methods have been proposed over time. The
27
technique used by researcher based on comparison on
RREP sequence number of packet received by the
sender from its neighbors broadcasting the
availability of fresher or shorter routes.
S. Fouchal et al.[6] carried out a parametric study to a
novel approach to detect denial attacks in Wireless
Sensor Networks. This is approach based on a
recursive clustering. They have approved our
proposition with two clustering algorithms on 100-
sensor networks. In fact, they used the LEACH (low
energy algorithm adaptive clustering hierarchy)and
FFUCA (Fast and flexible unsupervised clustering
algorithm) algorithms. The results are convincing in
terms of the detection of the groups. In addition, the
use of FFUCA induces a better management of
energy and thus a longer network lifetime.
Mansouri D et al.[7] presented study on a method for
detecting and preventing Denial of Service attacks in
WSNs. The detection method they have considered is
based on using special control nodes which are
monitoring the throughput of traffic in clusters.
Control nodes (Cluster Heads) are elected by using
recursively LEACH clustering algorithm. We have
presented by means of a set of experimentation, using
Simevents simulator. The numerical results obtained
show that our approach gives significant results in
term of detection rate and time detection.
Kiss I., Haller P. and Beres A.[8] proposed a
clustering based approach for detecting the influences
of cyber-attacks, especially those of denial of service
(DoS) attacks, in the observed of the system.
Proposed approach is presented in contrast with TEP
(Tennessee Eastman challenge process) therefore
several scenario of DoS attack are experimented to
validate the effectiveness of the method. Researcher
have used the SCADA (Supervisory Control And
Data Acquisition) simulator this approach.
Ballarini P., Mokdad L. and Monnennt Q.[9]
proposed a dynamic cNodes displacement schema
according to which cNodes are periodically elected
among ordinary nodes of each atomic cluster. Such a
solution results in a better energy balance while
maintaining good detection coverage. They analyze
the tradeoffs between static and dynamic solutions by
means of two complementary approaches: through
simulation with the NS-2 simulation platform and by
means of statistical model checking with the Hybrid
28
Automata Stochastic Logic. Researcher have used
different different models which are Non-
Markoveianmodling and verification of DoS,
Generalized stochastic Petri nets (GSPN), eGSPN
and HASL modeling checking with two algorithm
which are LEACH and k-LEACH.
Ghildiyal S et al.[10] focued on the Wireless Sensor
Network characteristics, constraints and types of DoS
attacks at different layer. Different layers of WSN
nodes have variety of roles to play for proper their
proper functioning at different layers like signaling,
framing, forwarding, reliable transportation and user
interaction at both receiving as well as sending end.
Many denial of service attacks are identified at each
layer which are meant for purposeful, planned attacks
to jeopardize the availability of service, restricting
the WSN utility for application.
Wazid M et al.[11] carried out a parametric study of
Blackhole attack is measured on the network
parameters followed by the proposal of a novel
technique for the detection and prevention of
Blackhole attack in WSN. The presences of
blackhole attack both parameters of network which
are End-to-end delay and Throughput are affected.
They have observed that in the presence of blackhole
attack the performance of network degraded very
rapidly. The End-to-end delay increases to 4.03 msec
and throughput decreases to 5027.85 bps. So it has
become very important to provide a detection and
prevention mechanism for blackhole attack.
Singh V. Pal, UkeyAnand A. S. and Jain S.[12]
proposed to detect and prevent hello flood attack
using signal strength of received Hello messages.
Nodes have been classified as friend and stranger
based on the signal strength of Hello messages sent
by them. Nodes classified as stranger are further
validated by sending a simple test packet; if the reply
of test packet comes back in a predefined time then it
is treated as valid otherwise it is treated as malicious.
The algorithm is implemented in ns-2 by modifying
the AODV-routing protocol. The performance of
algorithm has been tested under different network
scenarios. The simulation results show improved
performance of the new algorithm in terms of number
of packet delivery ratio as compare to AODV with
hello flood attack, hello flood attack is an important
attack on the network layer, in which an adversary,
IITM Journal of Management and IT
which is not a legal node in the network, can flood
hello request to any legitimate node using high
transmission power and break the security of WSNs.
Hai T. H. and Eui-Nam H.[13] proposed a
lightweight detection algorithm based only on the
neighborhood information. His detection algorithm
can detect selective forwarding attack with high
accuracy and little overhead imposed on detection
modules than previous works. His algorithm has been
evaluated and shows a good effectiveness even the
high density of network and the high probability of
collisions in WSNs. Besides, our detection modules
consume less energy than previous works by using
over-hearing mechanism to reduce the transmission
of alert packets.
Malik R. and Sehrawat H.[14] studied on some
important attacks over WSNs with possible ways to
detect and defend selective forwarding attack. WSNs
have issues like low memory and limited battery
availability, so conventional security establishments
are not effective here. A number of attacks are
possible over WSNs like black hole, wormhole and
selective forwarding attack. Selective forwarding
attack is a special case of black hole attack where
compromised nodes drop packets selectively.
Researcher have reviewed the some OSI layers such
as Application, Transport, Network, Data link and
Physical and these attacks and defense strategy.
Soni V., Modi P. and Chuadhri V.[15] have presented
some countermeasures against the sinkhole attack.
There are many possible attacks on sensor network
such as selective forwarding, jamming, sinkhole,
wormhole, Sybil and hello flood attacks. Sinkhole
attack is among the most destructive routing attacks
for these networks. It may cause the intruder to lure
all or most of the data flow that has to be captured at
the base station. Once sinkhole attack has been
implemented and the adversary node has started to
work as network member in the data routing, it can
apply some more threats such as black hole or gray
hole. Ultimately this drop of some important data
packets can disrupt the sensor networks completely.
Rassam M. A et al.[16] concluded that the
vulnerabilities of Mintroute protocol to sinkhole
attacks are discussed and the existing manual rules
used for detection are investigated using different
architecture. On here different types of protocols
Volume 10, Issue 1 ∙January-June 2019
were proposed for WSN, they cannot guarantee the
protection of the network from different attacks.
Sinkhole attacks which are launched by a new or a
compromised node attracts the network traffic to pass
through it. This attack leads to many other attacks
such as blackhole, wormhole, or even information
fabrication attacks.
Khanderiya M. and Panchal M.[17] proposed work
researcher has tried to give a method that could detect
Sybil attack in Wireless Sensor Networks.
Researchers have developed many schemes and
methodologies for detecting and preventing Sybil
attack, but these security mechanisms are not being
used satisfactorily in real scenario for Wireless
Sensor Networks. They have presented a robust and
lightweight solution to detect Sybil attack using RSSI
(Received Signal strength Indicator). RSSI value is
used for detecting Sybil attack, but there three
detectors were used for this scheme. Three detectors
were required because the nodes that are at same
distance from detecting node would have same RSSI
value, so single node was not enough for detection
process, as it would regard those nodes as Sybil too.
Dhamodharan U. K. R. and Vayanaperumal R.[18]
designed and experimental program on a scheme of
assuring security for wireless sensor network, to deal
with attacks of these kinds in unicasting
andmulticasting. Basically a Sybil attack means a
node which pretends its identity to other nodes.
Communication to an illegal node results in data loss
and becomes dangerous in the network.The existing
method Random Password Comparison has only a
scheme which just verifies the node identities by
analyzing the neighbors. A survey was done on a
Sybil attack with the objective of resolving this
problem.The survey has proposed a combined
CAM-PVM(compare and match-position verification
method) with MAP (message authentication and
passing) for detecting, eliminating, and eventually
preventing the entry of Sybil nodes in the wireless
sensor networks.
Amish P. and Vaghela V. B.[19] carried out a
parametric study on the techniques dealing with
wormhole attack in WSN are surveyed and a method
is proposed for detection and prevention of wormhole
attack. AOMDV (Ad hoc On demand Multipath
Distance Vector) routing protocol is incorporated into
29
these method which is based on RTT (Round Trip
Time) mechanism and other characteristics of
wormhole attack. As compared to other solution
shown in literature, proposed approach looks very
promising. NS2 simulator is used to perform all
simulation.
Shaikh F. A. and Patil U.[20] carried out the study to
explain a wormhole detection algorithm for Wireless
Mesh Networks which detect the wormholes by
calculating neighbor list as well as directional
neighbor list of the source node. The main aim of the
algorithm is that it can offer approximate location of
nodes and effect of wormhole attack on all nodes
which is helpful in implementing countermeasures.
The performance evaluation is complete in varying
no. of wormholes in the network.
Modirkhazeni A. et al.[21] carried out the parametric
study to focused on wormhole attack and proposed
distributed network discovery approach to mitigate
its effect. Researcher has presented selected
countermeasures and then we proposed network
discovery approach which needs no additional tools
or accurate time synchronization. According to the
simulation our approach can mitigated almost 100%
of wormhole attack overload in the environment
where 54% of nodes are affected with the wormhole.
Ahmad Salehi S. et. al.[22] have analyzed networks
require security plan due to various limitations of
resources and the prominent characteristics of a
wireless sensor network which is a considerable
challenge in WSNs the node nature causes limitations
like restricted energy, capability of processing, and
storage capacity. These restrictions create WSNs so
distinctive from conventional ad hoc wireless
networks. Specific methods and protocols have been
advanced to utilize in WSNs. All of the mentioned
security dangers including the Hello flood attack,
wormhole attack, Sybil attack, sinkhole attack, offer
one usual goal which is for compromising the
integrity of the network they attack. In this paper,
they principally concentrate on the threats in WSN
security and the abstract of the WSNs threats which
influence various layers along with their defense
techniques is presented.
Bhalla M., Pandey N. and Kumar B.[23] have
analyzed security issues and vulnerabilities in
wireless sensor networks. All the security protocols
30
mentioned should be analyzed using simulation and
some more features like speed-of- operation, Power
Consumption and Efficiency should be evaluated.
Aldhobaiban D., Elleithy K. and Almazaydeh L.[24]
carried out an experimental research has shown how
the network nodes can be rerouted to avoid the
attacked nodes. The deletion of the links for the
infected node and its neighbor leads to the security of
the rest of the node network. Since this approach
requires a table to monitor all the nodes, the load on
the network is overwhelming.
Sakthivel T. and Chandrasekaran R. M.[25] carried
out the proposed Path Tracing (PT) algorithm for
detection and prevention of wormhole attack as an
extension of DSR protocol. The PT algorithm runs on
each node in a path during the DSR route discovery
process. The PT algorithm detects and prevents the
wormhole attack using per hop distance between two
nodes. They proposed algorithm implementation
depends on DSR protocol. They then simulated the
proposed algorithm in NS-2. The parameters like
throughput, overhead and the average delay of the
proposed algorithm are compared with that of
existing wormhole prevention techniques.
Modirkhazeni A et al.[26] focused on wormhole
attack and proposed distributed network discovery
approach to mitigate its effect.Then they focused on
the wormhole attack in these kinds of networks and
presented selected countermeasures. Afterward they
generalized previous countermeasures, analyzed them
and selected the better one. And then base on the
presented results they proposed network discovery
approach base on distributed scheme which needs no
additional tools or accurate time synchronization.
According to the simulation proposed approach acted
efficiently and mitigated almost 100% of wormhole
attack overload in the environment where 54% of
nodes are affected with the wormhole.
S. I. Eludiora, et.al[27] reviewed the existing
approach to security solutions in WSNs and proposed
the use of a distributed approach. The approach will
allow SNs to communicate directly with the BSs
rather than forming cluster-heads among themselves.
Mobile Agents (MAs) were introduced to facilitate
communication among the BSs. MAs can easily
move from one host to another and perform
necessary tasks. Researchers developed a Distributed
IITM Journal of Management and IT
IDS for WSNs. Distributed IDS is implemented using
TMote sky wireless sensor for testing and simulation
over specified parameters.
Ioannou C., Vassiliou V. and Sergiou C[28] have
analyzed they proposed a general methodology of an
anomaly-based Intrusion Detection System (IDS),
named mIDS, that uses the Binary Logistic
Regression (BLR) statistical tool to classify local
sensor activity to either benign or malicious to detect
a malicious behavior within a sensor node. Attacks
have been implemented within the Contiki O/S and
we tested the results using the associated COOJA
simulator. All sensor nodes are equivalent to TelosB
nodes and have a 25m radio range. They created a
model that took into consideration both attacks and
evaluated in different network topologies.
Shanthi. S., E. G, Rajan[29] discussed many potential
issues of WSN security and detection mechanisms
and present a comprehensive analysis of various
Intrusion Detection approaches (signature based
detection system, anomaly based detection system,
hybrid based detection system) in Wireless Sensor
Networks.
III. CONCLUSION
The interest for security in WSNs turns out to be
more self-evident amid capacity development of
WSNs and they are utilized substantially more,
nonetheless, in WSNs the hub nature causes
impediments like confined vitality, ability of
preparing, and capacity limit. These confinements
make WSNs so particular from regular specially
appointed remote systems. Particular techniques and
conventions have been progressed to use in WSNs.
The majority of the said security perils including the
Welcome surge assault, wormhole assault, Sybil
assault, sinkhole assault, offer one common objective
which is for trading off the uprightness of the system
they assault.
The security of WSNs has turned into a noteworthy
subject since of the distinctive perils showing up and
the importance of information classification, in spite
of the fact that before, there was a little focus on
WSNs security. There are a few answers for secure
against all risks, albeit a few arrangements have
already been recommended. In this article, we
essentially focus on the dangers in WSN security and
Volume 10, Issue 1 ∙January-June 2019
the conceptual of the WSNs dangers which impact
different layers alongside their protection methods is
displayed. As of late, set up of concentrating on
different layers, researchers are striving for
incorporated framework for security component. The
most normal security risk in different layers and the
most sensible arrangement in this paper are exhibited.
REFERENCES
[1] MalekGuechari, Lynda Mokdad and Sovanna
Tan. "Dynamic Solution for Detecting Denial
of Service Attacks in Wireless Sensor
Networks" IEEE, pp. 173-177, (2012).
[2] Rolla P. and Kuar M.,"Dynamic Forwarding
Window Technique against DoS Attack in
WSN" IEEE, pp.212-216, DOI
10.1109/ICMETE.2016.93, (2016).
[3] Patil S. and Choudhari S. "DoS attack
prevention technique in Wireless Sensor
Networks" IEEE, pp.715-721, DOI:
10.1016/j.procs.2016.03.094, (2016).
[4] Naik S. and Shekokar N., "Conservation of
energy in wireless sensor network by
preventing denial of sleep attack" ELSVIER,
PP.370-379, doi:
10.1016/j.procs.2015.03.164, (2015).
[5] Chaudhary S. and Thanvi P., "Performance
Analysis of Modified AODV Protocol in
Context of Denial of Service (Dos) Attack in
Wireless Sensor Networks" International
Journal of Engineering Research and General
Science Volume 3, pp.486-491(2015).
[6] S. Fouchal et al., "Recursive-clustering-based
approach for denial of service (DoS) attacks
in wireless sensors networks" International
journal of communication systems, pp.309-
324, DOI: 10.1002/dac.2670, (2015).
[7] Mansouri D., Mokddad L., Ben-othman J.
and Ioualalen M., "Preventing Denial of
Service Attacks in WirelessSensor Networks"
IEEE, PP.3014-3019, (2015).
[8] Kiss I., Haller P. and Beres A., " Denial of
Service attack detection in case of Tennessee
Eastman challenge process" ELSEVIER,
pp.835-841, doi:
10.1016/j.protcy.2015.02.120, (2015).
31
[9] Ballarini P., Mokdad L. and Monnennt Q.,
"Modeling tools for detecting DoS attacks in
WSNs" Security and Communication
Networks, pp.420-436, DOI:
10.1002/sec.630, (2013).
[10] Ghildiyal S., Mishra A. K., Gupta A. and
Garg N., "ANALYSIS OF DENIAL OF
SERVICE (DOS) ATTACKS IN WIRELESS
SENSOR NETWORKS" IJRET:
International Journal of Research in
Engineering and Technology, pp.140-143,
(2014).
[11] Wazid M., Katal A., Sachan R. S., R H
Goudar and D P Singh., "Detection and
Prevention Mechanism for Blackhole Attack
in Wireless Sensor Network" IEEE, pp.576-
581, (2013).
[12] Singh V. Pal, UkeyAnand A. S. and Jain S.,
"Signal Strength based Hello Flood Attack
Detection and Prevention in Wireless Sensor
Networks" International Journal of Computer
Applications, pp.1-6, DOI: 10.5120/10153-
4987, (2013).
[13] Hai T. H. and Eui-Nam H., "Detecting
Selective Forwarding Attacks in Wireless
Sensor Networks Using Two-hops Neighbor
Knowledge" IEEE, pp.325-331, DOI
10.1109/NCA.2008.13, (2008).
[14] Malik R. and Sehrawat H., "Comprehensive
Study of Selective Forwarding Attack in
Wireless Sensor Networks" International
Journal of Advanced Research in Computer
Science, vol.8, pp.1835-1838, (2017).
[15] Soni V., Modi P. and Chuadhri V.,
"Detecting Sinkhole Attack in Wireless
Sensor Network" International Journal of
Application or Innovation in Engineering &
Management (IJAIEM), Vol. 2, pp.29-32,
(2013).
[16] Rassam M. A., Zainal A., Maarof A. and Al-
Shaboti M., "A Sinkhole Attack Detection
Scheme in Mintroute Wireless Sensor
Networks" IEEE, pp.71-75, DOI:
10.1109/ISTT.2012.6481568, (2012).
32
[17] Khanderiya M. and Panchal M.,"A Novel
Approach for Detection of Sybil Attack in
Wireless Sensor Networks" IJSRSET, Vol. 2,
pp.113-117, (2016),.
[18] Dhamodharan U. K. R. and Vayanaperumal
R., "Detecting and Preventing Sybil Attacks
in Wireless Sensor Networks Using Message
Authentication and Passing Method" Hindawi
Publishing Corporation Scientific World
Journal, pp.1-7,
http://dx.doi.org/10.1155/2015/841267,
(2015).
[19] Amish P. and Vaghela V. B., "Detection and
Prevention of Wormhole Attack in Wireless
Sensor Network using AOMDV protocol"
ELSEVIER, pp.700-707, doi:
10.1016/j.procs.2016.03.092, (2016).
[20] Shaikh F. A. and Patil U. "Efficient
Detection and prevention of Wormhole
Attacks in
[21] Wireless Mesh Network" International
Research Journal of Engineering and
Technology (IRJET), Vol. 4, pp.2208-2214,
(2017),.
[22] Modirkhazeni A., Aghamahmoodi S.,
Modirkhazeni A. and Niknejad N.,
"Distributed Approach to Mitigate Wormhole
Attack in Wireless Sensor Networks" IEEE,
pp.122-128, (2012).
[23] Ahmad Salehi S., Razzaque M. A., Naraei P.
and Farrokhtala A., "Security in Wireless
Sensor Networks: Issues and Challanges"
IEEE, (2013).
[24] Bhalla M., Pandey N. and Kumar B.
―Security Protocols for Wireless Sensor
Networks‖ IEEE, pp. 1005-1009, (2015),.
[25] Aldhobaiban D., Elleithy K. and
Almazaydeh L., "Prevention of Wormhole
Attacks in Wireless Sensor Networks" IEEE,
pp.287-291, DOI 10.1109/AIMS.2014.57,
(2014).
[26] Sakthivel T. and Chandrasekaran R. M.,
"Detection and Prevention of Wormhole
Attacks in MANETs using Path Tracing
IITM Journal of Management and IT
 Approach" European Journal of Scientific Detection Scheme for Wireless Sensor
Research, pp.240-252, (2012). Networks‖ IEEE, 2011.
[27] Modirkhazeni A., Aghamahmoodi S., [29] Ioannou C., Vassiliou V. and Sergiou C.,
Modirkhazeni A. and Modirkhazeni N., "An Intrusion Detection System for Wireless
"Distributed Approach to Mitigate Wormhole Sensor Networks" IEEE. 2017.
Attack in Wireless Sensor Networks" IEEE,
[30] Shanthi. S., E. G, Rajan.
pp.122-128, (2012).
―Comprehensive Analysis of Security Attacks
[28] S. I. Eludiora, O.O. Abiona, A. O. and Intrusion Detection System in Wireless
Oluwatope, S. A. Bello, M.L Sanni, , D. O. Sensor Networks‖ IEEE 2016, pp. 24-31,
Ayanda, C.E Onime E. R. Adagunodo and 2016
L.O. Kehinde ―A Distributed Intrusion

Volume 10, Issue 1 ∙January-June 2019 33

 A Review on Histogram of Oriented Gradient
Apurva Jain1, Deepanshu Singh2
1,2
Research Scholar, IITM, GGSIPU, New Delhi, India
apurva.2296@gmail.com, singh.deepanshu207@gmail.com

Abstract—Pedestrian detection systems are image of size 256 x 256 pixels. The input image is
receiving increasing attention in both industry and divided into 256 cells with a cell size of 16 x 16
academia with the rapid development of autonomous pixels and each cell is divided into four sub-cells as
automobiles which employ artificial intelligence.This with a sub-cell size of 8 x 8 pixels.
article describes the approaches to Histogram of
The gradients Gx and Gy of a particular pixel
Oriented Gradient and support vector machine,
location is computed using 1D masks in X and Y
focusing on studying the HOG feature and
direction as:
application, detailing the process of HOG feature
extracted and the design of classifiers in pedestrian Gx = Mx• Ix
detection. Gy = My• Iy
I. INTRODUCTION whereMx = [-1 0 1] and My = [-1 0 1]T are the masks
Real-time human detection from videos is one of the used on Ix in X-direction and Iy in Y-direction
most active areas in computer vision due to its respectively. The gradient magnitude, Gmag and
widespread applications such as intelligent orientation angle, Gdir are computed.
surveillance and home security in [1], personal The gradient vector for each pixel location in the sub-
protection and kidnapping detection in [2], automatic cells are drawn using Gmag and Gdir values The
detection of crimes in [3] and human computer gradient vectors in each sub-cell are normalized to
interfaces in [4]. The successful progress towards obtain a key descriptor as shown. Similarly, key
design of autonomous vehicles such as autonomous descriptors are computed from each cell to obtain
cars [5,6], self-driven rider less bicycles [7], HOG features of input image.
autonomous robots [8] and drones [9,10] have
spearheaded research in the area of pedestrian
detection. Pedestrian detection using HOG and neural
networks is reported in [11], pedestrian detection for
advance driver assistance system using HOG and
Adaboost is reported in [12], pedestrian detection
using Bayesian and Edgelet detectors in [13], using
local binary patterns (LBP) in [14], using Motion and
Appearance patterns in [15], using Shapelet features
in [16] and using Deep Networks in [17].
II. DESIGN OF PROPOSED PEDESTRIAN
B. SVM Classifier
DETECTION SYSTEM
Support vector machine (SVM) is a new method of
In this section, the design of proposed real-time
machine learning based on statistic learning theory
pedestrian detection system is explained with feature
proposed by Vapnik et, the goal of which is to find
extraction and classifier being the two main stages of
the optimal classification surface [21].
the system
SVM is considered to be the simplest and fastest
A. HOG Feature Extraction classifier for both linear as well as non-linear
The sequence of steps followed to obtain these HOG Classification problems [18]. SVM learning aims at
features from an input image, considering an input

34 IITM Journal of Management and IT

finding a good hyperplane in a higher dimensional
feature space, that best separates two classes as
shown in Fig. 5. The equation of SVM hyper plane
that classifies two classes is given as
where, z is the test data, K(xsv , z) is the kernel
function, NSV denotes the number of support
vectors, xsv is the input, is the target output and is the
Lagrange multiplier associated with each training
data.
Fig. 2 optimal classification surface in the linearly
separable case [21].
III. EXPERIMENTAL RESULTS
The performance of proposed system is evaluated by
carrying out various experiments using images
collected online, using images from INRIA Person
Dataset [20] and using real-time inputs. The photo
collage of samples considered for training from two
classes, background and pedestrians for images
obtained online. The performance of proposed
pedestrian detection system using images obtained
online and real-time input images, which is again a
photo collage of few results obtained, wherein a blue
box denotes the identification of apedestrian in an
image. The performance result of proposed system is
reported in Table I for all the experiments carried out
using different datasets. It is observed that a
maximum accuracy of 98.31% is INRIA datasets,
which is a mixture of cropped images of pedestrians
and backgrounds. Pedestrian detection accuracy of
96.35% is obtained on testing the system using full
Volume 10, Issue 1 ∙January-June 2019
scenery images provided in the INRIA database. It is
observed during experimentation that the proposed
real-time pedestrian detection system gave 93.22%
accuracy based on the distance between pedestrian
and camera, whereas 100% accuracy is obtained, the
moment pedestrian comes closer towards the camera.
This suffices the requirement of pedestrian detection
for autonomous vehicles.
IV. CONCLUSION
In this paper, design of a real-time pedestrian
detection system for autonomous vehicles is
proposed and its performance is evaluated by
carrying out various experiments using offline
images, images from standard dataset and real- time
input. The system is capable of detecting pedestrians
with an accuracy of 98.31% and it is observed that
the non-detected pedestrians are also detected, once
they come closer to the camera, thus achieving 100%
recognition accuracy.
REFERENCES
[1] Yi G, Myoungjin K, Yun C and H Lee,
―Design and Implementation of UPnP-based
Surveillance Camera system for Home
Security‖, in Proc. Int. Conf. on Information
Science and Applications, Suwon, South
Korea, June 2013.
[2] Miyahara A and Nagayama I, ―An Intelligent
security camera system for kidnapping
detection‖, Journal of Advanced
Computational Intelligence and Intelligent
Informatics, 2017
[3] Goya K, Zhang X, Kitayama K and Nagayama
I, ―A Method for Automatic Detection of
Crimes for Public Security by using Motion
Analysis‖, in Proc. Fifth Int. Conf. on
Intelligent Information Hiding and
Multimedia Signal Processing, Kyoto, Sep
2009.
[4] R. Moldovan, B. Orza, A. Vlaicu and C.
Porumb, ―Advanced human-computer
interaction in external resource annotation,‖ in
Proc. Int. Conf. on Automation, Quality and
Testing, Robotics, Cluj-Napoca, 2014,
[5] K. Bimbraw, ―Autonomous cars: Past, present
and future a review of the developments in the
last century, the present scenario and the
expected future of autonomous vehicle
technology,‖ in Proc. 12th Int. Conf. on
35
 Informatics in Control, Automation and
Robotics, Colmar, 2015.
[6] Pozna and C. Antonya, ―Issues about
autonomous cars,‖ in Proc. 11th Int. Sym. on
Applied Computational Intelligence and
Informatics, Timisoara, 2016
[7] A Brizuela-Mendoza, CM Astorga-Zaragoza,
A Zavala-Río, F Canales- Abarca and J Reyes-
Reyes,―Fault Tolerant Control for
Polynomial Linear Parameter Varying (LPV)
Systems applied to the stabilization of
ariderless bicycle,‖ in Proc. Control
Conference (ECC), 2013 European, Zurich,
2013.
[8] S. A. Miratabzadeh et al., "Cloud robotics: A
software architecture: For heterogeneous
large-scale autonomous robots," 2016 World
Automation Congress (WAC), Rio Grande,
PR, USA, 2016.
[9] S. Perkins, "From insects to drones," in
Engineering & Technology, Vol. 10, No. 11,
pp. 72-76, December 2015.
[10] Moskvitch, "Take off: are drones the future of
farming?," in Engineering & Technology, Vol.
10, No. 7-8, pp. 62-66, August- September
2015.
[11] Gangqiang Shi and Weiwei Hu, "The video
people detection based on neural network,"
2015 IEEE 16th International Conference on
Communication Technology (ICCT),
Hangzhou, 2015.
[12] Rettkowski, A. Boutros and D. Gohringer,
"Real-time pedestrian detection on a
xilinxzynq using the HOG algorithm," 2015
International Conference on ReConFigurable
Computing and FPGAs (ReConFig), Mexico
City, 2015.
[13] Wu B and Nevatia Ram, ―Detection and
Tracking of Multiple, Partially Occluded
Humans by Bayesian Combination of Edgelet
based Part Detectors,‖ International Journal of
Computer Vision, Nov 2007.
36
[14] X Wang, TX Han and S Yan, "An HOG-LBP
human detector with partial occlusion
handling," 2009 IEEE 12th International
Conference on Computer Vision, Kyoto, 2009.
[15] P Viola, MJ Jones and D Snow, "Detecting
pedestrians using patterns of motion and
appearance," Computer Vision, 2003.
Proceedings. Ninth IEEE International
Conference on, Nice, France, 2003.
[16] P Sabzmeydani and G Mori, "Detecting
Pedestrians by Learning Shapelet Features,"
2007 IEEE Conference on Computer Vision
and Pattern Recognition, Minneapolis, MN,
2007.
[17] Anelia A, Alex K, Vincent V, Abhijit O and
Dave F, ―Real-time Pedestrian Detection with
Deep Network Cascades,‖ in Proc. 26th British
Machine Vision Conference (BMVC),
Swansea UK, Sep 2015.
[18] Christopher Burges, ―A tutorial on Support
Vector Machines for Pattern Recognition,‖
Data Mining and Knowledge Discovery 2,
1998.
[19] J. Manikandan and B. Venkataramani, ―Study
and evaluation of a multi- class SVM classifier
using diminishing learning technique,‖
Neurocomputing, Volume 73, Issues 10–12,
June 2010.
[20] INRIA Person Dataset.
http://pascal.inrialpes.fr/data/human/.
[21] Huo, D., Liu, H., Shang, Z. and Li, R.,
―Research for pedestrian detection classifier.‖,
In 2016 7th IEEE International Conference on
Software Engineering and Service Science
(ICSESS) ,2018.
[22] Goel, R., Srivastava, S. and Sinha, A.K., 2017.
A Review of Feature Extraction Techniques
for Image Analysis. International Journal of
Advanced Research in Computer and
Communication Engineering, 2017.
IITM Journal of Management and IT
 Breast Cancer Risk Prediction
Pankaj Kumar Varshney1, Hemant Kumar2,
Jasleen Kaur3, Ishika Gera4
1, 2, 3,4
Department of Computer Science,
Institute of Information Technology and Management, Janakpuri
pankaj.surir@gmail.com, hemantmbmgla@gmail.com,
kaurjasleen420.jk@gmail.com, ishikagera1998@gmail.com

Abstract.-The number and size of restorative/medical Artificial Neural Network.

databases are expanding quickly yet the greater part
I.
of these information are not investigated for finding
the significant and concealed learning. Propelled
information and data mining methods can be utilized
to find concealed examples and connections. Models
created from these strategies are helpful for
medicinal specialists to settle on right choices. The
present research contemplated the utilization of
information mining strategies to create prescient
models for bosom (breast) malignant growth repeat
in patients who were followed-up for a long time.
Objective is to fabricate a model utilizing many
machine learning algorithms to foresee whether
bosom cell tissue is malignant (cancerous) or benign
(non-cancerous).We executed machine learning
techniques/algorithms, i.e., k-nearest algorithm,
Logistic Regression, Decision Tree, Random Forest,
Gradient Boosting, Support Vector Machine (SVM),
and Artificial Neural Network (ANN) to build up the
prescient models. The primary objective of this paper
is to think about the execution of these outstanding
calculations on our information through affectability,
explicitness, and precision to think about the
execution of these outstanding calculations on our
information through affectability, explicitness, and
precision. Our analysis shows that accuracy of DT,
RF, SVM and ANN are 0.937, 0.951, 0.965, and
0.958 respectively. The SVM classification model
predicts bosom malignancy repeat with least error
rate and most astounding precision. The anticipated
exactness of the DT demonstrate is the most minimal
of all. The results are achieved using 10-fold cross-
validation for measuring the unbiased prediction
accuracy of each model.
Keywords- Classification; Logistic Regression‟s-
nearest algorithm; Decision tree; Random Forest;
Gradient Boosting; Support vector machine;
INTRODUCTION
Breast cancer (BC) is the most common cancer in
women, affecting about 10% of all women at some
stages of their life. In recent years, the incidence rate
keeps increasing and data show that the survival rate
is 88% after five years from diagnosis and 80% after
10 years from diagnosis [1]. Breast cancer is the
second leading cause of cancer death among women.
It occurs as a result of abnormal growth of cells in
the breast tissue, commonly referred to as a Tumor.
A tumor does not mean cancer - tumors can be
benign (not cancerous), pre-malignant (pre-
cancerous), or malignant (cancerous). Tests such as
MRI, mammogram, ultrasound and biopsy are
commonly used to diagnose breast cancer performed.
Build model to predict whether breast cell tissue is
malignant or benign, we will construct a predictive
model using SVM machine learning algorithm to
predict the diagnosis of a breast tumor. They studied
951 breast cancer patients and used tumor size,
auxiliary nodal status, histological type, mitotic
count, nuclear pleomorphism, tubule formation,
tumor necrosis, and age as input variables [7].
Pendharker patterns in breast cancer. In this study,
they showed that data mining could be a valuable
tool in identifying similarities (patterns) in breast
cancer cases, which can be used for diagnosis,
prognosis, and treatment purposes [4]. These studies
are some examples of researches that apply data
mining to medical fields for prediction of diseases.
246,660 of women's new cases of invasive breast
cancer are expected to be diagnosed in the US during
2016 and 40,450 of women‟s death is estimated2.
Breast cancer represents about 12% of all new cancer
cases and 25% of all cancers in women[3]. Early
prediction of breast cancer is one of the most crucial

Volume 10, Issue 1 ∙January-June 2019 37

works in the follow-up process. Data mining
methods can help to reduce the number of false
positive and false negative decisions [2,3].
Consequently, new methods such as knowledge
discovery in databases (KDD) has become a popular
research tool for medical researchers who try to
identify and exploit patterns and relationships among
large number of variables, and predict the outcome
of a disease using historical cases stored in datasets.
Machine learning is not new to cancer research.
Artificial neural networks (ANNs) and decision
trees(DTs) have been used in cancer detection and
diagnosis for nearly 20 years. Today machine
learning methods are being used in a wide range of
applications ranging from detecting and classifying
tumors via X-ray and CRT images to the
classification of malignancies from proteomic and
genomic (microarray). According to the latest
PubMed statistics, more than 1500 papers have been
published on the subject of machine learning and
cancer. However, the vast majority of these papers
are concerned with using machine learning methods
to identify, classify, detect, or distinguish tumors and
other malignancies. In other words machine learning
has been used primarily as an aid to cancer diagnosis
and detection [4]. In this paper, using data mining
techniques, authors developed models to predict the
recurrence of breast cancer by analyzing data
collected from ICBC registry. The next sections of
this paper review related work, describe background
of this study, evaluate three classification models
(DT, SVM, and ANN), explain the methodology
used to conduct the prediction, present experimental
results, and the last part of the paper is the
conclusion. To estimate validation of the models,
accuracy, sensitivity, and specificity were used as
criteria, and were compared. In the present work only
studies that employed ML techniques for modeling
cancer diagnosis and prognosis are presented.
II. LITERATURE REVIEW
A literature review showed that there have been
several studies on the survival prediction problem
using statistical approaches and artificial neural
networks. However, we could only find a few studies
related to medical diagnosis and recurrence using
data mining approaches such as decision trees [5,6].
Delen et al. used artificial neural networks, decision
trees and logistic regression to develop prediction
38
models for breast cancer survival by analyzing a
large dataset, the SEER cancer incidence database
[6]. Linden et al. used ANN and logistic
regression models to predict 5, 10, and 15 -year
breast cancer survival. They studied 951 breast
cancer patients and used tumor size, auxiliary nodal
status, histological type, mitotic count, nuclear
pleomorphism, tubule formation, tumor necrosis, and
age as input variables
[7] Pendharker patterns in breast cancer. In
this study, they showed that data mining could be a
valuable tool in identifying similarities (patterns) in
breast cancer cases, which can be used for diagnosis,
prognosis, and treatment purposes [4]. These studies
are some examples of researches that apply data
mining to medical fields for prediction of diseases.
III. MATERIAL AND METHODS
Machine Learning, a branch of Artificial Intelligence,
relates the problem of learning from data samples to
the general concept of inference [5]. Every learning
process consists of two phases: (i) estimation of
unknown dependencies in a system from a given
dataset and (ii) use of estimated dependencies to
predict new outputs of the system. It has also been
proven an interesting area in biomedical research
with many applications, where an acceptable
generalization is obtained by searching through an n-
dimensional space for a given set of biological
samples, using different techniques and algorithms
[2]. There are two main common types of Machine
learning methods known as (i) supervised learning
and (ii) unsupervised learning. In supervised learning
a labeled set of training data is used to estimate or
map the input data to the desired output. In contrast,
under the unsupervised learning methods no labeled
examples are provided and there is no notion of the
output during the learning process. As a result, it is
up to the learning scheme/model to find patterns or
discover the groups of the input data. In supervised
learning this procedure can be thought as a
classification problem. The task of classification
refers to a learning process that categorizes the data
into a set of finite classes. Two other common ML
tasks are regression and clustering. In the case of
regression problems, a learning function maps the
data into a real-value variable. Subsequently, for each
new sample the value of a predictive variable can be
IITM Journal of Management and IT
estimated, based on this process. Clustering is a
common unsupervised task in which one tries to find
the categories or clusters in order to describe the data
items. Based on this process each new sample can be
assigned to one of the identified clusters concerning
the similar characteristics that they share. Suppose
for example that we have collected medical records
relevant to breast cancer and we try to predict if a
tumor is malignant or benign based on its size. The
ML question would be referred to the estimation of
the probability that the tumor is malignant or no (1 =
Yes, 0=No). It depicts the classification process of a
tumor being malignant or not. The circled records
depict any misclassification of the type of a tumor
produced by the procedure. Another type of ML
methods that have been widely applied is semi-
supervised learning, which is a combination of
supervised and unsupervised learning. It combines
labeled and unlabeled data in order to construct an
accurate learning model. Usually, this type of
learning is used when there are more unlabeled
datasets than labeled. When applying a ML method,
data samples constitute the basic components [4].
Every sample is described with several features and
every feature consists of different types of values.
Furthermore, knowing in advance the specific type of
data being used allows the right selection of tools and
techniques that can be used for their analysis. Some
data-related issues refer to the quality of the data and
the preprocessing steps to make them more suitable
for ML. Data quality issues include the presence of
noise, outliers, missing or duplicate data and data that
is biased-unrepresentative. When improving the data
quality, typically the quality of the resulting analysis
is also improved. In addition, in order to make the
raw data more suitable for further analysis,
preprocessing steps should be applied that focus on
the modification of the data. A number of different
techniques and strategies exist, relevant to data
preprocessing that focus on modifying the data for
better fitting in a specific ML method. Among these
techniques some of the most important approaches
include (i) dimensionality reduction (ii) feature
selection and (iii) feature extraction. Feature
Selection is one of the core concepts in machine
learning which hugely impacts the performance of
your model. The data features that you use to train
your machine learning models have a huge influence
on the performance you can achieve. Irrelevant or
Volume 10, Issue 1 ∙January-June 2019
partially relevant features can negatively impact
model performance. Feature selection and Data
cleaning should be the first and most important step
of your model designing [1,2].In this post, you will
discover feature selection techniques that you can use
in Machine Learning. Feature Selection is the process
where you automatically or manually select those
features which contribute most to your prediction
variable or output in which you are interested in.
Having irrelevant features in your data can decrease
the accuracy of the models and make your model
learn based on irrelevant features. There are many
benefits regarding the dimensionality reduction when
the datasets have a large number of features. ML
algorithms work better when the dimensionality is
lower [10]. Additionally, the reduction of
dimensionality can eliminate irrelevant features,
reduce noise and can produce more robust learning
models due to the involvement of fewer features. In
general, the dimensionality reduction by selecting
new features which are a subset of the old ones is
known as feature selection. Three main approaches
exist for feature selection namely embedded, filter
and wrapper approaches[8,9]. In the case of feature
extraction, a new set of features can be created from
the initial set that captures all the significant
information in a dataset. The creation of new sets of
features allows for gathering the described benefits of
dimensionality reduction. However, the application
of feature selection techniques may result in specific
fluctuations concerning the creation of predictive
feature lists. Several studies in the literature discuss
the phenomenon of lack of agreement between the
predictive gene lists discovered by different groups,
the need of thousands of samples in order to achieve
the desired outcomes, the lack of biological
interpretation of predictive signatures and the
dangers of information leak recorded in published
studies. Fig 1 depicts the feature correlation/selection
of diagnosis dataset.
39

Fig 1 Feature Correlation i.e, red dots correspond to
malignant diagnosis and blue to benign.
However, the application of feature selection
techniques may result in specific fluctuations
concerning the creation of predictive feature lists.
Several studies in the literature discuss the
phenomenon of lack of agreement between the
predictive gene lists discovered by different groups,
the need of thousands of samples in order to achieve
the desired outcomes, the lack of biological
interpretation of predictive signatures and the
dangers of information leak recorded in published
studies. Fig 2 depicts count of diagnosis in which
green color shows benign tumor and blue color
shows malignant tumor.
Fig 2 Diagnosis:- Benign(B)-357, Malignant(M)-219
A dataset used for machine learning should be
partitioned into three subsets — training, test, and
validation sets. Training set: A data scientist uses a
training set to train a model and define its optimal
parameters, a subset to train a model. Test set: A test
set is needed for an evaluation of the trained model
and its capability for generalization, a subset to test
40
the trained model. Validation set: The purpose of a
validation set is to tweak a model‟s hyper parameters
higher-level structural settings that can‟t be directly
learned from data. These settings can express, for
instance, how complex a model is and how fast it
finds patterns in data. The k-NN algorithm is
arguably the simplest machine learning algorithm. To
make a prediction for a new data point, the algorithm
finds the closest data points in the training dataset—
it‟s ―nearest neighbors.‖ Fig 3 depicts the training
dataset and test dataset accuracy when the
n_neighbors value is 3.The figure shows the training
and test set accuracy on the y-axis against the setting
of n_neighbors on the x-axis. Considering a single
nearest neighbor, the prediction on the training set is
perfect. But when more neighbors are considered, the
training accuracy drops, indicating that using the
single nearest neighbor leads to a model that is too
complex. This suggests that we should choose
n_neighbors=3.Then the Accuracy of K-NN
classifier on training set and test set is 0.96 and 0.92
respectively. Logistic Regression is one of the most
common linear classification algorithm is logistic
regression. Logistic regression examines the
relationship between a binary outcome (dependent)
variable such as presence or absence of disease and
predictor (explanatory or independent) variables such
as patient demographics or imaging findings. Fig 4
depicts the coefficient magnitude of features and the
accuracy level change when we set value of C i.e.,
C=1 provides quite good performance, with 96%
(0.955) accuracy on training and 0.94 accuracy on
test set. C=100 provides higher accuracy on both
training set (0.967) and test set (0.972). C=0.01
provides lower accuracy on the training set (0.948)
and much lower accuracy on the test set (0.895).
IITM Journal of Management and IT

Fig 3 Training and test set accuracy
Fig 4 Coefficient Magnitude of features
IV. MACHINE LEARNING TECHNIQUES
Artificial Neural Networks (ANNs), Bayesian
Networks (BNs), Support Vector Machines (SVMs)
and Decision Trees (DTs) have been widely applied
in cancer research for the development of predictive
models, resulting in effective and accurate decision
making [7]. These techniques have been utilized as
an aim to model the progression and treatment of
cancerous conditions. Decision trees are a non-
parametric supervised learning method used for
classification and regression. The goal is to create a
model that predicts the value of a target variable by
learning simple decision rules inferred from the data
features [5,6].Random Forest is a flexible, easy to use
machine learning algorithm that produces, even
Volume 10, Issue 1 ∙January-June 2019
without hyper-parameter tuning, a great result most
of the time [2]. It builds multiple decision trees and
merges them together to get a more accurate and
stable prediction. Random forests, also known as
random decision forests, are a popular ensemble
method that can be used to build predictive models
for both classification and regression problems.
Ensemble methods use multiple learning models to
gain better predictive results — in the case of a
random forest, the model creates an entire forest of
random uncorrelated decision trees to arrive at the
best possible answer [1, 3]. Gradient boosting is a
machine learning technique for regression and
classification problems, which produces a prediction
model in the form of an ensemble of weak prediction
models, typically decision trees. Gradient Boosting
Classifier is used to find out accurate prediction. A
Support Vector Machine (SVM) is a discriminative
classifier formally defined by a separating hyper
plane in an N-dimensional space (N — the number of
features) that distinctly classifies the data points
either side. Support vector machine (SVM) is an
emerging powerful machine learning technique to
classify cases. SVM has been used in a range of
problems and they have already been successful in
pattern recognition in bioinformatics, cancer
diagnosis [6]. SVM is a maximum margin
classification algorithm rooted in statistical learning
theory. It is the method for classifying both linear
and non-linear data. It uses a non-linear mapping
technique to transform the original training data into
a higher dimension. It performs classification tasks
by maximizing the margin separating both classes
while minimizing the classification errors[9].Neural
networks are a set of algorithms, modeled loosely
after the human brain, that are designed to recognize
patterns. They interpret sensory data through a kind
of machine perception, labeling or clustering raw
input [5]. The patterns they recognize are numerical,
contained in vectors, into which all real-world data,
are it images, sound, text or time series, and must be
translated.
V. DISCUSSION AND RESULTS
This section presents the result of all the machine
learning algorithms that we have used to do
prediction breast cancer risk. This paper has explored
risk factors for predicting breast cancer by using
machine learning techniques. Each technique has its
41
own limitations and strengths specific to the type of A. Random Forest
application. Our results show that SVM is the best Random Forest is a flexible, easy to use machine
predictor and indicator of breast cancer because it learning algorithm that produces, even without
gives the higher accuracy in predicting data. hyper-parameter tuning, a great result most of the
VI. DECISION TREE time [2]. It builds multiple decision trees and merges
them together to get a more accurate and stable
Decision trees are a non-parametric supervised
prediction. Random forests, also known as random
learning method used for classification and
decision forests, are a popular ensemble method that
regression. The goal is to create a model that predicts
can be used to build predictive models for both
the value of a target variable by learning simple
classification and regression problems. Ensemble
decision rules inferred from the data features [5, 6].
methods use multiple learning models to gain better
Decision Tree Classifier is a class capable of
predictive results — in the case of a random forest,
performing multi-class classification on a dataset. As
the model creates an entire forest of random
with other classifiers, it takes as input two arrays: an
uncorrelated decision trees to arrive at the best
array X, sparse or dense, of size [n_samples,
possible answer [1, 3]. Random forest classifier
n_features] holding the training samples, and an
creates a set of decision trees from randomly selected
array Y of integer values, size [n_samples], holding
subset of training set and the accuracy on training set
the class labels for the training samples. The
is 0.995 and accuracy on test set is 0.951. The
maximum depth of the tree „max_depth‟ is an
random forest gives us an accuracy of 95.8%, better
argument of Decision Tree Classifier. If the
than a single decision tree, without tuning any
max_depth of tree is none, then nodes are expanded
parameters. Similarly to the single decision tree, the
until all leaves are pure or until all leaves contain less
random forest also gives a lot of importance to the
than min_samples_split samples and then accuracy
―worst radius‖ feature, but it also chooses ―perimeter
on training set is 1.000 and accuracy on test set are
worst‖ to be the most informative feature overall. Fig
0.937. If the max_depth is set a value of 4 then the
6 illustrates similarly to the single decision tree, the
accuracy on training set is 0.986 and accuracy on test
random forest also gives a lot of importance to the
set is 0.937.Feature importance in decision trees it
―worst radius‖ feature, but it also chooses ―perimeter
rates how important each feature is for the decision a
worst‖ to be the most informative feature overall.
tree makes. It is a number between 0 and 1 for each
feature, where 0 means ―not used at all‖ and 1 means
―perfectly predicts the target.‖ Fig 5 illustrates
feature perimeter worst is by far the most important
feature. This confirms our observation in analyzing
the tree that the first level already separates the two
classes fairly well.

Fig 6 Feature importance of random forest

B. Gradient Boosting
Gradient boosting is a machine learning technique for
Fig 5 Feature importance of decision tree regression and classification problems, which
produces a prediction model in the form of an

42 IITM Journal of Management and IT

ensemble of weak prediction models, typically
decision trees. Gradient Boosting Classifier is used to
find out accurate prediction. When „random state‟
argument of GB Classifier is zero (0) then accuracy
on training set is 1.000 and accuracy on test set is
0.944. As the training set accuracy is 100%, we are
likely to be over fitting. To reduce over fitting, we
could either apply stronger pre-pruning by limiting
the maximum depth or lower the learning rate. When
„max_depth‟ argument of GB Classifier is one (1)
then accuracy on training set is 0.988 and accuracy
on test set is 0.937. When „learning rate‟ argument
of GB Classifier is (0.01) then accuracy on training
set is 0.984 and accuracy on test set is 0.930.Both
methods of decreasing the model complexity reduced
the training set accuracy, as expected. In this case,
none of these methods increased the generalization
performance of the test set. Fig 7 depicts that the
feature importance‘s of the gradient boosted trees are
somewhat similar to the feature importance of the
random forests, though the gradient boosting
completely ignored some of the features. It chooses
―perimeter worst‖ to be the most informative feature
overall.
Fig 6 Feature importance of random forest
C. Gradient Boosting
Gradient boosting is a machine learning technique for
regression and classification problems, which
produces a prediction model in the form of an
ensemble of weak prediction models, typically
decision trees. Gradient Boosting Classifier is used to
find out accurate prediction. When „random state‟
argument of GB Classifier is zero (0) then accuracy
on training set is 1.000 and accuracy on test set is
0.944. As the training set accuracy is 100%, we are
likely to be over fitting. To reduce over fitting, we
Volume 10, Issue 1 ∙January-June 2019
could either apply stronger pre-pruning by limiting
the maximum depth or lower the learning rate. When
„max_depth‟ argument of GB Classifier is one (1)
then accuracy on training set is 0.988 and accuracy
on test set is 0.937. When „learning rate‟ argument
of GB Classifier is (0.01) then accuracy on training
set is 0.984 and accuracy on test set is 0.930.Both
methods of decreasing the model complexity reduced
the training set accuracy, as expected. In this case,
none of these methods increased the generalization
performance of the test set. Fig 7 depicts that the
feature importances of the gradient boosted trees are
somewhat similar to the feature importances of the
random forests, though the gradient boosting
completely ignored some of the features. It chooses
―perimeter worst‖ to be the most informative feature
overall.
Fig 7 Feature importance of Gradient Boosting
D. Support Vector Machine
A Support Vector Machine (SVM) is a
discriminative classifier formally defined by a
separating hyper plane in an N-dimensional space
(N — the number of features) that distinctly classifies
the data points either side. Support vector machine
(SVM) is an emerging powerful machine learning
technique to classify cases. SVM has been used in a
43
range of problems and they have already been
successful in pattern recognition in bioinformatics,
cancer diagnosis [6]. SVM is a maximum margin
classification algorithm rooted in statistical learning
theory. It is the method for classifying both linear
and non-linear data. It uses a non-linear mapping
technique to transform the original training data into
a higher dimension. It performs classification tasks
by maximizing the margin separating both classes
while minimizing the classification errors [9]. By
importing Support Vector Machine the accuracy on
training set is 1.00 and accuracy on test set is 0.63
the model over fits quite substantially, with a perfect
score on the training set and only 63% accuracy on
the test set. SVM requires all the features to vary on
a similar scale. We will need to rescale our data that
all the features are approximately on the same scale.
Min Max Scaler is used to rescale the data so that all
features will vary on same scale. By importing Min
Max Scaler the accuracy on training set is 0.95 and
accuracy on test set is 0.94. Scaling the data made a
huge difference as training and test set performance
are quite similar but less close to 100% accuracy. C
and gamma are the parameters of Radial Basis
Function (RBF) kernel SVM. Now, we can try
increasing either C or gamma to fit a more complex
model. By increasing C=1000 in SVM the accuracy
on training set is 0.986 and accuracy on test set is
0.965 Here, increasing C allows us to improve the
model significantly, resulting in 96.5% test set
accuracy.
E. Neural Networks
Neural networks are a set of algorithms, modeled
loosely after the human brain, that are designed to
recognize patterns.
By importing MLP Classifier the accuracy on
training set is 0.63 and accuracy on test set is
0.63.Results are not good. Neural networks also
expect all input features to vary in a similar way, and
ideally to have a mean of 0, and a variance of 1. Now
we need to scale our data by importing Min Max
Scaler then the accuracy on training set is 0.962 and
accuracy on test set is 0.958.The results are much
better after scaling, and already quite competitive.
After scaling the data now we will again classify the
data with MLP Classifier. After scaling the data,
MLP Classifier is used then the accuracy on training
set is 0.923 and accuracy on test set is 0.895.
44
Fig 8 shows the weights that were learned connecting
the input to the first hidden layer. The rows in this fig
correspond to the 30 input features, while the
columns correspond to the 100 hidden units. Light
colors represent large positive values, while dark
colors represent negative values. One possible
inference we can make is that features that have very
small weights for all of the hidden units are ―less
important‖ to the model. We can see that ―mean
smoothness‖ and ―mean compactness,‖ in addition to
the features found between ―smoothness error‖ and
―fractal dimension error,‖ have relatively low
weights compared to other features. This could mean
that these are less important features or possibly that
we didn‘t represent them in a way that the neural
network could use.
Fig 8 Weight Matrix
VII. CONCLUSION
To analyze medical data, various data mining and
machine learning methods are available. An
important challenge in data mining and machine
learning areas is to build accurate and
computationally efficient classifiers for Medical
applications. In this study, we employed many
algorithms: SVM, ANN, DT, RF, GB, and k-NN on
the Wisconsin Breast Cancer (original) datasets. We
tried to compare efficiency and effectiveness of those
algorithms in terms of accuracy, precision, sensitivity
and specificity to find the best classification accuracy
of SVM reaches and accuracy of 97.13% and out
performs, therefore, all other algorithms. In
conclusion, SVM has proven its efficiency in Breast
Cancer prediction and diagnosis and achieves the
best performance in terms of precision and low error
rate. The results indicated that SVM are the best
classifier predictor with the test dataset, followed by
IITM Journal of Management and IT
ANN and DT. Further studies should be conducted to
improve performance of these classification
techniques by using more variables and choosing for
a longer follow-up duration.
Breast cancer has created a terrible situation in
almost all over the world according to this study and
discussion. It has been observed that the death rate is
gradually coming down in some developed countries
like the UK and US because of the developed
technologies used in diagnosis and awareness. But in
developing countries like India the situation is not
good and some effective steps should be taken in this
direction without any delay [3]. This study has been
made on methodologies by which the breast cancer
can be detected at early stages by using the breast
cancer data set. It is clear from this study that the
Association Rule Mining, Classification, and
Clustering and Evolutionary Algorithms are good at
detection and classification of breast cancer data.
It is also observed that if the properties of the
symptoms are identified correctly, the chances of
accurate detection will improve [8]. It is also
observed by the results of the previous methods that
the classification algorithm increases the
possibility or improved detection accuracy. The
characteristics of breast cancer symptoms
are different, so the chances of good results by
using single algorithm are less. But by the use of
combined algorithms at different levels will produce
good results. So it is concluded that the
framework based on data mining and evolutionary
algorithms can be a milestone in case of breast
cancer detection [10].
Volume 10, Issue 1 ∙January-June 2019
REFERENCES
[1] https://www.omicsonline.org/using-three-
machine-learning-techniques-for-
predictingbreast-cancer-2157
7420.1000124.php?aid=13087
[2] https://www.sciencedirect.com/science/article/p
ii/S2001037014000464https://www.sciencedire
ct.com/science/article/pi i/S1877050916302575
[3] https://www.ncbi.nlm.nih.gov/pmc/articles/PM
C2675494/
[4] https://pubs.rsna.org/doi/pdf/10.1148/rg.301095
057
[5] http://journal.waocp.org/article_31073_4ac28ea
9398b1d19335b6f44a0e79afd.pdf
[6] https://breast-cancer-
esearch.biomedcentral.com/track/pdf/10.1186/b
c r3110.
[7] https://s3.amazonaws.com/academia.edu.docum
ents/32919287/V3I1201402.pdf?AWSAccessK
eyId=AKIAIWOWYYGZ2Y53UL3A&Expires
=1555299017&Signature=rVk6dRW4pUDsqP
v6J Xmm3zL%2B2Sg%3D&response-content-
ition=inline%3B%20filename%3DV3I1201402
. Pdf
[8] https://pdfs.semanticscholar.org/7bf7/3b15b7fd
64c2b01a718a2848b4a3d35b939.pdfhttp://canc
erpreventionresearch.aacrjournals.org/content/c
anprevres/9/1/13.full.pdf
45
 CYBER: Threats in Social Networking Websites and
Physical System Security
Tripti Lamba1, Ashish Garg2
1
Associate Professor, Institute of Information Technology, Janakpuri, New Delhi.
2
Research Scholar, Institute of Information Technology, Janakpuri, New Delhi
triptigautam@yahoo.co.in, ashishgarg518123@gmail.com
Abstract- A social network may be a social system
made up of people or organizations referred to as
nodes, that are connected by one or additional
specific kind of reciprocity, like friendly relationship,
common interest, and exchange of finance,
relationships of beliefs, information or status. A
cyber threat will be each unintentional and
intentional, targeted or non- targeted, and it will
come back from a spread of sources, as well as
foreign nations engaged in undercover work and
knowledge warfare, criminals, hackers, virus writers,
discontent staff and contractors operating inside a
company. Social networking sites don't seem to be Fig.1. General Representation of a CPS
solely speaking or act with people globally, however
Nowadays, innumerable net users frequently visit
conjointly one effective means for business
thousands of social websites to stay linked with their
promotion. In this paper, Tendency to investigate and
friends, share their thoughts, photos, videos and
study the cyber threats in social networking websites.
discuss even regarding their daily-life. In 2003,
the aim of this paper is to review and analyze these
MySpace was launched and within the following
threats of social network and develop measures to
years, several different social networking sites were
shield the identity in cyberspace i.e., security of non-
launched like Facebook in 2004, Twitter in 2006 etc.
public data and identity in social networks are
studied.
Keywords- Cyber threats, Protection, Crime,
Malware, Hackers, Attacks, Breaches, Security.
I. INTRODUCTION
The Term cyber–physical systems (CPSs) emerged
simply over a decade ago as an endeavor to unify the
emerging application of embedded PC and
communication technologies to a range of physical
domains, including aerospace, automotive, chemical
production, civil infrastructure, energy, healthcare,
producing, materials, and transportation. The goal of
Fig2.- Total number of users with respect to different
the CPS program is to reveal crosscutting basic
social platforms
scientific and engineering principles that underpin
the combination of cyber and physical components There are such a big amount of social networking
across all application sectors. sites and social media sites that there's even
computer programs and search engines for them.
These social websites have had positive and negative
impacts.

46 IITM Journal of Management and IT

INTERNET SECURITY THREAT REPORT
2019:
Threat Report takes a deep dive into insights from
the world‘s largest civilian world intelligence
network, revealing:
• Form jacking attacks skyrocketed, with a
mean of 4,800 websites compromised every month.
• Ransom ware shifted targets from shoppers
to enterprises, wherever infections rose 12 %.
• More than 70 million records taken from
poorly organized S3 buckets, a casualty of fast cloud
adoption.
• Supply chains remained a soft target with
attacks flight by 78 %.
• Smart Speaker, get ME a cyber-attack‖ - IoT
was a key entry purpose for targeted attacks; most
IoT devices square measure vulnerable.
This analysis is informed by 123 million sensors
recording thousands of threat events each second
from 157 countries and territories.
Due the actual fact that the quantity of social
network users is increasing day by day, the number
of attacks disbursed by hackers to steal personal data
is additionally raised. Hacked data will be used for
several functions like causing unauthorized messages
(spam), stealing cash from victim's accounts, etc.
Section-1 gives the brief introduction about the need
of cyber security and Threat protection. Literature
Review has been discussed in section-II. Section-III
describes the Applications of Cyber Security. cyber
threats in social networking websites are discussed in
section-IV. Anti-Threat strategies and various ways
can be suggested for circumventing threats related to
social website are discussed in Section-V. Risk
Assessment Methodology are discussed in section-
VI. Section-VII describes the various cyber security
Threats and Trends. The 5-Laws of cyber security
are describes in Section-VIII. Section-IX describes
the Reasons cyber security is more important than
ever and Section-X gives the conclusion of the paper.
Volume 10, Issue 1 ∙January-June 2019
Fig 3. The number of malicious programs targeting
popular social networking sites
Figure-2, shows the Total number of users with
respect to different social platforms and shows the
number of users who actives on social networks.
Another side Figure-3, shows the number of
malicious programs targeting popular social
networking sites The Internet today, unfortunately,
offers to the cyber criminals, many chances to hack
accounts on social network sites and the number of
malicious programs that target the social web sites is
very huge.
II. LITERATURE REVIEW
The popularity of the term social networking internet
sites has been hyperbolic, since 1997, and numerous
individuals currently square measure victimization
social networking internet sites to speak with their
friends, perform business and lots of different usages
per the interest of the users.
The interest of social networking internet sites has
been hyperbolic and lots of analysis papers are
revealed. A number of them mentioned the
protection problems with social networking,
analyzing the privacy and therefore the risks that
threat the web social networking internet sites.
The article [7] identifies the protection behavior and
attitudes for social network users from completing
different human ecology teams and assess, however
these behaviors map against privacy vulnerabilities
inherent in social networking applications.
In the article [8], the scientific highlights the
industrial and social edges of safe and well wise use
of social networking internet sites.
It emphasizes the foremost vital threats of the users
and illustrates the basic factors behind those threats.
Moreover, it presents the policy and technical
recommendations to enhance privacy and security
47
while not compromising the advantages of the
knowledge sharing through social networking
internet sites.
In the article [11], addresses security problems,
network and security managers, which regularly
address network policy management services like
firewall, intrusion, intromission system, antivirus and
knowledge lose. It addresses security, framework to
safeguard corporate info against the threats
associated with social networking internet sites.
Also, several other scientific research papers are Fig 4-. Threats percentage-pose on social networks
revealed, wherever the new technology and methods
were mentioned associated with the privacy and
security problems with social networking websites.
III. APPLICATIONS OF CYBER
SECURITY
• Filtered Communication – Include a firewall,
anti-virus, anti-spam, wireless security, and
online content filtration.
• Protection – Cybersecurity solutions provide
digital protection to your data that will ensure
your employees aren‘t at risk from potential
threats.
Fig. 5- Phishing and Trojan Attacks on different
• Increased Productivity – Viruses can slow down softwares
computers to a crawl, and making work
V. ANTI-THREAT STRATEGIES
practically impossible. Effective cybersecurity
eliminates this possibility, maximizing the This section describes the different types of cyber
potential output. threats in social networks and the possible
contributing factors are also listed below:
• Denies Spyware – Spyware is a kind of cyber
contamination which is intended to behold on • Most of the users aren't concerned with the
your computer operations and deliver that data importance of the private info revelation and
back to the cyber-criminal. therefore they're underneath the danger of over
revelation and privacy invasions.
IV. CYBER THREATS IN SOCIAL
NETWORKING WEBSITES • Users, who are aware of the threats,
unfortunately choose the inappropriate privacy
Lately, social networks attract thousands of users
setting and manage privacy preference properly.
who represent potential victims to attackers from the
following type is shown in figure-4. First Phishers • The policy and legislation aren't equipped
and spammers who use social networks for sending enough to influence every kind of social network
fraudulent messages to victims‘ ―friend‖, threat that are increasing day by day with
Cybercriminals and fraudsters who use the social additional challenges, fashionable and
networks for capturing user‘s data, then carrying out complicated technologies.
their social-engineering attacks and Terrorist groups
• Lack of tools and acceptable authentication
and sexual predators who create online communities
mechanism to handle and influence completely
for spreading their thoughts, propaganda, views and
different security and privacy problems.
conducting recruitment.

48 IITM Journal of Management and IT

• Because of the mentioned factors that cause
threats, following ways can be suggested for
circumventing threats related to social
website
(a) Building awareness, the information
disclosure:
Users most beware and extremely aware concerning
the revealing of their personal information in profiles
on social websites.
(b) Encouraging awareness -raising the academic
campaigns:
Governments got to give and provide educational
categories regarding awareness -raising and security
problems.
(c) Modifying the present legislation:
Existing legislation has to be changed associated
with the new technology and new frauds and attacks.
(d) Empowering the authentication:
Access management and authentication should be
terribly sturdy in order that cyber crimes done by
hackers, spammers and alternative cyber criminals
might be reduce the maximum amount as doable.
(e) Mistreatment the foremost powerful antivirus
tools:
Users should use the foremost powerful antivirus
tools with regular updates and should keep the
suitable default setting, in order that the antivirus
tools might work additional effectively.
(f) Providing appropriate security tools:
Here, we have a tendency to provide
recommendation to the protection software system
suppliers and is that: they need to offer some special
tools for users that alter them to get rid of their
accounts and to manage and management the various
privacy and security problems.
VI. RISK ASSESSMENT METHODOLOGY
The quality of the cyber–physical relationship can
present unintuitive system dependencies. Acting
correct risk assessments needs the event of models
that offer a basis for dependency analysis and
quantifying ensuing impacts. This association
between the salient options among each the cyber
Volume 10, Issue 1 ∙January-June 2019
and physical infrastructure can assist within the risk
review and mitigation processes. This paper presents
a rough assessment methodology parenthetically the
dependency between the ability applications and
supporting infrastructure.
Risk is traditionally defined as the impact times the
likelihood of an event. Likely should be addressed
through the infrastructure vulnerability analysis step
that addresses the supporting infrastructure‘s ability
to limit attacker‘s access to the important
management functions. Once potential vulnerabilities
are discovered, the applying impact analysis ought to
be performed to see accomplished grid management
functions. This data ought to then be wanting to
judge the physical system impact.
A. Risk Analysis
The initial step within the risk analysis method is that
the infrastructure vulnerability analysis. Numerous
difficulties are encountered once crucial cyber
vulnerabilities among system environments thanks to
the high accessibility needs and dependencies on
inheritance systems and protocols.
A comprehensive vulnerability analysis ought to
begin with the identification of cyber assets as well
as code, hardware, and communications protocols.
Then, activities like penetration testing and
vulnerability scanning is utilized to see potential
security considerations among the atmosphere. in
addition, continuing analysis of security advisories
from vendors, system logs, and deployed intrusion
detection systems ought to be utilized to see extra
system vulnerabilities.
B. Risk Mitigation
Mitigation activities should attempt to minimize
unacceptable risk levels. This may be performed
through the readying of a lot of strong supporting
infrastructure or power applications.
Understanding opportunities to concentrate on
specific or mix approaches might gift novel
mitigation ways. Varied analysis efforts have self-
addressed the cyber– physical relationship among the
danger assessment method.
VII. CYBER SECURITY THREATS AND
TRENDS
Phishing Gets a lot of subtle — Phishing attacks,
49
during which rigorously targeted digital messages transportation systems, water treatment facilities,
square measure transmitted to fool folks into clicking etc., represent a significant vulnerability going
on a link that may then install malware or expose forward.
sensitive information, are getting a lot of subtle.
State-Sponsored Attacks — On the far side hackers
Now that workers at the most organizations square
trying to create a profit through stealing individual
measures a lot of alert to the risks of email phishing
and company information, entire nation states square
or by clicking on suspicious-looking links, hackers
measure currently victimization, their cyber skills to
square measure upping the ante — for instance,
infiltrate different governments and perform attacks
victimization machine learning to rather more
on important infrastructure. Cybercrime nowadays
quickly craft and distribute convincing pretend
could be a major threat not only for the non-public
messages within the hopes that recipients can
sector and for people except for the govt and also the
inadvertently compromise their organization‘s
nation as a full. As we tend to come in 2019, state-
networks and systems. Such attacks change hackers
sponsored attacks square measure expected to
to steal user logins, Mastercard credentials and
extend, with attacks on important infrastructure of
different forms of personal money data, moreover as
specific concern.
gain access to non-public databases.
IoT Attacks — The net of Things is changing into a
Ransom ware methods Evolve — Ransom ware
lot of presents by the day (the variety of devices
attacks square measure believed to price victims
connected to the IoT is anticipated to achieve nearly
billions of greenbacks once a year, as hackers deploy
31 billion by 2020). It includes laptops and tablets, of
technologies that change them to virtually enable a
course, however additionally routers, webcams,
person or organization‘s database and hold all of the
house appliances, good watches, medical devices,
data for ransom. The increase of cryptocurrencies
producing instrumentation, vehicles and even home
like Bitcoin is attributable to serving to fuel
security systems. Connected devices square measure
ransomware attacks by permitting ransom demands
handy for shoppers and lots of firms currently use
to be paid anonymously.
them to save lots of cash by gathering huge amounts
As firms still concentrate on building stronger of perceptual information and streamlining business
defenses to protect against ransomware breaches, processes. However, a lot of connected devices
some consultants believe hackers can progressively suggest that bigger risk, creating IoT networks a lot
target different, doubtless profitable ransomware of at risk of cyber invasions and infections. Once
victims like high-net-worth people. controlled by hackers, IoT devices will be wanting to
produce disturbance, overload networks or lock
Crypto jacking — The cryptocurrency movement
down essential instrumentation for gain.Good
additionally affects cyber security in different ways
Medical Devices and Electronic Medical Records
that. For instance, crypto jacking could be a trend
(EMRs) — The health care business continues to be
that involves cyber criminals hijacking third-party
inquiring a significant evolution as most patient
home or work computers to ―mine‖ for
medical records have currently affected on-line, and
cryptocurrency. As a result of mining for
medical professionals notice the advantages of
cryptocurrency (like Bitcoin, for example) needs
advancements in good medical devices. However,
huge amounts of laptop process power, hackers will
because the health care business adapts to the digital
build cash by on the QT piggybacking on somebody
age, there square measure variety of issues around
else‘s systems. For businesses, crypto jacked systems
privacy, safety and cyber security threats. Third
will cause serious performance problems and tear
Parties (Vendors, Contractors, Partners)
down time because it works to trace down and
resolve the difficulty. Cyber-Physical Attacks — A Third parties like vendors and contractors)-
similar technology that has enabled the United States create an enormous risk to companies, the bulk of
of America to modernize and computerize important that doesn't have any secure system or dedicated
infrastructure additionally brings risk. The continued team in situ to manage these third-party workers. As
threat of hacks targeting electrical grids, cyber criminals become increasingly sophisticated

50 IITM Journal of Management and IT

and cyber security threats still rise, organizations are
getting a lot of and a lot of alert to the chance third
parties create. Many years ago, Wendy‘s fell victim
to an information breach that affected a minimum of
one,000 of the fast-food chain‘s locations and was
caused by a third-party merchant that had been
hacked.
Connected Cars and Semi-Autonomous Vehicles
— whereas the driverless automobile is shut,
however, not nonetheless here, A connected
automobile utilizes a board sensors to optimize its
own operation and also the comfort of passengers.
This can be usually done through embedded, bound
or smartphone integration. As technology evolves,
the connected automobile is changing into a lot of
and a lot of prevalent; by 2020, associate degree
calculable 90% of recent cars are going to be
connected to the net.
For hackers, this evolution in automobile producing
and style suggests that yet one more chance to use
vulnerabilities in insecure systems and steal sensitive
information and/or hurt drivers. Additionally, to
safety issues, connected cars create serious privacy
issues.
As manufacturers rush to market with high-tech
automobiles, 2019 will likely see an increase in not
only the number of connected cars, but in the number
and severity of system vulnerabilities detected.
A Severe Shortage of Cyber Security
Professionals — The cyber-crime epidemic has
escalated speedily in recent years, whereas firms and
governments have struggled to rent enough qualified
professionals to safeguard against the growing threat.
Volume 10, Issue 1 ∙January-June 2019
This trend is anticipated to continue into 2019 and on
the far side, with some estimates indicating that there
are some 1 million empty positions worldwide
(potentially rising to 3.5 million by 2021).
VIII. THE 5 LAWS OF CYBER SECURITY
It's time to determine a universal language and
understanding of these foundational facts that govern
our data-security levels.
So, while not additional ruction, here are 5 laws of
cyber security, and whereas there may simply be a
lot of, these 5 can forever be the immutable universal
constants that govern this subject and our existence
in relevance to it.
Law No. 1: If there's A Vulnerability, it'll be
Exploited
―Consider, for a flash that once the primary bank
was formed and designed, there was a minimum of
one person out there UN agency wished to rob it.‖
within the a lot of epoch, since the first ―bug‖ was
found in a system, we‘ve been trying to find ways
that bypass the framework or laws that govern a
trojan horse, a tool or perhaps our society. Think
about that there are those in our society who will
attempt to hack everything at intervals their
capability. This might be obvious with a lot of basic
exploits, just like the one that discovered the way to
impede their car‘s vehicle plate to travel through a
stall for complimentary, or the a lot of obscure, like
infecting a fancy ADPS to derail a remarkable
nuclear weapons program. Finding ways that around
everything for each sensible and dangerous function,
thus present nowadays that we have a tendency to
even have a term for it: ―Life Hacking.‖
Law No. 2: Everything is Vulnerable in a way
We cannot assume that something is off the table and
utterly safe anymore. State-sponsored hacking is a
superb example of this. Government intelligence has
been astonishing over the years in gaining access to
AN opponent‘s systems after they were thought to be
secure. Publicly, we‘ve seen a series huge
information breach over the years from companies
that pay millions annually on cyber defense methods.
Law No. 3: Humans Trust Even after they should
not
Trust, quite honestly, sucks. Yes, it‘s an essential
51
part of the human expertise. A tendency to trust our
vital others, trust by virtue fails in, no matter
religion, and tendency to adhere to and conjointly
trust within the infrastructure around us. An
expectation that the switch can ON the light or that
the mechanic we have a tendency to pay to perform
the automotive in our car can truly know. We cannot
have a functioning society while not a way of trust,
and this is often why it‘s our greatest weakness in
cyber security. People fall for phishing scams,
assume that the computer program they bought for
$20 can flip their PC into Fort Knox (it won‘t) or
believe the shape they‘re filling out is legit (it
typically isn't).
It sounds weird to mention we'd like to combat
thrust; however, we have a tendency to do if we‘re
planning to survive against the nonstop hacking that
takes place.
Law No. 4: With Innovation Comes Opportunity
for Exploitation
The world is full with good people. computer
scientist created a world computing platform to
induce humanity on a similar page. However, with
every innovation and evolution in our technology
comes sure exploits. we have a tendency to sleep in
the age of IoT, and by virtue of this, our lives have,
hopefully, been created higher. one in all the primary
huge samples of this is often the Ring button. It
created, adding a video camera to your front button
simple and extremely simple to watch through a
mobile app. Life was sensible with the clearly
innovative Ring device -- till a security vulnerability
was discovered. the corporate has since mounted that
exploit, however as is usually the case, we have a
tendency to are awaiting consequent vulnerability to
be discovered. And naturally, it‘s created even worse
by Law No. 3.
Law No. 5: Once Unsure
This one isn‘t a cop-out. Each single law written here
comes right down to the easy incontrovertible fact
that despite what the issues or concerns are with
relevance cybersecurity, all of them stem from a
vulnerability of some kind. If we have a tendency to
ever forget this, we have a tendency to do nothing
however posing for the bother.
Our ability to properly defend ourselves comes from
52
understanding that attribute makes these laws
immutable. Once we begin thinking sort of a hacker
is once we will truly stop them, thus here‘s to
hacking the long term along for our own security.
IX. REASONS CYBER SECURITY IS
MORE IMPORTANT THAN EVER
The threat of crime to businesses is rising quick. in
step with one estimate, the damages related to crime
currently stands at over $400 billion, up from $250
billion 2 years past, with the prices incurred by
United Kingdom of Great Britain and Northern
Ireland business conjointly running within the
billions. in a very bid to foreclose e-criminals,
organizations are more and more finance in ramping
up their digital frontiers and security protocols,
however, several are still deferred by the prices, or
by the unclear vary of tools and services on the
market. 5 reasons why finance in cyber security
could be a smart call to form.
The rising value of breaches
The fact is that cyber attacks are extraordinarily dear
for businesses to endure. Recent statistics have
instructed that the common value of an information
breach at a bigger firm is £20,000. However, this
truly underestimates the important expense of
associate degree attack against a corporation. It's not
simply the money injury suffered by the business or
the price of remediation; an information breach can
even communicate much reputational injury.
Suffering a cyber attack will cause customers to lose
trust in a very business and pay their cash elsewhere.
In addition, having a name for poor security can even
cause a failure to win new contracts.
Increasingly Sophisticated Hackers
Almost each business features a website and
outwardly exposed systems that might offer
criminals with entry points into internal networks.
Hackers have a great deal to realize from roaring
knowledge breaches, and there are unnumbered
samples of well-funded and coordinated cyber-
attacks against a number of the most important
corporations within the United Kingdom of Great
Britain and Northern Ireland. Ironically, even
Deloitte, the globe‘s largest cybersecurity adviser,
was itself rocked by the associate degree attack in
October last year.With extremely refined attacks
IITM Journal of Management and IT
currently commonplace, businesses ought to assume
that they'll be broken at some purpose and implement
controls that facilitate them to sight and reply to
malicious activity before it causes injury and
disruption.
Fig-6. Interconnection of different sectors
Above image shows the different-different sectors
are interconnected to each other and people consume
and use various things through online services and
they registered and login itself, that‘s why hackers
easily hack all the information about the people.
Widely accessible hacking tools
While well-funded and extremely masterly hackers
create a big risk to your business, the wide avails of
hacking tools and programs on the net additionally
means that there's additionally a growing threat from
less masterly people. The exploitation of law-
breaking has created it simple for anyone to get the
resources they have to launch damaging attacks, like
ransomware and crypto mining.
A proliferation of IoT devices
More good devices than ever are connected to the
net. These are referred to as net of Things, or IoT,
devices and are progressively common in homes and
offices. On the surface, these devices will alter and
speed up tasks, in addition, as supply larger levels of
management and accessibility. Their proliferation,
however, presents a retardant.
If not managed properly, every IoT device that's
connected to the net may give cyber criminals with
some way into a business. IT services large Cisco
estimates there'll be 27.1 billion connected devices
globally by 2021 – thus this drawback can solely
Volume 10, Issue 1 ∙January-June 2019
worsen with time. To use of IoT devices doubtless
introducing a good variety of security weaknesses,
it's wise conduct regular vulnerability assessments to
assist determine and address risks conferred by these
assets.
Tighter Regulations
It is not simply criminal attacks that mean businesses
got to be additional endowed in cyber security than
ever before. The introduction of laws like the GDPR
means organizations got to take security additional
seriously than ever, or face serious fines.
The GDPR has been introduced by the EU to force
organizations into to taking higher care of the non-
public information they hold. Among the wants of
the GDPR is that the want for organizations to
implement applicable technical and organizational
measures to shield personal information, often
review controls, and find, investigate and report
breaches.
X. CONCLUSION
Social networking community‘s area unit associate
inherently a part of today‘s net. People love
victimising them to remain in touch with friends,
exchange photos, or simply to pass the time once
bored. Firms have conjointly discovered social media
as a brand-new approach of targeting their customers
with relevant info. With user teams with many
countless members, there is a unit forever some
black sheep with malicious intent. We've got seen
several worms unfold through social networks. In
most cases, they need to use social engineering tricks
to post attractive messages on behalf of the associate
infected user. Curious friends Who follow the link
also will get infected with malware and unwillingly
unfold the message further. Many people can click
on nearly any link that they see announce and add
anybody to their personal network that asks, while
not knowing Who extremely is behind it. This
inherent trust, particularly in messages returning
from friends that have had their account
compromised, makes it simple for attacks to succeed,
regardless if it's a phishing attack, a spam run, or a
malicious worm spreading through machine-
controlled scripts.
Some of the newer attacks area unit terribly refined
associated area unit typically arduous to identify for
53
a primitive eye. Use comprehensive security https://csjournals.com/IJCSC/PDF9-
software package to safeguard against these threats. 1/28.%20Divya.pdf. Accessed 3 Aug 2019.
[3] Admin. ―Hacking Social Media. Threats &
You should ne'er share your PIN with others. This
Vulnerabilities- ' Threats & Anti-Threats
includes services that promise to assist you get more
Strategies for Social Networking Websites'.‖
friends or one thing similar. Don't lose manage of
Hakin9, 2 Sept. 2014, hakin9.org/hacking-
your PIN. If you enter your PIN, make sure that
social-media-threats-vulnerabilities--threats-
you're on the original website and not a phishing
anti-threats-strategies-for-social-networking-
scam page that simply sounds like the initial site.
websites/.Espinosa,nick.2018.
Must you suspect that you just have fallen for a
[4] http://www.forbes.com/sitesforbestechcouncil
phishing attack and your account has been
/2018/01/19/the-five-laws-of-
compromised, use a clean system to log into the
cybersecurity/#17c9f4a82265 2019.
initial service and alter your PIN.
[5] http://resource.elq.symantec.com/LP=6
REFERENCES 821?cid=70138000001Qv0FAAS,Vol. 24. ,
2019.
[1] Giraldo, Jairo, et al. ―Security and Privacy in
[6] https://www.quora.com/what-is-cyber-
Cyber-Physical Systems: A Survey of
security-why-is-it-impoetant‖Computer
Surveys.‖IEEE Design & Test, vol. 34, no. 4,
Security", En.Wikipedia.Org, 2019,
2017, pp. 7–17.,
[7] https://en.wikipedia.org/wiki/Computer_securi
doi:10.1109/mdat.2017.2709310.
ty.,"What Is Cyber Security? Definition, Best
[2] Shree, Divya. "Cyber Attack". Social
Practices & More". Digital Guardian, 2019,.
Networking Websites, vol 9, no. 1, 2017, p. 6.,
[8] https://digitalguardian.com/blog/what-cyber-
security.

54 IITM Journal of Management and IT

 Comparative Analysis of Different Encryption Techniques
in Mobile Ad-Hoc Networks (MANETs)
1 2
Apoorva Sharma , Gitika Kushwaha
1,2
Research Scholar, Institute of Information Technology & Management, New Delhi, India
apoorva.sharma098@gmail.com, gitika1512@gmail.com
Abstract - This paper is in depth analysis of Data among the transmission vary will acquire the
Encryption Standard (DES), Triple DES information. a way to stay data confidential is to also
(3DES) and Advanced Encryption Standard (AES) be a threat to confidentiality if the scientific
even secret writing algorithms in painter was done discipline keys aren't encrypted and hold on within
victimization the Network Simulator 2(NS-2) in the node [2].
terms of energy consumption, information transfer
Another challenge once it involves Edouard Manet
time, End-to-End delay time and out turn with varied
security is that the key management issue. so as to
information sizes. 2 simulation models were
stop the malicious nodes from connection within the
adopted: the primary simulates the network
networks, it is necessary to evidence the nodes once
performance assumptive the supply of the common
they square measure connection in Because of the
key, and also the second simulates the network
restricted energy and machine capability of
performance as well as the employment of the Diffie-
MANETs, it is necessary to style a light-weight
Hellman Key Exchange (DHKE) protocol within the
weight and storage economical key management
key management part. The obtained simulation
theme [3] [4].
results showed the prevalence of AES over DES by
sixty fifth, seventieth and eighty three in term of the Numerous security solutions, key management and
energy consumption, information transfer time, and scientific discipline techniques are designed to
network out turn severally. On the opposite hand, the support Manet, a number of them square measure
results showed that AES is healthier than 3DES by custom-made to suit the network necessities
around ninetieth for all of the performance metrics. (minimum delay, minimum power
Supported these results the AES was the suggested Consumption and most throughput) whereas others
secret writing theme square measure famed to be computationally stern.
Keywords - MANET, AES, DES, Key management. They consume a substantial quantity of computing
resources like information measure and power
I. INTRODUCTION
[5].
In recent years, MANETs emerged as a serious next
generation wireless networking technology. There is not any enough data regarding the potency
However, the safety problems on Manet became one of incorporating totally different coding techniques
amongst the first issues. MANETs square measure in unplanned networks.
liable to attacks over wired networks. As a result, This study was done to investigate DES, 3DES and
attacks with malicious goals can invariably devise to AES coding techniques potency and suitableness for
use these vulnerabilities and to disrupt the Edouard MANETs.
Manet operation. The matter display by potential
breaching of the systems by passive observations and Table 1 shows a comparison between these coding
masquerading is more sophisticated by the variable techniques in line with [6].
nature of the wireless atmosphere [1].
Security is provided through security services like
confidentiality. The goal of confidentiality is to
manage or prohibit access to sensitive data to the
sole approved people. Edouard Manet uses associate
degree open medium, therefore sometimes all nodes

Volume 10, Issue 1 ∙January-June 2019 55

 DH algorithmic rule was the primary revealed public
key algorithmic rule by Diffie, and is generally
spoken as DHKE. Several industrial products use this
key exchange technique [7]. The purpose of the
algorithmic rule is to permit 2 users to firmly
exchange a key that may then be used for encryption.
The algorithmic rule itself is proscribed to the
exchange of secret values. The DH algorithmic rule
depends for its potency on the problem of computing
separate logarithms. DHKE algorithmic rule general
steps square measure shown in Figure 1.
Figure 1. Diffie-Hellman Key Exchange Algorithm
General Steps
The rest of this paper is organized as follows; section
a pair of demonstrates the connected add the sphere
of our study. Section three describes the
implementation procedure of the cryptographic
schemes in NS-2. Section four contains experimental
results. Finally, this paper is complete in section five.
II. RELATED WORK
MANET security problems square measure quite
common topic. we are going to survey some analysis
efforts during this topic.
Some researchers targeted on the analysis of the
performance of various coding schemes, others
targeted on the key management and distribution
problems that precede the particular encoding.
Mandal, et al.[8] projected a study that investigated
the 2 most generally used symmetrical coding
techniques DES and AES. The coding schemes had
been enforced victimization MATrix
LABoratory (MATLAB) software package. once the
implementation, these techniques were compared on
some points, were theses points avalanched the result
56
thanks to one bit variation in plain text keeping the
key constant, avalanche result thanks to one bit
variation in key keeping the plain text constant,
memory needed for implementation and simulation
time needed for coding. The authors finished that the
DES coding algorithmic rule incorporates a
disadvantage in term of high memory demand.
Moreover, in AES the avalanche result is
incredibly high in order that AES is good for
encrypting messages sent between objects via
unsecured channels, and is helpful for objects that
square measure a part of financial transactions,
and gave a future direction to incorporate
experiments on alternative sorts of information like
pictures.
Umaparvathi and Varughese in [9] bestowed a
comparison of the foremost unremarkably used
symmetrical coding algorithms AES (Rijndael),
DES, 3DES and Blowfish in terms of power
consumption. A comparison had been conducted for
those coding algorithms victimization completely
different information sorts like text, image, audio and
video. The assorted coding algorithms had been
enforced in Java. Within the experiments, the
software package encrypts completely different file
formats with file sizes (4MB - 11MB). The
performance metrics like coding time, decipherment
time and outturn had been collected. The bestowed
simulation results showed that AES incorporates a
higher performance than alternative common coding
algorithms used. Since AES had not showed any
celebrated security weak points within the bestowed
study, this makes it a wonderful candidate. 3DES
showed poor performance results sinceit needs
additional process power. Since the battery power is
one among the most important limitations in Manet
nodes, the AES coding algorithmic rule is that the
most suitable option.
Sahu and Kushwaha in [10] enforced symmetrical
key coding algorithms DES, AES and Blowfish
victimization NS-2 network machine to check their
performance with completely different information
sorts like text and image supported some
performance metrics. within the experiments, the
algorithms write a special file sorts like text, image
and video sizes (0.3KB - 1KB). The performance
metrics like coding and decipherment time, battery
consumption, residual battery and outturn had been
IITM Journal of Management and IT
recorded for every file sort.
The projected symmetrical key coding algorithms
were enforced victimization NS-2 (v-2.34) with
completely different packet size, the gettable
simulation results showed that AES is easy and
higher in term of residual battery and coding time
than alternative enforced algorithms. Blowfish had
higher performance in term of outturn, however it
consumes additional battery power compared with
the opposite enforced algorithms.
Norouzi, et al. [11] targeted on the improvement of
security performance associate degree exceedingly
|in a wireless impromptu with an coding formula and
transmission rate that planned. Simulation had been
done victimisation MATLAB the input was text
files with minimum size of fifty bytes and most size
used is three hundred bytes,then these information
transmitted victimization 2 modes; with coding and
while not coding. For the primary mode, the
information transmitted while not victimisation any
coding. meantime for the second technique
information transmitted with 3 coding algorithms;
DES, AES and Blowfish. These algorithms were
chosen as a result of they were usually employed in
previous researches. throughout the conducted
experiments only 1 key was accustomed inscribe and
decipher information, that is that the largest size key
within the specific formula.For the coding,
information was encrypted with software system,
Encrypt On Click for AES formula with 256 bit,
Blowfish 2000 for Blowfish formula and Kryplite for
DES formula. supported the input that is distance
and size, time that Accustomed send information to
receiver and turnout might be calculated. All of those
calculation drained the MATLAB programming and
also the output produces time of knowledge transfer.
supported the gained results the authors suggested
selecting AES to attain quick delivery of knowledge
and high turnout, and selecting Blowfish formula
once larger size of knowledge causation with smaller
transmission rate.
Kashani and Mahriyar in [12] analyzed video
streaming characteristics in impromptu networks
victimisation many cryptography algorithms. The
authors conferred associate degree application setup
for secured video streaming in impromptu networks.
Public key infrastructure approach was chosen to
Volume 10, Issue 1 ∙January-June 2019
produce authentication at the network layer. They
planned a completely distributed certification
authority (CA) for Optimized Link State Routing
(OLSR) primarily based impromptu networks. The
initial assumption was that the network contains
predefined special nodes known as shareholders.
Shareholders will generate partial signatures. A node
connection the network, will get a certificate given
that it receives a minimum of k partial signatures
type k completely different shareholders ,a investor
providing service will be known from the
broadcasted how-do-you-do messages.
On the opposite hand, completely different
cryptography schemes were enforced and analyzed
within the study; RC4, 3DES, AES-128, AES-256,
Salsa20-128 and Salsa20-256 and also the time
needed to inscribe completely different sizes of
knowledge were adopted as a performance metric.
The results showed that for RC4, 3DES, AES-128,
AES-256, Salsa20-128 and Salsa20-256 took but
one500 ms to inscribe the 1
MB computer file. 3DES consumes the biggest
coding time followed by Salsa20-256, Salsa20-128,
AES-256, AES-128 and RC4 severally.
Sandhiya, et al. [13] planned associate degree
intrusion detection system named Enhanced
Adaptive ACKnowledgment (EAACK) that consists
of 3 parts; ACK, Secure ACKnowledgment (S-
ACK), and misdeed Report
Authentication (MRA). All the acknowledgement
packets were signed and verified to forestall cast
acknowledgement packets. For linguistic
communication and corroborative the
acknowledgement packets, keys were generated and
distributed earlier. The planned system uses one-hop
ACK that accustomed enhance the misdeed of
detection rates.
To eliminate the need of pre-distributed keys the
planned system thought-about DHKE that depends
on the problem of computing separate logarithms and
permits user to firmly inscribe messages. NS-2
machine tool was used for running simulation, and
also the results showed the advance of misdeed
detection rates which ends up in lower routing
overhead than the prevailing Intrusion Detection
Systems (IDS) once victimisation the DHKE
57
Mechanism.
Du and Xiong in [3] planned a hop-by-hop
authentication and routing driven dynamic key
management them enamed HARD-KM. associate
degree improved Elliptic Curve Diffie-Hellman
(ECDH) protocol with mutual authentication was
accustomed generate 2 combine keys, that were keep
in caches before their expiration.
HARD-KM handling all nodes within the network
equally rather than putt some cluster heads or a base
station within the network, the theme used associate
degree off-line certificate authority (CA) to sign
certificates and distributed authentication materials
matrix for all the mobile nodes.NS2 to machine was
accustomed valuate HARD-KM feasibleness and
potency.
The results showed that HARD-KM key
management theme was resilient to the adversaries
and reduces key cupboard space. The benefits of the
planned key management theme were; neighboring
pair-wise keys on demand creation to save lots of
cupboard space, the pair-wise keys were derived
from associate degree authentication materials matrix
to wear down eavesdropping attack and
compromised nodes had restricted threats to different
uncompromised nodes.
Taneja, et al. [14] planned a standard secret key
institution for even coding over impromptu networks
victim misation DH key agreement protocol. The
idea will be accustomed develop a brand new
routing protocol for MANETs to produce most
security against every kind of attacks. whereas DH
key agreement protocol uses regular system to
inscribe the information associate degreed an uneven
system to inscribe the symmetric keys, the authors
planned a protocol consists of 5 stages; the key
generation and an exchange, shared secret creation,
encrypting victimisation even key and encrypted
information transmission.
CrypTool machine had been employed in modeling
associate degreed testing the DH key agreement
protocol that is an open supply e-learning
application, employed in the implementation and
analysis of cryptological algorithms. As a primary
step in simulation, public parameters should be set.
Since the general public parameters were freely
58
accessible to any or all and thus, not solely supply
and destination area unit ready to access these
parameters rather each third party can also observe a
similar. Once the general public parameters set,
secret numbers of the supply and also the destination
area unit chosen by pushing the button select secrets
in CrypTool. Then the supply sends the shared key
to the destination and contrariwise. As a final step,
the supply and destination produce common and
secret session key by pushing the button generates
common session key in CrypTool. The
implementation of a brand new security
extension and cryptanalytic schemes square measure
written as a brand new implementation within the
NS-2[15].
This section discusses the new security agent and
functions that been wont to simulate the performance
of the encoding schemes of our interest. The NS-2
could be a common separate event machine
developed chiefly for networking analysis. NS-2 is
Associate in Nursing open supply software package
provides wide simulating network sorts, network
applications, routing protocols, information sources
and network components. In NS-2, the system is
sculptured as ordered events that take Associate in
nursing discretionary quantity of your time. NS-2 is
meant having 2 basic building blocks; C++ for the
core practicality that handle processing and therefore
the Object TCL (OTCL) for scripting functions that
is just a special purpose language used for writing
management script to run the simulation.
III. IMPLEMENTATION OF THE
CRYPTOGRAPHIC SCHEMES IN NS-2
The protocol implementation needs the C++
language for packet process. and therefore the use of
script language makes the modification of simulation
configuration quicker and freely adjustable with
dynamic parameters [15].
NS-2 is additionally supported with the Network
AniMator (NAM) that offers a GUI of the network
that's simulated. For MANET, NS-2 provides an
oversized library for circumstantial routing, topology
generators, propagation models, quality models and
information sources. To run any simulation situation
in NS-2, it should be written exploitation TCL script
within the OTCL file [15]. though NS-2 offers
various style alternatives it doesn't provide all. Our
IITM Journal of Management and IT
 enforced cryptanalytic schemes and security The Simulation program accepts four inputs: the
extensions wasn't enclosed within the original NS- secret writing formula, secret writing mode, key and
2, we've got enforced our supply codes and compiled an computer file .once a winning execution, the
viable files and record results supported some cipher text generated.
network metrics [15].
A. Simulation Parameters
The security agent file throughout the protection Along with usual configuration of the wireless
institution method must be feed with the encoding network simulation in NS-2, we have a tendency to
kind from the supply and destination nodes through had set the routing protocol as AODV mistreatment
the TCL file. The encoding kind received from the the command, set val(rp) AODV the macintosh
TCL file hooked up with the encoding kind variable layer, data rate, transmission vary, simulation space,
kind exploitation the bind statement. once a node simulation time, range of nodes and alternative
receives the encoding kind and details conjointly set within the network
therefore the key worth the particular encoding start configuration TCL file. we have a tendency to used
by reading an information file with variable size the AODV routing protocol for power
exploitation the subsequent pseudo code: optimisation, as a result of it needs less
management packets. the main points of the pc
Get pointer to file ("test.txt");
system that we've got wont to compile NS-2 and run
if (not permitted access file) the simulation ar given in Table a pair of, and
therefore the NS-2 simulation parameters that we
return (error);
have a tendency to utilized in our experiments as
read data items from ("test.txt"); shown in Table 3. Table 3
read data as a separate block test
for end of file:
if yes end with read data;
return (done);
IV. SIMULATION AND RESULTS
DISCUSSION
The two main functions of the enforced secret
writing schemes performance analysis we have a
tendency to had exhausted the unplanned network
were; to perform a short study of the enforced
symmetric performance, and to see the overhead that
the DH formula adds to the general network
performance.
During this Chapter we are going to gift the
simulation results that we have a tendency to had
recorded consistent with completely different
performance metrics.
By considering completely different sizes of
knowledge files (2 kilobyte to 64KB) the DES,
3DES and AES (128 key) secret writing algorithms
were evaluated in terms of the energy consumption,
knowledge transfer time and network output. All the B. Simulation Factors and Metrics
implementations were balanced to form certain that The performance of enforced cryptanalytic schemes
the results are comparatively truthful and correct.

Volume 10, Issue 1 ∙January-June 2019 59

within the spontaneous network depends upon many Performance analysis assumptions:
factors:
1. Free house network with no multipath
1. Encryption schemes: This study evaluates 3 and/or attenuation
completely different symmetric algorithms; DES, 2. No noise touching the network
AES (128 key) and 3DES. 3. 20 repetitions for every experiment.
2. Number of hops: within the conducted Results and Discussion
experiments the performance of the enforced
This Section discusses the performance supported the
cryptanalytic schemes was evaluated individually
chosen metrics upon the variable factors
upon 3 main scenarios; one hop, 2 hops and 3 hops
that elaborated within the previous section.
between the supply and therefore the destination
nodes. 1) Energy Consumption
In our experiments the energy consumption was
3.Data file size: the enforced algorithms encipher
evaluated exploitation constant technique delineated
completely different file sizes; 2KB, 4KB, 8KB,
in [16]. we tend to gift a basic value of secret writing
16KB, 32KB and 64KB.
and decoding conferred by the merchandise of the
4. Simulation models: In our study we have a whole range of clock cycles taken by the secret
tendency to applied 2 simulation modes; the primary writing and also the average current drawn by every
mode simulates the network behavior forward the CPU clock cycle. The author in [17] showed the
supply of the common key, and therefore the second price of some secret writing algorithms on Pentium
mode simulates the network behavior as well as the processor as clock cycles per computer memory unit,
key management innovate the link sensing between that we tend to employed in our calculations as
the supply and therefore the destination nodes to shown in Table 4. To calculate the whole energy
confirm areliable and secure key management that value, we tend to divide the price in Amperes for all
precedes the particular coding. secret writing and decoding clock cycles by the
processor clock speed in cycles/sec.
We have performed many tests on our enforced
cryptanalytic schemes to watch its performance For a Pentium processor the clock speed is 7590
exploitation many performance metrics that area unit cycle/sec as shown in [18] that employed in our
outlined in Table 4 calculations as shown in Table 4. The energy value
calculations per computer memory unit done
Table 4. Simulation Metrics
exploitation the subsequent equation, and also the
Metric Definition Energy consumption for various file sizes area unit
The energy The energy consumption is the
Consumption (Joule) average amount of energy shown in figure 2 for DES, 3DES and AES secret
consumed by th encryption and writing schemes.
decryption during algorithm
processing.
The data transfer The time from starting the
time (sec) encryption of the first packet in
a selected data file till the end
of the decryption of the last
encrypted packet that reached
the destination node including
the End-to-End delay time.
End-to-End delay The time taken for a packet to be
time (sec). transmitted across a network
from source to destination.
The network The network throughput that
Throughput (Kb/sec) evaluated by dividing the total
plaintext size that been
encrypted on the total
encryption time consumed
during encryption.

60 IITM Journal of Management and IT

 Figure 2: Energy Consumption for Varying Data File
Sizes
In general the results showed the prevalence of
AES rule over DES and 3DES in term of the energy
consumption
(when inscribe identical knowledge file).
Actually, we tend to found that the
AES needs around sixty fifth, eighty fifth energy less
that the energy consumed by DES and 3DES
algorithms severally.
DES rule consumes around fifty eight
energy but 3DES rule.
2) Data Transfer Time
The data transfer time calculations in our conducted
experiments were supported an
equivalent technique utilized by[11] that thought-
about because the time from beginning the secret
writing of the primary packet in an exceedingly
hand-picked record until the tip of
the cryptography of the last encrypted packet that
reached the destination node together with the End-
to-End delay time. so as to work out the transfer
time the subsequent equation was used:

Table 5. Energy Consumption Results

For the implemented encryption schemes in our

study the transfer time results are shown graphically
in Figure. 3.

Volume 10, Issue 1 ∙January-June 2019 61

Fig.3 : The Implemented Encryption Schemes
Transfer Time Results for the Two Simulation
Modes
As we will notice from Figure 4 a bonus of
victimization the AES secret writing theme is that it
takes less information transfer time than DES and
3DES secret writing schemes. The experimental
results showed that the AES transfer time is some
ninetieth but DES secret writing once running
simulation mode one. On the opposite hand, AES
consumes associate degree some twenty fifth transfer
time but DES secret writing for tiny information files
and (57%-80%) but DES for larger information files
once applying the DHKE rule in simulation mode 2
applied experiments (loading constant information
sizes for each secret writing schemes).
3) Network Throughput
In our study the throughput of the network whereas
running the enforced coding schemes is calculated
exploitation the formula given by [11], that done by
normalizing the whole encrypted file size in bytes by
the information transfer time exploitation the
subsequent formula:
Throughput = size of plain text / time consumed
during encryption
For different record sizes the throughput results
whereas running the 2 simulation modes area unit
shown in Figure 4.
62
Fig.4: Network Throughput Results for the
Implemented Encryption Schemes End-to-End Delay
Time
The End-to-End delay time in our study measured
because the measure from the instant that the supply
node sends a primary packet of information once
secret writing procedure completion till the instant
that the destination node within the network receives
the last encrypted packet. in keeping with the End-to-
End delay definition the DHKE transactions adds a
definite preprocessing time overhead to the particular
End-to-End delay time between supply and
destination nodes now is fastened for DES, 3DES
and AES as a result of it's associated with the transfer
packets throughout session initiation stage and not
the particular encoding. forward totally different
variety of hops between the supply and destination
nodes, and exploitation 16KB record size the End-to-
End delay time results square measure shown in
―Fig. 5‖ for the 2 applied simulation modes. usually
the file size VS. the proportion of the DHKE
overhead is shown in Table 6.
IITM Journal of Management and IT
 Table 7. Performance Evaluation Results Summary
AES DES
AES superiority
Performance superiority superiority
over
over over
Metric DES (%) 3DES (%) 3DES (%)
Energy 65 85 59
Consumption
Transfer Time 70 95 63
Network 83 95 64
Throughput
Security in unintentional networks is associate
degree open analysis issue, and fact-finding work
remains in progress for brand spanking new security
solutions. The scientific discipline solutions, and
their suitableness with unintentional limitations, can
Figure 5 Ad hoc Network End-to-End Delay Time perpetually be a challenge so as to produce
Calculations for 16KB Data protection from malicious attacks. The followings ar
some future work suggestions:
Table 6 The Data File Size VS. DHKE Overhead
•Analyze and measure the performance of another
The following conclusions were obtained:
isobilateral block cipher like the Blowfish cipher.
File Size (KB) DHKE Overhead (%)
•Analyze and measure the performance of stream
1 66.7 cipher cryptography like the RC4 and SEAL ciphers.
2 50 A comparative analysis of stream cipher
4 33.3 cryptography with block cipher cryptography is
8 20 assumed to be valuable.
16 11.1
•Evaluate the performance of the network
From the results shown within the on top of table we victimisation another network machine like Opnet
will conclude that the overhead caused by applying network machine so as to validate the obtained thesis
DHKE protocol to the general Manet performance is results.
suitable scrutiny with its edges particularly for giant
•Evaluate the performance of the network with
information files.
totally different network topologies.
V. CONCLUSIONS AND FUTURE
•Evaluate the performance of the network forward
DIRECTIONS
new nodes joining/leaving the network.
In this study we have a tendency to tried to judge the
performance of DES, 3DES and AES bilateral REFERENCES
coding algorithms beneath painter setting. On the
[1] Nadeem, A. and Howarth, M. P. (2013), A
opposite hand, we have a tendency to applied a
secure key management resolution exploitation the Survey of MANET Intrusion Detection &
Prevention Approaches for Network Layer
DHKE protocol. and eventually we have a tendency
Attacks. IEEE Communications Surveys &
to offered the power to decide on the coding sort by
Tutorials, 15(4), 2027-2045.
the user supported the specified security level.
[2] Chen, J. and Wu, J. (2010), A Survey on
Cryptography Applied to Secure Mobile Ad
hoc Networks and Wireless Sensor networks.
Handbook of Research on Developments and
Trends in Wireless Sensor Networks: From
Principle to Practice, IGI Global, AH

Volume 10, Issue 1 ∙January-June 2019 63

 ALTALHI, 5, 2414-2424. Network. In International Journal of Emerging
[3] Du, D. and Xiong, H. (2011), A Dynamic Key Technology and Advanced Engineering
Management Scheme for MANETs. In Cross IJETAE, 4(6).
Strait Quad-Regional Radio Science and [11] Norouzi, M. esmaeel Akbari, M. and Souri, A.
Wireless Technology Conference (CSQRWC), (2012), Optimization of Security Performance
IEEE, 1, 779-783. in MANET. Journal of American Science,
[4] Mokhtarnameh, R. Muthuvelu, N. Ho, S. B. 8(6).
and Chai, I. (2010), A Comparison Study on [12] Kashani, A. A. and Mahriyar, H. (2014), A
Key Exchange-Authentication Protocol. New Method for Securely Streaming Real-
International Journal of Computer time Video in Ad hoc Networks. Advances in
Applications IJCA, 7(5), 5-11. Environmental Biology, 8(10), 1331-1338.
[5] Abdul, D. S. Elminaam, H. M. A. K. and [13] Sandhiya, D. Sangeetha, K. and Latha, R. S.
Hadhoud, M. M. (2009), Performance (2014), Adaptive ACKnowledgement
Evaluation of Symmetric Encryption Technique with Key Exchange Mechanism for
Algorithms. International Journal of Computer MANET. In Electronics and Communication
Science and Network Security, 8(12), 78-85. Systems (ICECS), 2014 International
[6] Alanazi, H., Zaidan, B. B., Zaidan, A. A., Conference, IEEE, 1-5.
Jalab, H. A., Shabbir, M. and Al-Nabhani, Y. [14] Taneja, S. Kush, A. and Hwang, C. J. (2011),
(2010). New comparative study between DES, Secret Key Establishment for Symmetric
3DES and AES within nine factors. arXiv Encryption over Adhoc Networks. In
preprint arXiv:1003.4085. Proceedings of the World Congress on
[7] Stallings, W. (2006), Cryptography and Engineering and Computer Science (Vol. 2).
Network Security: Principles and Practice, [15] Fall, K. and Varadhan, K. (2002). The NS
(5^th ed.). India: Pearson Education. Manual. Notes and Documentation on the
[8] Mandal, A. K. Parakash, C. and Tiwari, A. Software NS2-Simulator.
(2012), Performance Evaluation of [16] Elminaam, D. S. Kader, H. M. A. and
Cryptographic Algorithms: DES and AES. In Hadhoud, M. M. (2009), Energy Efficiency of
Electrical, Electronics and Computer Science Encryption Schemes for Wireless Devices.
(SCEECS), 2012 IEEE Students' Conference, International Journal of Computer Theory and
IEEE, 1-5. Engineering, 1, 302-309.
[9] Umaparvathi, M. and Varughese, D. K. [17] Biham, E. (Ed.). (2006), Fast Software
(2010), Evaluation of Symmetric Encryption Encryption. 4th International Workshop,
Algorithms for MANETs. In Computational FSE'97, Haifa, Israel, January 20-22, 1997,
Intelligence and Computing Research Springer, Proceedings (Vol. 1267).
(ICCIC), IEEE International Conference, 1-3. [18] Rhett, (1999), x86 CPU Reference, Part 2.
[10] Sahu, S. K. and Kushwaha, A. (2014), Retrieved May 25, 2015, from
Performance Analysis of Symmetric http://alasir.com/x86ref/index.htm
Encryption Algorithms for Mobile Ad hoc

64 IITM Journal of Management and IT

 Fuchsia OS - A Threat to Android
Taranjeet Singh1, Rishabh Bhardwaj2
1,2
Research Scholar, Institute of Information Technology and Management
t23singh@gmail.com , rishabh77k@gmail.com
Abstract-Fuchsia is a fairly new Operating System
whose development was started back in 2016.
Android supports various types of devices which is
having different types of screen size, Architecture,
etc. But problem is that whenever google releases
new updates due to a large variety of devices lots of
devices doesn't receive updates that are the main
issue with android.
This review is about fuchsia and its current Status
and how is it different from the Android operating
system.
Keywords: Internet Of Things( IOT ), Operating
System (OS), Microkernel, Little Kernel, Software
Development Kit (SDK), GitHub
I INTRODUCTION
Fuchsia is an open source Hybrid Real-time
Operating System which is under development.
Prior to Fuchsia we already had android OS which is
used in almost all kinds of devices. Android
development was started in 2003 by android inc later
on google purchased them in 2005. Android‘s first
beta version was released in 2007 by Google and
OHA, and then it was officially accepted by the most
of the companies which were in Open Handset
Alliance [1]. But android wasn‘t dedicatedly
developed for IOT, so Google came up with an idea
to develop an Operating system for Internet of
Things devices.
In 2016 Google was uploading some of the
documentation[2] on the GitHub and at that time and
no-one was aware of this new operating system after
some time google further uploaded some of the other
documents and made it clear that it is a new OS and
its main focus is on IOT devices. Fuchsia runs on
modern 64-bit Intel and ARM processors[1].
Purpose of development (Fuchsia)
Fuchsia is an Open source Operating System
Optimised and developed in a way that it supports
Volume 10, Issue 1 ∙January-June 2019
both personal computers as well as low power
running devices, particularly IOT devices.
Initially, Android was developed for cameras and
then it is extended to other electronic devices,
developing apps for these devices are still a complex
task because of compatibility issues of native
devices.
Android operating system supports various types of
devices such as android wear devices, auto cars,
tablets, smart phones, etc. so to develop an android
app for all these devices is a very tedious task.
The Major problem with android is, not all the
devices receive updates on time.
Fuchsia is developed to overcome these problems,
with fuchsia we can develop apps for all these
devices and they can be implemented flawlessly.
A. Architecture of Fuchsia
Fuchsia uses Microkernel which is an evolution of
little kernel (LK) named as Zircon. Zircon is a much
segmented model and designed for smaller electronic
devices, this is what makes Zircon different from
Android‘s Linux kernel [3]. The biggest problem
with Android Operating System its fragmentation,
there are nearly 840 million versions of android are
available which makes it complex for developers to
decide the targeted versions.
B. Languages supported in Fuchsia for
software development
It supports almost all modern and trending languages
like Swift, C++, Go, Rust, python, which might
attract lots of developers on the environment.
C. Technologies Recommended by Fuchsia
for Software Development
Flutter is Google's mobile UI framework for crafting
high-quality native experiences on iOS and Android
in record time[4].
It has lots of amazing features such as:
65
  Fast Development: Due to hot reload we
can develop apps faster[4].
 Expressive UI: Flutter has built-in Material
Design Widgets, rich API, Smooth natural
scrolling and platform awareness[4].
 Native Performance: Flutter‘s widgets
incorporate all critical platform differences
such as scrolling, navigation, icons and fonts
to provide full native performance on both
iOS and Android[4].
DART [5](recommended programming language for
developing application based on fuchsia): an open-
source project that aims to enable developers to build
more complex, highly performant apps for the
modern web.
Using the Dart language, you can quickly write
prototypes that evolve rapidly, and you also have
access to advanced tools, reliable libraries, and good
software engineering techniques. Meanwhile, dart is
also receiving a little criticism for being a modern
language that doesn‘t handle Null references.
II CURRENT STATUS OF FUCHSIA OS
Fuchsia is still under development and may take few
more years for being operational. Fuchsia is not
google‘s 20% project. Armadillo was the first
renderer engine of fuchsia but now it has been
replaced by Escher which is Vulkan based rendering
engine. Escher provides faster graphics rendering.
Google has released Fuchsia SDK and anyone can
download and execute it.
Currently, Google is using HonorPlay[6] to
demonstrate the very first look of Fuchsia, and a web
Interface demo[7] by Manuel Goulão[8]which has
attracted the audiences due to the spectacular User
Interface.
A. Scope
Fuchsia is an operating system which is primarily
built to support Internet Of Things devices and
Internet Of Things is a booming industry, more and
more devices are connected with each other but we
currently don‘t have any operating system so Fuchsia
definitely have huge scope in near future.
Fuchsia uses Flutter Software development kit which
is used to create native appsso developers can
66
develop fuchsia apps android apps and ios apps via
using single code base which is kind of amazing
thing because nowadays we have to write codes for
different platforms individually, so fuchsia has huge
scope when it comes to native software development
for different devices within same software
development kit using Common language for all.
B. Fuchsia: A threat to android?
As observed above that android‘s architecture has
some issues with compatibility with smaller devices.
It is observed on the basis of Google docs regarding
the Fuchsia OS code[9] that Android apps will run on
Fuchsia too. But regarding this, there is no official
statement has been made by Google that Fuchsia will
completely make Android obsolete.
Google wants to peg Fuchsia as a unifying OS, the
single operating system, the company had been
mulling since 2015. But at this point, it is just a
kernel – the core of an operating system.
C. Fuchsia’s effect on Windows
Implementation of fuchsia seems a nice strategy to
catch up with Microsoft‘s Windows 10 IoT
Enterprise [10]. However, Microsoft has sheer
dominance in Desktop oriented Operating system.
And Fuchsia‘s Target is to dominate smaller devices
like Phones and Tablets.
Microsoft is also working on similar projects to
create a new unified mobile platform, for example,
the Universal Windows Platform and Continuum,
this basically enable the apps to resize and formulate
the interface for different screen sizes and
orientations. Fuchsia‘s objective is also something
similar to handle portability based on different
requirements. If Google successfully develops a new
OS, the market could definitely shift, experts
indicate. When Microsoft Windows Mobile failed to
take off, they left the field open for Google and
Apple.
III OBSERVATION
Fuchsia is designed to have the ability to scale up
from small devices like watches, mini cameras to
larger devices like laptops and computers.
The recommended language by google for fuchsia
i.e. DART is highly preferred for prototyping but can
IITM Journal of Management and IT
also receive little criticism for being a modern REFERENCES
language that doesn‘t handle Null references[11].
[1]
Dart in reference with billion dollar mistake (Null [2]
References)
[3]
Dart being a modern language is missing null able
[4]
types. That might kill it as a modern language, and
[5]
do not solves the Billion Dollar Mistake[12] referred
[6]
by Tony Hoare[13](British Computer Scientist -
inventor of quick sort), null references historically do
not have positive references If the billion dollar
[7]
mistake was the null pointer, the C gets function is a
[8]
multi-billion dollar mistake that created the
[9]
opportunity for malware and viruses to infiltrate[11].
This mistake was recovered in many modern
[10]
language like Kotlin, C# etc.
IV CONCLUSION [11]
It is intended that fuchsia will run on everything from
phones and laptops to IoT devices and more.
Fuchsia is going to accomplish much of what [12]
Microsoft and Apple already have in Windows 10
and iOS-to-macOS Sierra Continuity, respectively, [13]
but in a very Google way.
Volume 10, Issue 1 ∙January-June 2019
https://www.openhandsetalliance.com/
https://android.googlesource.com/platform/preb
uilts/fuchsia_sdk/
https://github.com/fuchsia-mirror
https://flutter.dev/
https://www.dartlang.org/
https://www.digit.in/news/mobile-
phones/google-fuchsia-os-successfully-booted-
on-kirin-970-powered-honor-play-44842.html
https://mgoulao.github.io/fuchsia-web-demo/
https://mgoulao.github.io/cv.html
https://fuchsia.googlesource.com/third_party/an
droid/platform/frameworks/native/
https://developer.microsoft.com/en-
us/windows/iot
https://www.infoq.com/presentations/Null-
References-The-Billion-Dollar-Mistake-Tony-
Hoare
https://www.youtube.com/watch?v=kz7DfbOuv
OM
https://www.cs.ox.ac.uk/people/tony.hoare/
67
 Sentiment Analysis using Lexicon based
Approach
Rebecca Williams1, Nikita Jindal2, Anurag Batra3
Research Scholar, Institute of Information Technology & Management, Janakpuri, New Delhi, India
rebecca3williams@gmail.com, nikitajindal68@gmail.com, anuragbatra1999@gmail.com
Abstract - Triple talaq is also known as talaq-e-
biddat instant divorce. It is a kind of Islamic divorce
used by Muslims in India. It allows Muslims man to
divorce their wife legally by simply stating the word
‗Talaq' three times in any form which can be in any
way (verbal, written, or in electronic form). Now a
day, the huge amount of data is posted on daily basis
on the social media platform. Twitter is a well known
social networking platform where the user can post
their views, opinions, and thoughts freely. The
sentimental analysis is a process of understanding
opinions, thoughts and feelings of people about a
given subject. This paper analyses tweets posted on
Twitter on the subject Triple from the year 2002 to
the year 2019. We have transformed unstructured
data into well-informed data for getting the insights
of people. The main focus of the work is to analyze
the feelings of people using two well-known API like
TextBlob, and SpaCy. These APIs are based on
Lexicon approach. This paper predicts sentiment into
three classes positive, negative and neutral.
Keywords: - Talaq, ApplicationProgramming
Interface(API),SpaCy,TextBlob,Parts-of-Speech
(POS),Natural Language Processing(NLP)
I. INTRODUCTION
Triple talaq also named as talaq-e-biddat instant
divorce. It is an Islamic divorce in the Muslim
religion. It allows Muslims man to divorce their wife
legally by simply stating the word ‗Talaq‘ three
times in any form which can be in any way (verbal,
written, or in electronic form).22nd August 2017 was
the date when the Indian supreme court instantly
deemed the triple talaq. Out of five judges, three
judges have the same opinion that triple talaq is
illegal and the remaining two stated that the
government should ban this practice by simply
following the law. The Modi government made a bill
known as the Muslim women bill, 2017 and it was
passed on 28th December 2017 by the Lok Sabha.
The bill stated that if a man gives instant triple talaq
68
in any form- spoken or written or by any electronic
means like by email, message or any other mean of
communication will be considered illegal and if any
such practice found then there will be a three years
imprisonment and will be fined. In Islam marriage is
considered as a contract between husband and wife
and in that various procedures have been written on
how to annul it. As per Islamic traditions, a woman
can ask divorce via "khula", whereas the husband can
end the marriage instantly by pronouncing talaq
thrice. But many have highlighted the misuse of
instant divorce by men as a reason to ban it. The man
will pay some maintenance and custody of the child
to the mother. The bill passed should not be viewed
from point of politics. The bill passed should not also
be viewed from the point of religious motive or for
vote bank. The bill is passed for the rights and
respect of women. The sentimental analysis is a
process of analyzing views of people about a given
subject or a topic which can be in the form of written
or spoken language. Today in this world where a
huge amount of data generated every day,
sentimental analysis has become a vital tool for
making sense of each processed data. This has
allowed companies to get the results of various
processes they are doing nowadays. It is a type of
text research or mining. In this paper, we are
applying statistics, natural language processing
(NLP), and machine learning to identify, analyze and
extract some important information from tweets. The
main objective is to observe the reviewer‘s feelings,
expressions, thoughts or judgments about a particular
topic, events, or a company. This type of analysis is
also known as opinion mining which has its main
focus on extraction. The goal of this analysis is to get
the opinion of our audience on a particular subject by
analyzing a large amount of data from heterogeneous
sources. Today sentimental analysis has a different
number of uses. With the increase in the use of social
networking sites and the rise in feedbacks forms and
ranking sites, companies are becoming more
IITM Journal of Management and IT
interested in this type of analysis. Consumers can
freely share their thoughts easily on the web. But
with a huge amount of information out there, it can
be tedious for companies to hone in on the most
valuable parts of consumers comment.
This is the main reason why we are applying
sentimental analysis. Organizations are leaning on
the basis of sentiment analysis and then filtering out
valuable information so that they successfully
understand consumers behavior on a topic so that
they can take targeted decisions. Now a day, the
huge amount of data is posted on daily basis on the
social media platform. Twitter is a well known social
networking platform where the user can post their
views, opinions, and thoughts freely.
Fig 1. Flow chart of the process
Sentiment Analysis can be done by either machine
learning or lexicon-based approach. In this paper, we
have applied a Lexicon based approach. This is a
feasible and practical approach which can analyze
tweet text without training or using machine
learning. Lexicon is a collection of words or one can
say it is like a dictionary in which words are arranged
alphabetically. This approach is subdivided into a
dictionary-based approach and corpus-based
approach. Here we are using a corpus-based
approach. Corpus is a large body of words or text
which formulate a set of conceptual rules that govern
a natural language from texts in that language and
examine how that language relates to other
languages.
Volume 10, Issue 1 ∙January-June 2019
II. RELATED WORK
In the last few decade there has been the vast
expansion of social media due to which social media
has started playing a vital role when it comes to
gathering sentiments of masses. Something similar
was done there are various ways to perform
sentiment analysis using lexicon based approach and
some of the methods are discussed by Taboada, M.,
Brooke, J., Tofiloski, M., Voll, K., & Stede, M.in
their research paper Lexicon-based methods for
sentiment analysis.
III. EXPERIMENTAL SET-UP
The experimental setup of the approach presents the
research methodology employed, the tools, and
libraries used to analyze the opinion of people of
India on Triple Talaq. We used a laptop of DELL, i3,
3.6 GHz with 8GB DDR3 RAM. In this study, open
source libraries, packages, APIs are extensively used.
This section is further categorized into two sub-
sections.
A. System Architecture
In this section, we are discussing the overall
architecture of a system. The tweets that were
extracted do not contain their corresponding labels.
So, labeling of tweets was required in order to get the
desired results. In this paper, we used a lexicon-
based approach to classify our tweets. Currently,
many web services are available which automatically
provide more clear-cut labels as compared to human
annotators. The main focus of the work is to analyze
the feelings of people using four well-known API
like Text Blob, NLTK, SentiWords, and Vivekn. The
services provided by these four approaches vary
from one another.
a. Data Extraction
With an increase in the importance of text analysis in
many research areas, so many researchers started
analyzing the sentiments of people by their posts on
many social sites. In this paper, we are analyzing the
tweets posted by people on triple talaq. The first of
analyzing is data extraction. Data extraction is the
process of retrieving the data posted by people from
available sites (in this research we are taking data
from twitter). After this we preprocessed our data
i.e., we cleaned our data (removed noisy data).
69

b. Labeling of Tweets
In this part, we have labeled our tweets using four
well known APIs. The four well-known API that we
have used over here are TextBlob, SentiWord,
NLTK, and Vivekn. These four APIs label our tweets
into three well-known categories (Positive, Neutral
and Negative).
c. Data Visualization
In this section, we discussed how after applying all
our techniques we Visualize all our gathered
information.
● Pie charts of our results of APIs
In this, we have visualized how many tweets are
classified into which category(positive, negative or
neutral) in the form of pie charts for each API.
● Comparison of our APIs used
In this, we have compared all our four API and have
determined the accuracy of each API that we have
used. This accuracy has been shown using a
histogram.
B. Tools Used
In this section, we are discussing the various tools
that we have during this research. Here we are
discussing programming languages, APIs, open
source libraries in brief.
● Tweepy
Tweepy is open source which enables python to
communicate with Twitter and enable us to use its
API.
● Python programming language
Python is a popular programming language which
nowadays is being used for text mining and analysis.
It is object-oriented and high-level programming
language.
● Text Blob
It is a python library for textual data. Using text blob
we can tokenize our paragraph into sentences or
words.
70
● SpaCy
It is a free and open source library which is used for
advanced NLP ( natural language processing).
● Matplotlib
It is a python plotting library (in 2D). It produces
quality figures in a variety of formats. It also helps in
plotting various graphs for data visualization.
● Seaborn
It is a python library for making statistical graphs.
IV. Data Extraction
A. Extraction of Data
It is one of the most important step cause from here
only things start. Now Twitter is the most famous
social network where people from all around the
world share their views which are also called tweets.
Twitter helps us to access its API through python
library called Tweepy which allows us to extract the
data from the twitter of any user. Tweepy provides
an easy to use Cursor interface by which we can
iterate through different types of objects. This python
library can be installed by using pip command on
cmd or terminal.
The first thing we need to know are the keys, that
are: the consumer key, the consumer secret key, the
access key and the access secret key from Twitter's
developer site available easily for each user. These
keys will help the API for authentication.
Steps to obtain these keys are as follows :
● Log in to Twitter's developer site.
● Go to ―Create an App‖.
● Fill the details in the form.
● Then Click on ―Create your Twitter
Application‖.
● Details of our new application will be shown
along with consumer key and consumer
secret key.
● To obtain access token key, click on ‖Create
my access token‖.
The page will refresh and generate an access token.
For authorization of our twitter account, we use
OAuth Interface. By this, we can authorize our app
to access our account.
Now, Twitter limits the tweets to be extracted to
IITM Journal of Management and IT
3200 maximum at a time. A JSON file is created
when we fetch tweets from Twitter using tweeps. csvWriter.writerow([tweet.created_at,
Since JSON file data is complex and it contains a lot tweet.text.encode('utf-8')])
of unrequired data, we created a corpus file, which
B. Pre-Processing of Data
contains only required data which can be analyzed.
This is another important step to proceed further is
We also converted time to the timestamp to fetch
preprocessing. Preprocessing of data is required to
data between the year 2002 to 2019. Timestamp data
clean the data to acquire the required data. In this
is easily read by machine. We have Created separate
step, all the noisy characters are removed from the
files for name and text to improve the performance
text to further analyze it. The misspelled words,
of the code.
grammatical errors, punctuation errors, unnecessary
Code snippet for extracting tweets capitalization, stop words and use of non-dictionary
words such as abbreviations or acronyms of common
import tweepy
terms are few examples of noise in the text.
import csv
import pandas as pd A classic tweet contains variations of words,
def TWITTER(Save): emoticons, mentions of users, hash tags etc. The
#input your credentials here main goal of the preprocessing step is to standardize
CONSUMER_KEY = input("Enter the text into a relevant form to derive the sentiments
Consumer Key: ") of the user. Following are the steps to pre-process
C0NSUMER_SECRET) = input("Enter text into useful data for classification:
Consumer Secret Key: ") a. Tokenization-
ACCESS_TOKEN= input("Enter Access
First of all the text is tokenized. Tokenization is the
Token: ")
process of natural language processing(NLP) by
ACCESS_TOKEN_SECRET = input("Enter
which large textual data is divided into smaller parts
Secret Access Token: ")
called tokens. In other words, tokenization helps to
subdivide sentences into a group of words and
AUTH =
paragraphs in a group of phrases. This step is a
tweepy.OAuthHandler(CONSUMER_KEY,
crucial step in NLP.
C0NSUMER_SECRET)
AUTH.set_access_token(ACCESS_TOKEN Tokenization can be of two types:
, ACCESS_TOKEN_SECRET)
i) word tokenization
API =
ii ) sentence tokenization
tweepy.API(AUTH,wait_on_rate_limit=True)
file = STANDARDIZED_CSV(Save) We are using nltk word_tokenizer() to tokenize our
csvFile = open(file, 'a') textual data split a sentence into words. Then the
output of the tokenization is converted into a data
Code snippet to tokenize content
frame.
import nltk
from nltk.tokenize import word_tokenize Now, tokenization of text from the corpus is done in
m = nltk.word_tokenize(tex[1].strip().lower() three ways using nltk i.e. unigram, bigram, and n-
#Using csv Writer gram. These text models can also be used in
temp0 = get_input() tokenized sentences.
for temp1 in temp0:
csvWriter = csv.writer(csvFile) b. POS Tagging-
for tweet in
tweepy.Cursor(api.search,q=temp1,count=10, The second step of pre-processing the data is POS
lang="en", since="2001-04-03").items(): Tagging. Parts-of-speech Tagger is very useful as it
print (tweet.created_at, tweet.text) reads the text and assigns parts of speech or tokens

Volume 10, Issue 1 ∙January-June 2019 71

(i.e. noun, pronoun, verb, adjective, etc.) to each
word.
Code example:
text = word_tokenize("This sentence is written to
check pos tagging in nltk")
nltk.pos_tag(text)
Output: [('This', 'DT'),('sentence', 'NN'),('is',
'VBZ'),('written', 'VBN'),('to', 'TO'),('check',
'VB'),('pos', 'NN'),('tagging', 'VBG'), ('in', 'IN'),('nltk',
'NN')]
c. Lemmatization:
The third step of preprocessing is Lemmatization. It
is an algorithmic process of finding the lemma of a
word depending on its meaning. Lemmatization
usually refers to the linguistic analysis of words. The
main aim of this process is to remove any inflection
in the ending of a word.
Now in text pre-processing both stemming as well as
lemmatization. They both seem similar but are
different because stemming method cuts the suffix
from the word i.e. either the ending or the beginning
of a word which sometimes makes the word
meaningless. For example: Stemming for studies is
studi , which indeed have no meaning in the
dictionary.
On the other hand, lemmatization is a much better
method and is more powerful as it also considers the
morphological analysis of a word which helps in
conversion of the word into its base form without
changing its meaning. For example Lemma for
studies is study.
So we can say that lemmatization is a smart method
and stemming is a generic method. Hence,
lemmatization will help in creating better machine
learning characteristics.
Code Example:
from nltk.stem import WordNetLemmatizer
wordnet_lemmatizer = WordNetLemmatizer()
wordnet_lemmatizer.lemmatize("teaches")
Output: 'teach'
Also, for verbs, we use ―v‖ as an argument to pos as
the default pos argument is ―n‖.
72
Code Example:
wordnet_lemmatizer.lemmatize("is", pos="v")
Output: 'be'
d. Stop words removal
Stop words removal is one of the major pre-
processing steps as it is used to filter out useless data.
In natural language, stopwords are the frequently
used words such as is, am, are, an, the etc which
have very little meaning. These words are ignored by
the search engine while indexing entries for
searching and retrieving them. The programming
languages are programmed to ignore such words. We
are removing these words as they are considered as
they do not add any value to our analysis.
Code snippet for removing stop words and
punctuation in the tweet data
import math
from nltk.corpus import stopwords
self = ["`","``","''"]
stop_words = set(stopwords.words("english") +
list(string.punctuation) + list(self))
e. Translation and Language Detection
Last but not least this step is used to detect and
translate a given language into a required language.
We are using Textblob which is a python library for
this task. Now, textblob is an amazing tool which
makes NLP faster and easier to work with and this
translation feature of textblob is one of the best
features. Let us do some code examples and see how
it works:
Code Example:
from textblob import TextBlob
blob= TextBlob("‫)"مرحبا بالعالم‬
blob.detect_language()
Output: ‗ar‘
Code Example:
blob.translate(from_lang="ar", to='en')
Output: TextBlob("Hello World")
V. Labeling of Data
This step is further subdivided into four steps
including various APIs to label the text into negative,
positive or neutral. The APIs we are using are -
IITM Journal of Management and IT
TextBlob, Nltk, SentiWord, and Vivekn. We are tem = 1
using these four APIS to label the tweet text and then text_neg = text_neg + 1
we will compare and contrast them. These APIs are
B. SpaCy
freely available and are usually used in the lexicon-
based approach. The pre-processing steps are almost It is a free and open source library which is used for
common in all of these APIs i.e. tokenization, POS advanced NLP ( natural language processing). Spacy
tagging, stemming and lemmatization, stop words is gaining popularity at a very fast rate and is said to
removal etc. Vivekn labels sentiment using words, n- overtake NLTK. SpaCy is lightning fast, highly
grams, and phrases whereas TextBlob uses Parts-of- accurate and easy to run. It also works well with
Speech (POS) tagging to label the sentiments. Four other tools like TensorFlow, Scikit-Learn, PyTorch
python scripts are written for labeling the sentiment and Gensim and provides models for Named Entity
by accessing their APIs. Recognition, Dependency Parsing and Part-of-
Speech tagging. This library works best while
A. TextBlob
preprocessing data for deep learning. Some of its
TextBlob is a python library and it helps in various other features include pre-trained word vectors,
natural language processing tasks providing a simple support more than 31 languages and easy model
API. packaging and deployment. State-of-the-art speed is
the best unique feature and spaCy v2.0 features
Textblob has easy to use interface hence it is
neural models for tasks such as tagging, parsing and
beginner friendly. It is fairly simple to learn Textblob
entity recognition.
when compared to other open source libraries which
is one of its key feature. Code snippet:
import spacy
It is basically used for text processing which includes
from spacy.tokens import Doc
tasks such as tokenization, Parts-of-Speech tagging,
from nltk.sentiment.vader import
stemming, lemmatization, stopwords removal,
SentimentIntensityAnalyzer
translation, noun phrase extraction, sentiment
sentiment_analyzer =
analysis by labeling, classification by using machine
SentimentIntensityAnalyzer()
learning algorithms, and more. It has a sentiment
def polarity_scores(doc):
property which returns a tuple of the form Sentiment
return sentiment_analyzer.polarity_scores(doc.text)
(polarity, subjectivity). The polarity score is a
Doc.set_extension('polarity_scores',
floating point number within the range [-1.0,
getter=polarity_scores)
1.0](where -1 means negative statement and 1 means
for min tweet: tem =3
positive statement). The subjectivity is a floating
doc = nlp(m)
point number within the range [0.0, 1.0] (where 0.0
if doc._.polarity_scores["pos"] ==
is very objective and 1.0 is very subjective). The
doc._.polarity_scores["neg"]: tem = 0
subjectivity is basically the person‘s opinion or
spacy_confused = spacy_neg + 1
emotion on a particular topic. If you want to work on
elif doc._.polarity_scores["neg"] >
basic NLP tasks, TextBlob is the best open source
doc._.polarity_scores["pos"]: tem = 1
software. In fact, TextBlob performs better than
spacy_neg = spacy_neg + 1
NLTK for textual analysis.
elif doc._.polarity_scores["pos"] >
Code snippet: doc._.polarity_scores["neg"]: tem = 2
for m in tweets: tem = 3 spacy_pos = spacy_pos + 1
analysis = TextBlob(m)
VI. Visualization
if analysis.sentiment.polarity > 0: tem = 2
text_pos = text_pos + 1 Data in visual form makes more sense than data in
elif analysis.sentiment.polarity == 0: tem = 0 textual form. The representation of data using charts
text_confused = text_confused + 1 elif and graphs is more presentable and helps understand
analysis.sentiment.polarity < 0: facts and figures better.

Volume 10, Issue 1 ∙January-June 2019 73

In this step, we are using statistics and mathematical
functions to organize our data in graphical format.
Data visualization helps us in understanding
changing patterns or trends in the data over time. It
also helps in the comparison of the different sets of
data.
We are using the following three graphs for basic
data visualization:
● Line Plot
● Bar Chart
● Histogram Plot
● Pie Chart
We will do the following analysis and visualize the
Fig 3. Pie chart of data labelling using
data in forms of various graphs.
TEXTBLOB
A. Pie charts of our results of APIs
In this, we have visualized how many tweets are
classified into which category(positive, negative or
neutral) in the form of pie charts for each API.
B. Bar graph of our results of APIs
In this, we have visualized how many tweets are
classified into which category(positive, negative or
neutral) in the form of bar graph for each API.
C. Comparison of our APIs used
In this, we have compared our APIs and have
determined the accuracy of each API that we have
used. This accuracy has been shown using a
histogram.
Fig 4. Pie chart of data labelling using SpaCy
● Sentimental analysis Visualization Results
Using various API’s
Pseudocode: To create pie chart
# Data to plot
labels = 'Positive', 'Negative', 'Neutral' sizes =
[text_pos,text_neg,text_confused]
colors = ['gold', 'yellowgreen', 'lightskyblue'] explode
= (0.1, 0, 0) # explode 1st slice
# Plot
plt.pie(sizes, explode=explode, labels=labels,
colors=colors, autopct='%1.1f%%',
shadow=True, startangle=140)
plt.axis('equal')
plt.title("Sentimental Analysis Using TEXTBLOB ",
bbox={'facecolor':'0.8', 'pad':5})
plt.show()

74 IITM Journal of Management and IT


Fig 5. Bar graph of SpaCy and Textblob APIs
Fig 6. Histogram of SpaCy and Textblob APIs
showing comparison
These Pie charts shows Results For SpaCy and
Textblob APIs. SpaCy shows better results than
Textblob. As neutral is less in percentage. Also,
SpaCy took more time than Textblob for analysis.
Hence, SpaCy is slow. To get the most accurate and
final results we created the text file which contains
the result for each individual tweet and we take the
mode of both API to check weather the tweet is
Positive, Negative or Neutral.
Table 1. Difference between Textblob and SpaCy
on the basis of recent Repository
Repository TextBlob SpaCy
Stars 6,130 13,094
Watchers 283 537
Forks 805 2,214
Volume 10, Issue 1 ∙January-June 2019
Last Commit About 2 months ago 2 days ago
Code Quality L3 -
Language Python HTML
License MIT License
Tags Text Processing, Natural
Natural Language Language
Processing, Linguistic Processing,
Scientific,
Engineerin
g
VII. CONCLUSION
In this paper, we used Twitter's tweets on triple talak
and analysed how positive or negative or neutral a
tweet is. We have used two well-known APIs i.e.
Textblob and SpaCy. We compared the results of
two API and found that Textblob is faster than
SpaCy. But SpaCy produced more accurate results
(refer fig 3. and 4.). With gentle learning curve and
surprising amount of functionality Textblob has
become one of the best language on beginner level.It
makes text processing simple by providing an
intuitive interface to NLTK. TextBlob can be used
for initial prototyping for almost every NLP project.
On the other hand, SpaCy is the new kid on the
block, and it is becoming quite sensational. It‘s
marketed as an ―industrial-strength‖ Python NLP
library that‘s geared toward performance. SpaCy is
minimal and opinionated, and it provide you with
plenty of options like NLTK does. It give the best
algorithm for each purpose so we don't have to waste
time on choosing an optimal algorithm and we just
have to focus on productivity. SpaCy is built on
Cython and lightning-fast but not as fast as
TextBlob. SpaCy is known as ―state-of-the-art,‖ but
its main weakness is that it currently only supports
English.
SpaCy is new but growing at good pace and will
probably over take NLTK. If one is building a new
application or revamping an old one then one must
try SpaCy.
REFERENCES
[1] Taboada, M., Brooke, J., Tofiloski, M., Voll, K.,
& Stede, M. (2011). Lexicon-based methods for
75
 sentiment analysis. Computational linguistics,
37(2), 267-307.
[2] M. Ravichandran, G. Kulanthaivel, and T.
Chellatamilan.Research Article on Intelligent
Topical sentiment Analysis for the
Classification of E-Learners and Their Topics of
Interest, Hindawi Publishing Corporation.
The Scientific World Journal.Volume 2015,
Article ID 617358, 8 pages
[3] Sudhir Kumar Sharma,Sentiment Predictions
Using Deep Belief Networks Model for Odd-
Even Policy in Delhi.International Journal
of Synthetic Emotions , Volume 7 • Issue 2 •
July-December 2016
[4] Ana ̈ıs Collomb ,Crina Costea ,Damien Joyeux
,Omar Hasan and Lionel Brunie .A Study and
Comparison of Sentiment Analysis Methods for
Reputation Evaluation. University of Lyon,
INSA-Lyon, F-69621 Vil leurbanne, France
76
[5] Braja Gopal Patra, Dipankar Das, and Amitava
Das.Sentiment Analysis of Code-Mixed Indian
Languages: An Overview of SAIL_Code-
Mixed Shared Task @ICON-2017.Article ·
March 2018
[6] Prabu Palanisamy, Vineet Yadav and Harsha
Elchuri,Serendio: Simple and Practical lexicon
based ap proach to Sentiment Analysis.
Serendio Software Pvt Ltd Guindy, Chennai
600032, India
[7] Haseena Rahmath P, Tanvir Ahmad. Sentiment
Analysis Techniques - A Comparative Study.
IJCEM International Journal of Computational
Engineering & Management, Vol. 17 Issue 4,
July 2014 25 ISSN (Online): 2230-7893
[8] Mining the Social Web, 3rd Edition.Book by
Matthew A. Russell, Mikhail Klassen
IITM Journal of Management and IT