Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Implementation of IoT System for Securing

Telecommunications Infrastructure Based on


LoRaWAN Operator's Network
Marko Lj. Jovanovi Mladen Koprivica Nataša Neškovi
Department of Telecommunications, Department of Telecommunications, Department of Telecommunications,
School of Electrical Engineering, School of Electrical Engineering, School of Electrical Engineering,
University of Belgrade University of Belgrade University of Belgrade
Telekom Srbija (2nd Affiliation) Belgrade, Serbia Belgrade, Serbia
Belgrade, Serbia kopra@etf.rs natasha@etf.rs
marexgatuzo@yahoo.com

Abstract—This paper considers a practical IoT (Internet of should be implemented in the upcoming period over the
Things) solution implemented in a LoRaWAN (Long Range operator’s network.
Wide Area Network) network of the Serbian
telecommunications operator Telekom Srbija. The IoT solution The paper is organised as follows. The second section
was developed to secure the operator’s infrastructure located brings a discussion about LoRaWAN technology and
in manholes and prevent stealing. The paper presents methods benefits of using it instead other LPWA (Low Power Wide
and techniques, which telecommunications operator has used Area) technologies. The next section is referred to the
to incorporate LoRaWAN technology into the existing implementation per se. It is divided into three parts, where
IP/MPLS (Internet Protocol/Multi-Protocol Label Switching) are described topology, radio planning and application
network. The results of this technology implementation are services. The following section presents results of LoRa
shown and significance of it is pointed out. implementation and its positive outcome. Last section gives
conclusion about LoRaWAN technology and its
Keywords—base stations, Internet of Things, long range, low implementation.
power, LoRaWAN, LPWAN, M2M, radio planning, sensor,
server.
II. THE NETWORK TECHNOLOGY
I. INTRODUCTION With the IoT development, the traditional Internet has
been revolutionized: objects got an opportunity to connect
The IoT (Internet of Things) refers to the billions of and communicate through the Internet. Hence, services
physical devices around the world which are connected to the outspreaded to machine-to-machine and not only human-
Internet collecting and sharing data. This term is mainly used centric services [1]. That implies the possibility of (almost)
for gadgets that would not usually be expected to have an every gadget is going to be connected to the Internet.
Internet connection. Therefore, desktop and notebook Comparing the Internet with the IoT [2], a younger brother
computers, or smartphones, would not be considered as the has got everything with smaller amount: smaller bandwidth,
IoT devices. On the other hand, smartwatch or some other less memory, few processing power. That is why the IoT
wearables could be counted as an IoT device. The leader in should avoid using classical Wi-Fi and cellular networks and
network engineering, the company Cisco Systems, predicts turn to something which consumes less energy. Also, many
that there will be 500 billion devices connected to the wireless technologies are short-ranged like Zigbee. Because
Internet by 2030. This forecast shows the importance of the of all that, LPWA technologies became very popular. It is
IoT development and a future progress in electronics important to distinct two main groups of these technologies.
(because of the IoT devices development), The first one consists of standards that use a licensed
telecommunications (because of need for evolving new types spectrum, therefore more expensive than rest.
of network and protocols suitable for low energy Representatives of these standards are: NB-IoT (Narrow
consumption) and cloud computing (because of application Band – IoT), LTE-M (Long-Term Evolution – Machine-to-
services and information about devices that should be always machine) and EC-GSM (Extended Coverage – Global
available for the user). System for Mobile communications). On the other hand,
The Serbian telecommunications operator Telekom Srbija group of LPWA standards which utilize unlicensed spectrum
has faced with a challenge regarding the stealing its consist of LoRa (Long Range) and Sigfox technologies.
infrastructure. Because of an extensive and widespread Not only telecommunications operator does not have to
telecommunications infrastructure, this became quite big pay for spectrum utilizing, but the costs of equipment and
issue requiring to be solved in the most efficient way. For LPWA network realization are significantly less. For
this reason, the operator decided to use LoRaWAN (Long instance, price for one LoRa gateway is between $100 and
Range Wide Area Network) to secure its infrastructure. The $1000, while for one NB-IoT base station is needed to get
paper considers a practical IoT solution implemented in a out of pocket $15000 [3]. But now, there is a question: which
LoRaWAN operator's network. It is explained why the one to use LoRa or Sigfox? Talking about finances, Sigfox
telecommunications operator has chosen this technology and does not allow installing your own base station and are
not any other. It is important to point out that this solution limited to where the stations are set up. Therefore, operator
also served as a pilot project for a future IoT services which has no possibility to use its existing network and yet has to
School of Electrical Engineering, University of Belgrade. pay for building a new one because Sigfox technology is not

 
      1
an open protocol. Also, license must be paid even if it used on a sensor side, sensor will retransmit message 8 times in a
only peer-to-peer mode; Sigfox subscription is a must. LoRa row. If sensor fails to receive an ACK even after these
allows communication in this mode free of charge and retransmissions, it will wait 4 hours to repeat the action.
allows creating one’s own network with user-defined
network coverage. LoRa offers slightly better security Other type of downlink traffic considers firmware
providing AES (Advanced Encryption Standard) encryption. configuration. When the window is open, whether it is
On the other hand, Sigfox scores longer ranges. Huge because of heartbeat or alarm-triggered reason, server has a
downside of Sigfox technology is very small size of payload possibility to give an instruction to the sensor, i.e. sensor’s
which grants only 12 bytes (with a limited number of parameters could be changed then. This is very useful
messages per day), while that number is 243 bytes for LoRa. option, because it provides to set up parameters differently
Also, Sigfox has no option for downlink traffic, unlike to in case they are configured as too sensitive or too robust.
LoRa [4]. No doubt, that was decisive for Telekom Srbija to Not only it’s possible to configure temperature, luminosity
choose LoRa technology, considering that in this particular or heartbeat time parameters, but there is an opportunity to
project downstream traffic is used for sending ACKs set up sensor to function optimally by spreading
(acknowledgment messages) from server to sensor. This way optimization with ADR (Adaptive Data Rate). Enabling
sensor knows that it should stop to retransmit the packet ADR, it will automatically select the best SF (Spreading
because server has already got it. Another crucial feature that Factor). SFs are in range of 7 to 12. SF7 is considered the
triggered this operator to select LoRa is scalability and its best one, because it needs less time than other ones for
fantastic link budget results, which are better than all other transmitting the same message, which implies less power
LPWA technologies [5]. consumption and longer battery life. For instance, SF7 needs
sec. to transmit a message that SF12 transmits for 1.5
sec. Usually, the nearer sensor to radio base station is,
the better SF is going to use. In case the sensor is located
under a double manhole cover, thick concrete, or special
positioning of the buildings, SF7 will not be used, even if
it is near the radio base station. Battery life of sensor is
estimated to last 10 years at least.

III. THE IMPLEMENTATION

A. Topology
This section describes topology of a LoRaWAN solution
Fig. 1. LoRaWAN stack.
used in the network of Telekom Srbija (Fig. 2). Because of its
LoRaWAN represents wireless telecommunications purpose, which is securing telecommunications infrastructure
network designed to provide long-distance communication inside the manholes, sensors should be installed right under
between end devices with low bitrates [6]. It comes up with the manhole cover, so they could detect a manhole cover
interoperability between the IoT devices without need for opening or an unauthorized person entering, or both. Also,
complex set ups including secure bidirectional this device should have such a construction that the
communication, mobility and service localization. Typical possibility of its vandalization would be at the minimum
architecture contains star-of-stars topology. Usually, range level. On the other side, sensor should be as thin as possible,
reaches 15 km in rural environments or 5 km in urban ones. so it would not affect work on the field. Omniimpex, a
LoRa defines physical layer which enables long-distance vendor of these sensors, decided to use rounded aluminum
communication links, and is based on CSS (Chirp Spread profile with rounded head and flexible tip, or to be more
Spectrum) modulation that has feature of low-power precise, stainless steel tube in ball bearing because rubber tip
consumption just like FSK (Frequency Shift Keying) is easy to be destroyed. The sensors send signals to LoRa
modulation, but with significantly longer range. LoRaWAN gateway(s). This communication could be scheduled or event-
defines MAC (Medium Access Control) layer and based, as it is described in the previous section.
communication protocol (Fig. 1). LoRa gateway is nothing else but a nexus between end
In a project related to this paper all devices are class A, device (that is, sensor) and server. Actually, gateways are
which means there are no latency restrictions and this is the concentrators with a function of transparent bridges
most energy-efficient class. Class A devices use pure responsible for data transport from end devices (aka. nodes)
ALOHA access for the uplink. After sending a frame, class to network server. Every end device broadcasts messages
A device listens for a response during two downlink receive and gateways in its range accept them. One message with the
windows, while downlink transmission is only allowed after same content might be received from greater number of
a successful uplink transmission. Practically, sensor responds gateways. Network server drops packets with the same
with predefined scheduled period known as heartbeat time. content solving the multiple messaging problem. However, it
This response (which occurs every 24 hours) helps to is not quite a challenge, considering that the packets arrive at
indicate normal operation and to synchronize the system. more than one gateway in less than 10% cases. Unlike to end
During this time a channel is opened for communication, devices, communication between gateways is possible.
therefore, it is possible to send message from server to Because of need for very long range and large coverage
sensor. As it is earlier mentioned, ACKs are one particular area, related telecommunications operator has chosen to use
type of downlink traffic in this case. If ACK does not arrive Ufispace macro gateway. This gateway model has got two

2
LoRa antennas, GPS (Global Positioning System) antenna The first prediction, made by Actility, has been done with
and LTE antenna. LTE antenna serves to enable a backup software tool LoRa Air Interface Dimensioning. This
solution for gateway connectivity. If it is not possible to prediction disregards the exact terrain configuration. Instead,
provide the Ethernet connection, gateway will communicate it uses antenna’s location height and an average attenuation
through mobile network. Additionally, this gateway model for every type of environment to predict the appropriate radio
owns PoE (Power over Ethernet) adapter, which is very planning. By default, there are four environment types, but
convenient solution to supply gateway with power. For the only three are used in this case study (Fig. 3): urban (red
purposes of getting coverage as long as possible, antennas zone with tall buildings, usually located in downtown),
should be installed as high as possible. Furthermore, suburban (blue zone with smaller buildings or blocks of them
antennas must be obstacles free, especially the metal ones. separated with wide streets, industrial areas), rural (green
wide area with scattered houses and lots of field and forest).
Signal travels further from gateway(s) to digital units,
which do digitalization of that signal. In this form, signal Talking about the outdoor coverage, there is no problem
arrives at one of two devices, so called TCU (Transmission for whole this area. According to this calculus, four gateways
Control Unit) or SIU (Site Integration Unit). They represent are enough to make good signal strength through the selected
devices for “all-IP-zation” of radio base stations. Mentioned zones (Fig. 4). Reason for that lays in installed antennas on
network devices are Ericsson proprietary and act as an very high locations: sites 1, 6 and 8. Challenge here would be
advanced switch or some kind of router, which makes to provide such a good signal strength under the ground, i.e.
L3VPN (Layer 3 Virtual Private Network) tunnel to its under the manhole covers with as less as possible gateways.
superiors in IP/MPLS network, i.e. PE (Provider Edge) It is obvious that mentioned four gateways are not enough.
router [7]. Considering the security aspects, this is way more Adding nine more gateways and keeping the same
preferable and sophisticated solution than using the public parameters results with not-so-brilliant manhole coverage
Internet, which is the most common method generally. (Fig. 5). Causes of such outcome are manhole covers, which
contribute with large attenuation (roughly ~ 20 dB). Because
of that, there is a need for repetition per message, which
means there should be one more transceiver. Consequently, it
will compensate the loss and enable manhole coverage with
minimal number of gateways (Fig. 6).

Fig. 3. Terrein morphology map with zones for LoRa coverage.

Fig. 2. LoRaWAN architecture deployed for Telekom Srbija network.

Related interface on PE router ought to be provided with


an appropriate configuration, where VLAN (Virtual Local
Area Network) parameters must match VLAN parameters
configured on TCU or SIU device. Additionally, interface
VLAN should be configured on PE router, if we want to
ensure sensors management. From this point, connection
between a network and an application server is provided with
static route that directs packet to its final destination. More
about application layer will be described in subsection C.

B. Radio Planning
Radio planning is the key step in creating a stable
LoRaWAN network or any other wireless network. First of Fig. 4. Outdoor coverage with LoRa signal (sideview).
all, it is important to perceive characteristics of the
environment. In this particular case study, terrain is mostly Ideally, LoRa antenna should be at least 3 meters above
plain landscape, but there is a critical area near the river cellular antenna or 10 meters further in horizontal plane. If
which is quite densely populated with a lot of buildings. isolation cannot be accomplished by spacing, there must be

3
used cavity filter between antenna and gateway. This filter TABLE I. SITE LOCATIONS
represents resonant circuit for frequency filtering. According Site ID Code Priority Comment
to this prediction, 12 gateways are required to cover all
manholes from the zones at the Fig. 3 (TABLE I). 1 BG1 1 Outdoor coverage
2 BG2 1 Outdoor coverage

3 BG3 3 Outdoor redundancy


4 BG4 2 Manhole coverage

5 BG5 2 Manhole coverage


6 BG6 1 Outdoor coverage
7 BG7 2 Manhole coverage

8 BG8 1 Outdoor coverage


9 BG9 3 Outdoor redundancy
Fig. 5. Manhole coverage with one transceiver (sideview). 10 BG10 Not needed Not used, whatever the target
11 BG11 Not needed Not used, whatever the target
12 BG12 Not needed Not used, whatever the target

13 BG13 4 Manhole redundancy


14 BG14 2 Manhole coverage

15 BG15 2 Manhole coverage

The second radio planning calculus has been done by


Telekom Srbija. This time is used more powerful software,
Telko, which considers actual terrain configuration and
occurrences such are reflection or refraction. With this
Fig. 6. Manhole coverage with two transceivers (sideview). prediction model, manhole coverage is determined more
precisely (Fig. 7), where is also calculated the lowest
possible signal level which is -105 dBm (blue color). Green
color represents area with the strongest signal level, i.e. in
range [-90, 0] dBm. The final prediction recommends exactly
30 gateways to secure ideal manhole coverage, which is way
more than the first one prediction model.

C. Application
Application server provides application services to an
end device, or to be more precise, a user of the end device.
This server manages application layer of LoRaWAN stack
and handles payloads received from nodes. Also, it serves as
data storage. Packets are arriving to mentioned server via
static route from relevant router in the IP/MPLS network.
Users can access software platform on the application server.
All they need to do is to type appropriate URL (Uniform
Fig. 7. Telko radio planning. Resource Locator). Subsequently, they will be asked for their

Fig. 8. Viewing sensor logs at the subscriber tab

4
credentials. After logging in, there will be shown a home Similarly, task was accomplished and had positive outcome
page with operator, supplier and vendor tabs. Firstly, it is (TABLE II). Of course that RSSI (Received Signal Strength
essential to add an administrator user at the operator tab. This Indicator) and SNR (Signal to Noise Ratio) parameters
way it is enabled for the user to handle all introduced tabs. scored here better results than during the drive test, because
Supplier tab contains information of base stations and mobility and velocity put obvious impact on it.
manhole sensors. Some of these pieces of information are:
base station ID, address and type of base station, connection
TABLE II. RESULTS OF WALK AND DRIVE TESTS (PART OF THEM)
method, server visibility, CPU (Central Processing Unit) and
RAM (Random Access Memory) usage, uplink/downlink NBr
RSSI SNR RSSI SNR
Gateway ID
traffic. User may view base stations on a map or sort them in (walk) (walk) (drive) (drive)
the list. It is possible to set GPS coordinates manually on the 1 -69 9.25 -111 -3.5 00000524
map or set them more precisely on LRC (LoRa Controller) 2 -81 10.25 -110 -1 00000524
after accessing to the mentioned controller. Vendor tab 3 -73 10.75 -113 -6.5 000005C7
defines subscribers and end users, and orders of theirs. 4 -91 8.75 -93 -4.5 00000524
5 -80 9 -111 -1.5 00000524
Subscribers manage the sensors and sensor logs (Fig. 8). 6 -98 7.75 -108 1.75 000005C7
7 -85 8.25 -109 -0.25 00000524
8 -86 9.75 -110 -4.5 000004C1
IV. TEST RESULTS 9 -97 7.25 -115 -7.75 00000524
10 -111 0.75 -111 -0.25 000004C1
When the implementation is finished, next step is to test 11 -106 4.25 -105 6.5 000004C1
an operability of all services. LoRa network coverage test 12 -100 4.75 -115 1 000004C1
results are shown in this section. Tests were carried out with
two of modes: as drive test and as walk test. A device that
served for the testing purposes was Adeuinis Field Test C. CPU, RF and WAN parameters tests
which imparts maximal range 15 km. Other features are:
This section reviews the results of testing base station
superb battery autonomy, very precise GPS signal, support
parameters. It is very useful, because these results can show
for accelerometer, LoRaWAN v1.0 interoperability [7].
a health and a status of base station, thus, warn an amenable
person when there is a need for some changes.

Fig. 10. CPU usage of one base station.

Fig. 9. Map with drive testing path.

A. Drive tests
Base Station was up and correctly provisioned through
GUI (Graphical User Interface). There was one LoRaWAN Fig. 11. Duty cycles for upstream and downstream of one base station.
demonstrator device connected and correctly provisioned
through GUI (Adeunis Test Field Device), with a
connectivity plan that lock the SF to SF12. This means that
each bit was encoded into 4096 (27) chirps [8], a signal that
constantly made frequency variations. It indicated lower
physical bitrate reaching only 250 b/s and longer time on air
(1318.9 ms). Drive test was performed in the area of PoC
(Proof of Concept) LoRaWAN network [9] (Fig. 9). After
logging in on the application platform as a subscriber, a goal
was to verify that gateways and test device can be located Fig. 12. Sent/Received packets of one base station.
and viewable on the map. LoRaWAN network coverage was
verified (TABLE II) and this test was marked as passed. One of four LoRa gateways (base stations) with priority 1
from section II. B. has been examined to numerous tests.
B. Walk Tests CPU test shows that there was no heavy system processes
Expectation for this kind of test was to prove a visibility considering that an average usage is 5% and the maximum
between test device and gateways through the whole trace. scores only 36% (Fig. 10). Next test deals with uplink and

5
downlink traffic. It shows that there was low downlink traffic days). It looks like thieves wanted to play on a card of non-
(Fig. 11). Last presented test in this text (Fig. 12) clarifies working days. Also, a number of potential threats is lowered
that there were more sent than received packets, which is by one third in November. Very important fact is break-ins
logically because gateway sends to sensors and server, both. diminishing in the following month (Fig. 14). Almost every
day in November has got less generated critical alarms
V. OUTCOME OF THE IMPLEMENTATION AND ITS BENEFITS comparing to October; the exception is a tail of November.
After the implementation and successful tests, it was time There has been a total of 408 generated alarms in October
to get some statistics about real-life results, not only ones and that number is reduced to 274 ones in November, what is
scored in “laboratory environment”. LoRaWAN has been a very decent score. It is possible to view on the next graph
officially put into operation for Telekom Srbija network at (Fig. 15) a trend line (median line marked with red color)
the end of September 2018. Here is going to be presented that falls down, which proves an indication of decreasing in
statistics for October and November, 2018. the infrastructure stealing attempts. During these two
months, three persons are arrested and convicted for stealing.
This is a good indicator that LoRa was a good choice for this
case study.

VI. CONCLUSION
This paper has pointed the importance of the IoT and
M2M technologies stating differences between them and
explaining their purposes. Choice for LoRa of related
telecommunications operator does not mean other LPWA
technologies are not enough good, nor inappropriate. On the
contrary, all of them own features that could be used for the
mater of interest, whether it is bitrate, latency, distance, etc.
For example, Sigfox could have been easily set up instead
LoRa, if the operator had got the will for implementing that
Fig. 13. Graph with generated alarms submitted by categories. kind of network. Anyway, LoRa represented itself in a good
light securing manholes and providing expected solution. A
problem with the infrastructure stealing is finally solved
thanks to this technology.

ACKNOWLEDGMENT
The authors would like to thank to Telekom Srbija and all
its partners for allowing this paper to be published.

REFERENCES

Fig. 14. Graph with generated critical alarms by months. [1] O. Khutsoane, B. Isong, A. M. Abu-Mahfouz, “IoT Devices and
Applications based on LoRa/LoRaWAN”, IECON 2017 – 43rd
Annual Conference of the IEEE Industrial Electronics Society, South
Africa, 2017.
[2] A. Augustin, J. Yi, T. Clausen, W. M. Townsley, “A Study of LoRa:
Long Range & Low Power Networks for the Internet of Things”,
Sensors, Vol. 16, Issue 9, 2016. https://doi.org/10.3390/s16091466.
[3] R. S. Sinha, Y. Wei, S. Hwang, “A survey on LPWA technology:
LoRa and NB-IoT”, ICT Express, Vol. 3, Issue 1, pp. 14-21, 2017.
[4] K. Mekki, E. Bajic, F. Chaxel, F. Meyer “A comparative study of
LPWAN technologies for large-scale IoT deployment”, ICT Express,
Article in Press, 2018. DOI: 10.1016/j.icte.2017.12.005.
[5] D. Sjostrom, “Unlicensed and licensed low-power wide area
networks: Exploring the candidates for massive IoT”, Master’s
Thesis, KTH Royal Institute of Technology, School of Information
Fig. 15. Graph with trend of critical alarms. and Communication Technology (ICT), Stockholm, Sweden,
September 2017.
As it can be seen (Fig. 13), many attempts are noted to [6] M. A. Alereqi, F. N. Al-Wesabi, A. T. Zahary, M. N. Ali, “A Survey
of Internet of Things Services Provision by Telecom Operators”, EAI
break in manhole and steal the infrastructure. Basically, Endorsed Transactions on Internet of Things, Vol. 4, Issue 13, 2018.
every October day has got an average of 10 unauthorized [7] Telekom Srbija, ThingPark Wireless, “Technical Solution – LPWAN
break-ins, while that number decreased in November by RFP”, September 2017, unpublished.
32%. There was almost the same amount of false positives in [8] N. Blenn, F. Kuipers, “LoRaWAN in the Wild: Measurements from
these two months. It is interesting that mentioned illegal The Things Network”, 2017, https://arxiv.org/pdf/1706.03086.pdf
actions were more often during the working days (average [9] Telekom Srbija, Actility, Teri Engineering, Bitgear, "Telekom Serbia
11.7 per day) than during the weekends (average 5.13 per IoT Project - Interoperability PoC Test Results V1.33", December
day) in October, while these difference was drastically 2017, unpublished.
reduced in November (daily 7.45-5 in favor of working

You might also like