IEEE Sponsored 2nd International Conference on Innovations in Information Embedded and Communication Systems ICIIECS'15

Privacy preserving heuristic approach For

Association Rule Mining in Distributed
Database
Bhoomika R Mistry Amish Desai
Computer Science and Engineering Computer Science and Engineering
Parullnstitute of Technology, Vadodara Assi.Prof Parullnstitute of Technology, Vadodara
Gujarat,lNDIA-391760 Gujarat,lNDIA-391760
bhumikamistry92@gmail.com desaiamishI986@gmail.com

Abstract:
Association rule mmmg is a powerful model of data is some technique to deal with security .Privacy
mining used for finding hidden patterns in large preserving in data mining is one of the technique that
databases. The challenges of data mining is to secure deal with security of the knowledge that extracted by
the confidentiality of sensitive patterns when releasing data mining technique. There are various Data
database of third parties. Privacy Preserving in this
Mining Tasks:
paper is used as hide association rule. Association rule
hiding algorithm sanitize database such that certain
sensitive association rule cannot be discovered through
Association rule mining techniques. There are various
Classification
approach this describe in this paper but used the
Heuristic approach in Data Distortion Technique. The
Clustering
proposed algorithm is the extension of MDSRRC
algorithm, which hides multiple R.H.S items. In
Association Rule Mining
Proposed work MDSRRC algorithm works on the
distributed database. We will show experimental results
Sequential Pattern Mining
in comparisons with MDSRRC algorithm in single
database and MDSRRC algorithm in distributed
Regression
database.

II. PRIVACY PRESERVING DATA MINING

Keywords: Data Mining Privacy preserving, association

rule hiding, sensitivity

I. INTRODUCTION
Today there is large amount of data proceed
in every day from different sources. that large amount
of data stored in different database. this data store in
storage devices in from of row data. Data mining is
the process of discovering interesting pattern and
knowledge from large amount of data. Following
Example where data mining techniques are used are
Direct mail marketing, bioinformatics, credit card
fraud detection , text analysis and market basket
analysis. Extracting knowledge from row data, There
Data mining is the process of gathering
information about the user specific data, also called
knowledge discovery, on internet. The problem with
data mining output is that it also discloses some
information, which is considered to be private and
personal. Effortless access to such personal data
causes a peril to individual privacy[9]. Recent
research in the area of privacy preserving data mining
has considerate effort to determine a trade-off
between privacy and the need for knowledge
discovery, which is necessary in order to improve
decision-making processes and other human
activities. PPDM cope with the problem of learning

978-1-4799-6818-3/15/$31.00 © 2015 IEEE

 IEEE Sponsored 2nd International Conference on Innovations in Information Embedded and Communication Systems ICIIECS'15

accurate models over aggregate data, while protecting Confidence of rule A-B is the transaction database
privacy at the level of individual records. that contain A also contain B. The confidence for rule
(A-B) can be calculated using below formula in (2).
The main purpose of privacy preserving data
IAUBI
mmmg is to design competent frameworks and confidence(A-B) = -- (2)
A
algorithms that can extract relevant knowledge from
a large amount of data without revealing of any IV. ASSOCIATION RULE HIDING
sensItIve information[9]. It protects sensitive
information by providing sanitized database of Association rule hiding is one technique to
original database on the internet or a process is used PPDM(Privacy Preserving Data Mining). Association
in such a way that private data and private knowledge rule hiding methodology aim is to sanitized the
remain private even after the mining process. It is original data. so it may be applied to following
PPDM due to which the benefits of data mining be condition:
enjoyed, without compromising the privacy of
(1) sanitized database is not reveal any sensitive
concerned individuals.
rules.
III. ASSOCIATION RULE MINING
(2) sanitized database is mining of all non-sensitive
rules.
Association rule mining one of the task of
the data mining. Association rule mining is important
(3) sanitized database is not add any new rules, not
field to under privacy preserving data mining. R.
present in database D.
Agrawal was first proposed the basic concept of the
Association rule mining. Association rule is basically Association rule hiding is the depend on
using the concept of IF-THEN relationship among support and confidence of the rule, There is two way
the different data. following example of shows the to hide any rule (i) Decrease support up to certain
concept of the Association rule. "If the customer buy threshold. (ii) Decrease confidence up to certain
a laptop ,then he/she is 85% likely to also purchase threshold.
anti-virus ". Analysis of the above example that
laptop is somewhat related to anti-virus because
every time customer buy a computer then he/she buy
anti-virus. Association rule is used for market basket OrigIDal
analysis. Let 1={II,I2,...,In} be a set of item. Let D be database
a database of transactions where each transaction T is
a set of item such that T !:;; I. For every transaction is
associated to an identifier, called TID. A transaction Frequent Associatio,n
T is contain A if and only if A!:;; T. An association item mining RDlehiding
rule is applied of the form A-B. where AQ, BQ. algorithms
and AnB = <1>. every association rule must be satisfy
two contain support and confidence. Generation of Associ­
Association ation
Support of rule A-B is the transaction database that Association rules
ruies rules
contain support count of AUB. support for rule
(A-B) can be calculated using below formula in (1).

IAUBI Figure] - Basic Architecture of rule Hiding

support(A-B) =-­ (1)
D
In figure shows the framework of
where D is the total number of transaction in association rule hiding. In that modification of the
transaction database. database may causes some side effect. following side
effect may occurs in the database.
 IEEE Sponsored 2nd International Conference on Innovations in Information Embedded and Communication Systems ICIIECS'15
Lost rule: there are non sensItIve rule in original
database it can be mined by mining algorithms but it
can not mining after using hiding algorithm from
modify database
False rule: The sensitive rule can not hidden by
hidden algorithm and can be mined by using mining
algorithm for modify database.
Ghost rule: this rule are not present In original
database but generate using the hiding algorithm.
Association rule hiding technique can be classified
into heuristic based approaches, reconstruction based
approaches, border based approaches, exact
approaches, reconstruction based approaches and
cryptography based approaches.
(1) Heuristic based approaches:
Heuristic based approach aIm for taking
locally best decisions with respect to hiding the
sensitive knowledge which, however, which are not
necessary globally best. In heuristic result, heuristic
fail to provides guarantees to identified hiding
solution. This approach is fast, efficient and scalable
algorithm to select the sanitize data set in which set
of transaction from database to hide the association
rule.[l]
Two types of heuristic approach is used: (i) Data
distortion. (ii) Data Blocking.
(i) Data distortion: This type is done by alteration of old
attribute value to new value. It changes 1 's to D's or vice
versa in selected transactions to increase or decrease
support or confidence of sensitive rule[ I,2] . In heuristic
approach give optimum solution because of some side
effect to non sensitive rule.
two basic technique for data is reduced the confidence of
rule and reduce the support of the rule. example
explain........ the problem is finding an optimum solution
using NP-hard. Proposes algorithm using the heuristic
based data distortion technique.
Advantages: It is a more efficient, fast and scalable.
Disadvantages: This approach is difficult to Handel the
changes in Database.
(ii) Data Blocking: This technique is using the maximum
confidence or not reduce the sensitive rule. In database
there are D's and I's must be hidden during blocking,
because D's or 1 's replace with " ? ". In some applications
where publishing wrong data is not acceptable, then
unknown values may be inserted to blur the rules. so, that
support of certain items goes down to certain level and rule
mining algorithm nit able to mine the sensitive rules[3].
Advantages: It Maintain database , instead of inserting
false value to block the original value.
Disadvantages: Difficult to reproduce the original
database. and it is the various side effect like lost rule,
ghost rule, false rule etc.
(2) Border Based Approaches:
Border based approach is hide the sensitive
association rules by modifying the border in the
lattice of the frequent and the infrequent item sets in
original database[3]. this approach is make the border
between the frequent and infrequent items. that way
this border is divided the frequent and in frequent
item sets. The first frequent item set hiding
methodology that is based on the notion of the
border[l]. It maintains the quality of database by
greedily selecting the modifications with minimal
side effect.
Advantages: It maintain the database quality by selecting
the modification with the minimal side effect.
Disadvantages: The border is not easily identify. then it is
difficult to understand based on the heuristic approaches.
(3) Exact Approaches:
Another name of the exact approach is the
non-heuristic algorithm which is formulated to
constrain satisfaction problem(CSP) and solve by
using the binary integer programming(BIP). It
Provide the optimal solution to all constrain. In [4] is
first used the exact approach for hide the rules. and it
provides an optimal solution of rule hiding problem.
In [7] to hide sensitive rules by formulating
constraint satisfaction problem without any side
effects with the concepts of positive and negative
border sets. By using adopting divide and conquer
technique on constraints.
Advantages: it gives guarantees to provides the optimal
solution without any side effect.
 IEEE Sponsored 2nd International Conference on Innovations in Information Embedded and Communication Systems ICIIECS'15
Disadvantages: the approach is require
complexity due to the binary integer programming.
(4) Reconstruction Based Approaches:
Reconstruction based approaches is generate
privacy for database by using sensitive characteristic
from the original database. it produce lesser side
effect in database. In [5] it define the FP tree based
algorithm which reconstruct the original database and
efficient generate number of secure database.
Advantages: Reconstruction based approaches is create the
privacy of database and lesser side effect than heuristic
based approaches.
Disadvantages: The problem is number of transaction is
restricted in new database.
(5) Cryptographic Based Approaches:
Cryptographic based approaches used in
multi-party computation .In which data in distributed
from different location. The owner of the database is
want to share their data, and at the same time they
also want the privacy at their end. cryptographic
Based approaches can be classified two way that is
Horizontal partition distributed data and vertical
partition distributed data.
In horizontal partition distributed data is
provide the different rows are placed in different
tables that are distributed in different locations. In
vertical partition data some column keep in one table
and remaining column in another table. In [6] is
define the basic rule of the association rule over
horizontal partition distributed database. In [8] is
efficiently mine association rules over vertically
partitioned data. they introduce a partial topology to
lower communication cost as much as possible.
Empirical results.
Advantages: It provide the security in
computation over the partition database.
Disadvantages: It does not provide the security of the
output of a computation. Communication and Computation
cost should be low.
V. RELATED WORK
the high There are following paper refer as the
literature survey. This all paper included the
advantages and limitation.
Dharmendra Thakur, Prof. Hitesh Gupta
[9], they present various association rule hiding
algorithm. And they work on the heuristic based
algorithm because this approach is efficient, fast
algorithm to hide the sensitive knowledge. This paper
are using the support based algorithm and confidence
based algorithm.
Pallavi Dubey[ll], she is solve the huge
storage and its management. she is processing to
increase throughput of output data over distributed
database. she conclude that to improve response time,
number of processor in distributed environment.
R.Natarajan, Dr.R.Sugumar, M.Mahendran,
K.Anbazhagan[15], They using the privacy
preserving data mining to provide the confidentiality
and improve the performance of at the time when
database stores and retrieves huge amount of data.
they present ISL(Increase support of L.H.S) and
DSR(Decrease support of R.H.S).In which DSR
algorithm to hide useful association rule from
transactions data with binary attributes. and ISL
algorithm confidence of a rule is decreased by
increasing the support value of LHS of the rule. they
conclude that using this two algorithm only the 19%
efficiency increase.
Chirag N. Modi, Udai Pratap Rao and
Dhiren R Patel[2], They proposed heuristic algorithm
its name was DSRRC which is hide rule only the
certain level. the heuristic algorithm better used then
the other hiding algorithm. DSRRC algorithm is hide
rule that rule contain single item on R.H.S of the rule.
Komal shah, Amit Thakkar, Amit
Ganatra[I], they proposed two algorithm ADSRRC
multi-party
and RRLR to hiding sensitive association rule. this
two algorithm are overcome the limitation of DSRRC
algorithm. ADSRRC overcomes limitation of
multiple sorting in database as well as it selects
transaction to be modified based on different criteria
than DSRRC algorithm. Algorithm RRLR overcomes
limitation of hiding rules having multiple R.H.S.
items. they also reduce the side effect and
complexity.
 IEEE Sponsored 2nd International Conference on Innovations in Information Embedded and Communication Systems ICIIECS'15

Vikram Garg, Anju Singh, Divakar
Singh[12], Data mining is extract the hidden
predictive information from the data warehouse
without revealing their sensitive information. they
show all PPDM(privacy-preserving data mining)
technique and also show the recently research area to
deal with the association rule hiding. Show the
comparative analysis of the all association rule hiding
approaches.
J. Arokia Renjit, Dr .K.L
.Shunmuganathan[16], Data mining algorithm are
used to centralized algorithm. in that centralize data
mining algorithm to discover useful patterns in
distributed databases. in that the distributed database
is not always feasible because merging data sets from
different sites incurs huge network communication
costs. they are using the FDM algorithm fro
geographically Distributed Data sets, and use
effective pruning technique to minimized the
Step: 2 - Algorithm Finds IS = {isO, isl, ... isn}
k<=n, By arranging those in decreasing order of their
counts.
Step: 3- Calculate Sensitivity of each Transaction.
Step: 4 - The Transaction which support isO are
shorted in decreasing order of their sensitivity.
Step: 5 - In Shorted Transaction which is having
higher Sensitivity, delete item isO from the
transaction.
Step: 6 - Update support and confidence when delete
item of transaction. Also update the database.
This following figure show the proposed
flow of the work.
I �
I I I
Database(D) Apply

nl
Combme me
message for the support. conclude that mining
association rules in distributed databases which is
res.ult Association

rule II
reduced communication , efficiency, and scalability
1 Distributed T 1
Shorting item
then communication of sequential algorithm in Divide Database
According to
-i
Distributed database. i.e. ( 01 ,02,03) sensitivity

Up ate

Nikunj H. Domadiya, Udai Pratap Rao[lO],

this paper they proposed the MDSRRC algorithm to
1
Apply Apriory Algorithm &

Generate Association rule

Se sitivity

--11
1
Rul. Hiding
of all divided DB

1
hide the sensitive association rule with multiple items
in consequent (R.H.S) and antecedent (L.H.S). They � Update

Select any one fro m DB &

I
overcome the problem of DSRRC algorithm. my Combine with the
Upd ,te
Transaction DB
proposed work is related to this paper. they also show remaining

the comparative analysis of the DSRRC algorithm 1

and MDSRRC algorithm in which they shoe the The result will share with

aliOS using
MDSRRC algorithm is increase the efficiency of the Syn c hronization (using

message passing)
DSRRC algorithm.

Figure2- Proposed Flow

VI. PROPOSED WORK
2. Description Of Algorithm(Proposed)
1. Description OfAlgorithm (Existing)
There are following step of the MDSRRC Algorithm
There are following steps of the MDSRRC in Distributed Database:
Algorithm:
Step: 1 - Take the Large Database. because using the
Step: 1 - Count the Occurrence of each item in R.H.S distributed database.
of Sensitivity.
Step: 2 - Concept of the distributed database are they
are divided(i.e Dl,D2,D3)
 IEEE Sponsored 2nd International Conference on Innovations in Information Embedded and Communication Systems ICIIECS'15

Step: 3 - Apply the Association rule and Generate the VIII. FUTURE WORK
Association rule of all dividing Database.
Implementation of the proposed work is still under
Step: 4 - suppose select the Database Dl then progress and is left as future work.
combine the remaining two database D2,D3.

Step: 5 - Database Dl result is there then share with REFERENCES

all database D2,D3 using synchronization with
message passing technique. [ 1 ] Komal shah, Amit Thakkar, Amit Ganatra," Association
Rule Hiding by Heuristic Approach to Reduce Side Effects

Step: 6 - Combine All the result that is the generated & Hide Multiple R.H.S. Items " International Journal of

Association rule. Computer Applications (0975 - 8887) Volume

45- No.1 , May 20 1 2.

Step: 7 - Count the Occurrence of each item in R.H.S

[2] Chirag N. Modi, Udai Pratap Rao, Dhiren R. Patel,
of Sensitivity.
" Maintaining Privacy and Data Quality in Privacy
Preserving AssociationRule Mining" Second International
Step: 8 - Algorithm Finds IS = {isO, isl, ... isn}
conference on Computing, Communication and
k<=n, By arranging those in decreasing order of their
Networking Technologies,20 I O.
counts.
[3] Tapan Sirole and Jaytrilok Choudhary, "A Survey of
Step: 9- Calculate Sensitivity of each Transaction. Various Methodologies for Hiding Sensitive Association
Rules" International Journal of Computer Applications
Step: 10 - The Transaction which support isO are (0975 - 8887) Volume 96- No.18, June 20 1 4.
shorted in decreasing order of their sensitivity.
[4] A. Gkoulalas-Divanis, and V. S. Verykios, " Exact
Step: 11 - In Shorted Transaction which is having knowledge hiding through database extension" IEEE Trans
higher Sensitivity, delete item isO from the Knowledge Data Eng 2009, pp. 699-7 1 3.

transaction.
[5] Y. Guo, " Reconstruction-Based Association Rule

Step: 12 - Update support and confidence when Hiding," In Proc. of SIG MOD2007 Ph.D. Workshop on
Innovative Database Research 2007(ID A R2007), June
delete item of transaction. Also update the database.
2007.

VII. CONCLUSION
[6] Tamir Tassa; "Secure Mining of Association Rules in
Horizontally Distributed Databases" ieee transactions on
MDSRRC algorithm is hide the sensitive association
knowledge and data engineering, vol. 26, no. 4, april 20 1 4.
rule with few modification on database. Which
maintain the data quality and reduced the side effect [7] N V Muthu Lakshmi and Dr. K Sandhya Rani, " An
of database. I will try to extend the MDSRRC improved algorithm for hiding sensitive association rules

algorithm in distributed database. I will try to make using exact approach" I RACST - International Journal of

association rule more secure. And I will also try to Computer Science and Information Technology & Security
(IJCSITS), I S S N: 2249-9555 Vol. 2, No. 1 , 20 1 2.
the MDSRRC algorithm, increase the efficiency and
reduced the side effect by minimizing the [8] Due H. Tran, Wee Keong Ng and Wei Zha,
modification on database. and will used distributed "CRYPP AR: An Efficient Framework for Privacy
database to hide the association rule. Comparison of Preserving Association Rule Mining over Vertically
DSRRC,MDSRRC in single database, and my Partitioned Data" 978- 1 --4244--4547-9/09/$26.00 c 2009

proposed work MDSRRC in distributed database IEEE.

Will also be included.

[9] Dharmendra Thakur, prof. Hitesh Gupta, " An
Exemplary Study of Privacy Preserving Association Rule
Mining Techniques" International Journal of Computer
Science and software engineering Volume 3, Issue 1 1 ,
November 20 1 3.
 IEEE Sponsored 2nd International Conference on Innovations in Information Embedded and Communication Systems ICIIECS'15

[ 1 0] Nikunj H. Domadiya, Udai Pratap Rao, " Hiding
Sensitive Association Rules to Maintain Privacy and Data
Quality in Database " 3rd IEEE InternationalAdvance
Computing Conference IEEE 20 1 3.
[II] Pallavi Dubey, " Association Rule Mining on
Distributed Data ", International Journal of Scientific &
Engineering Research, Volume 3, Issue I, January-20 1 2.
[ 1 6] 1. Arokia Renjit, Dr.K.L.Shunmuganathan, " Mining
the data from distributed database using an improved
mining algorithm ", International Journal of Computer
Science and Information Security, Vol. 7, No. 3, March
20 1 0.

[ 1 2] Vikram Garg, Anju Singh, Divakar Singh, "

A Survey of Association Rule Hiding Algorithms", Fourth
International Conference on Communication Systems and
Network Technologies,IEEE 20 1 4.

[ 1 3] J. Arokia Renjit, Dr.K.L.Shunmuganathan, " Mining

the data from distributed database using an improved
mining algorithm ", International Journal of Computer
Science and Information Security, Vol. 7, No. 3, March
20 1 0.

[ 1 4] Nikunj H. Domadiya , Udai Pratap Rao , " Hiding

Sensitive Association Rules to Maintain Privacy and Data
Quality in Database " 3rd IEEE InternationalAdvance
Computing Conference, IEEE 20 1 3.

[ 1 5] R.Natarajan, Dr.R.Sugumar, M.Mahendran,

K.Anbazhagan, " Design and Implement an Association
Rule hiding Algorithm for Privacy Preserving Data Mining
", international Journal of Advanced Research in Computer
and Communication Engineering Vol. I, Issue 7,
September 20 1 2.