Self-recovery Service Securing Edge Server in IoT Network against
Ransomware Attack
In-San Lei, Su-Kit Tang, Ion-Kun Chao and Rita Tse
School of Applied Sciences, Macao Polytechnic Institute, Macao, China
Keywords: Edge Computing, Ransomware, Recovery.
Abstract: Edge server takes an important role in IoT networks that distributes the computing power in the network and
serves as a temporary storage for IoT devices. If there is any ransomware attack to the edge server, its network
segment will be paralyzed, raising the data integrity and accuracy issues in the IoT system. In this paper, we
propose a self-recovery method, called Self-Recovery Service (SRS), which can detect ransomware signature
and recover victim files automatically. No interruption to the operation of edge server would be caused by
ransomware. SRS is evaluated in the simulation test and the result shows that SRS takes insignificant system
resources for its operation that does not degrade the performance of the edge server.
1 INTRODUCTION
The rapid development of smart cities has driven the
deployment of Internet of Things (IoT) system in
solving specific problems in different fields, resulting
in a dramatical growth in the number of deployed IoT
devices. IoT devices are heterogeneous and of low
computing capability. An IoT device collects and
generates data continuously and data are forwarded to
backend servers for processing. If an IoT system is
connecting to a large number of IoT devices, a
bottleneck problem may be created at the backend.
Adding edge servers into the system can ease the
bottleneck problem and complicated computing tasks
can be simplified at the early stage.
In general architecture, edge servers are
interconnected and they serve as a gateway for their
IoT local subnet. (Fernández et al, 2018) (Lopez et al,
2015) (Shi et al, 2016) An edge server comes with
computing capabilities to pre-process dynamic data
generated from IoT devices. If intensive computing
tasks on the data are involved, the pre-processing not
only reduce the workload of backend servers, but it
also simplifies the logic of the data processing. This
distributes the power across a network and improves
the efficiency of the entire IoT system. For this
reason, the market size of edge servers grows steadily
and is expected to climb up to 1031 million by 2025
while the number of IoT devices will be projected to
reach 75 billion by 2025 (Statista, 2017) (Statista,
2020). In this context, edge server is likely to raise
interest from the academia and the industry. It is also
very attractive to cyber extortionists, as attacking
edge servers rather than IoT devices makes more
significant impact to IoT systems. It would raise
many challenges related to security and privacy
concerns (Shi et al, 2016).
A number of studies (Xiao et al, 2019) (Alrowaily
& Lu, 2018) (Hafeez et al, 2016) (Caprolu et al, 2019)
(Zhang et al, 2018) have been conducted regarding to
the security and privacy of edge server. Edge servers
in IoT networks are normally accessible in the public.
Any design flaws, misconfigurations and
implementation bugs may put them at risk, suffering
a number of cyber-attacks including ransomware
attack. Recently, ransomware has rapidly become one
of the severe network threats for enterprises and
individual users, causing billions of dollars in loss
globally. The maturity of ransomware has even
reached a new height that can attack millions of
computers at a time. Files encrypted by ransomware
are often unable to be decrypted, unless the
decrypting key is obtained. Reasonable defense
measures for edge servers to reduce the chance of
being attacked is definitely needed. Suppose that a
ransomware is injected into an edge server under the
malware injection attack, the server would stop
accepting data from IoT devices. No processing is
done for the IoT system. This results in data lost that
prevents the system from making critical decision and
thus downgrades the performance of the system.
Lei, I., Tang, S., Chao, I. and Tse, R.
Self-recovery Service Securing Edge Server in IoT Network against Ransomware Attack.
DOI: 10.5220/0009470303990404
In Proceedings of the 5th International Conference on Internet of Things, Big Data and Security (IoTBDS 2020), pages 399-404
ISBN: 978-989-758-426-8
Copyright
c
2020 by SCITEPRESS Science and Technology Publications, Lda. All rights reserved
399
In this paper, we aim at easing the damage caused
by ransomware to edge server by proposing a self-
recovery method, called Self-Recovery Service
(SRS), for edge server. SRS can detect ransomware
signature and recover victim files automatically. Its
concept is to monitor important files by a system
service. If ransomware is detected, SRS recovers
infected files by restoring the corresponding backup
of raw data. No interruption would be caused to the
operation of edge server. The service only takes
insignificant system resources for its operation that
does not degrade the performance of the server.
The remaining of this paper is organized as
follows. In Section 2, we review the ransomware
attack. In Section 3, we present the design of SRS.
Section 4 will highlight the verification of SRS.
Finally, we conclude this paper in Section 5.
2 RANSOMWARE ATTACK
Ransomware is a kind of malicious trojan horse
program which is secretly injected into victim's
devices (computers, smartphones, servers, etc.) and
interferes with their operation by encrypting some
important files, such as user data files or system files.
To rescue the files and even save the devices, the
decrypting key is needed which can be obtained by
paying for the ransom (O'Gorman & McDonald,
2012). In 1989, the first ransomware, called PC
Cyborg, had successfully forced the user to pay a
ransom of $189 by encrypting its hard disk (Gazet,
2010).
There are five types of ransomware reported
(Johansen, 2018), which are Crypto malware,
Lockers, Scareware, Doxware and RaaS. Crypto
malware is a well-known ransomware that can spread
over thousands of computers and make damages
worldwide. One of the noticeable examples is
WannaCry; Lockers is another type of ransomware
that aims to attack operating system. It locks down a
victim’s computer. No files or applications can be
accessed; Similarly, Scareware would lock down a
victim’s computer and pop up annoying messages to
ask for ransom; Doxware is another type of
ransomware that may reveal a victim’s sensitive
information. It threatens the victim by posting the
information online, if ransom is not paid; The last
one, called “Ransomware as a Service (RaaS)”,
contributed greatly to the growth of ransomware
attack because it enables anyone to be a cyber
attacker. RaaS is deployed as a portal that enables
legitimate venders to unintentionally setup malicious
services to their customers (victims).
Ransomware has been evolved gradually since
1989. It has expanded their scope of attack on devices
from computers to mobile devices, covering
individuals, enterprises, governments, medical
institutions, banking systems, etc. A number of attack
cases have been reported, including Hollywood
hospital network system and Muni subway in San
Francisco. They targeted the healthcare industry and
government organizations because it makes
significant impact to the public. In 2016, the first
ransomware for mobile devices, called “Gooligan”
has been reported, which led to 1 million Android
devices being attacked, by maliciously obtaining the
root permission of the devices (Adhikari, 2016).
There are a number of methods to inject
ransomware into edge servers: 1) by injecting
malicious scripts/codes or 2) by compromising an
edge server to spoof other edge servers. In 1), an
escape character can be used to attach malicious
string into a SQL query that can spoof the database to
execute the string, loading unexpected file remotely
into an edge server (victim) (Anley, 2002). Similarly,
XSS in HTML/JavaScript allows loading expected
codes remotely into the edge server (victim) as the
victim does not verify XSS codes (Martin & Lam,
2008). On the other hand, in 2), as edge servers are
interconnected and would work collaboratively with
each other, they would exchange data for processing.
If an edge server (victim) is spooled to listen to a
compromised edge server (attacker), the victim will
execute malicious codes from the attacker. XML
signature wrapping is commonly used when
launching the attack (McIntosh & Austel, 2005).
Nevertheless, ransomware attack is not detectable
immediately when it is taking place in an edge server.
In case of being attacked, the IoT system would face
the data integrity and data accuracy problems, which
may not cause significant harm to them. However, the
problems for data-oriented IoT systems (Tse et al,
2018) (Tse & Pau, 2016) (Aguiari et al, 2018) may
affect the quality of critical decision the systems
make. Thus, a self-recovery function against
ransomware attack should be enabled in edge servers.
In the next section, we present the Self-recovery
service (SRS), which secures edge servers against
ransomware attack.
3 SELF-RECOVERY SERVICE
In this paper, the Self-recovery service (SRS) is
proposed to ease the damage caused to an edge server
by ransomware. SRS is a system service that runs at
the kernel level, monitoring the activity of the file
IoTBDS 2020 - 5th International Conference on Internet of Things, Big Data and Security
400
system in an edge server. It is particularly sensitive to
the ransomware attack. If any suspected activity in
storage is detected, it checks for the ransomware
signature. If the signature is found in any IoT data
files, data files should be unexpected encrypted. It is
confirmed that the server is under the ransomware
attack. To resolve this situation, SRS takes an
immediate action to recover the encrypted file, by
recalling its backup copy from the backup node. If
there is no suspected activity, SRS continues to send
IoT data to the backup node.
It is noteworthy that an edge server may be free
from ransomware attack if IoT data is not kept in its
storage. However, some existing IoT systems (Tse et
al, 2018) (Tse & Pau, 2016) (Aguiari et al, 2018)
working in the environment without stable Internet
connection is required to keep the data in local storage
temporarily. In this context, those systems are
vulnerable to the ransomware attack during the
blackout period. Thus, SRS is highly recommended.
3.1 Architectural Design
SRS provides a seamless integrating solution that
does not require complicated modification to existing
IoT systems. Figure 1 shows the architectural design
of SRS.
Figure 1: Architectural Design of SRS.
As can be seen in Figure 1, there are a number of
IoT devices (i.e., camera, smartphone and lightbulb)
connecting to the edge server via Local Area Network
(LAN). The edge server collects IoT data from the
devices and immediately forwards them to the backup
node before doing any pre-processing on the data.
The backup node, which is located in the same subnet,
acts as a secondary storage for the edge server only.
When the recovery process is activated in the edge
server, it will recall the data from the backup node.
Once the recovery process is done and the data is
processed, it is then sent to the backend server
through the internet for further processing. It ensures
that the entire IoT system operates continuously.
3.2 Backup Node
In SRS architecture, the backup node will receive IoT
data periodically from the edge server using a secure
network connection, e.g., sftp. Data is received in a
form of raw data format. When it is stored in the file
system, extra information, such as its originality, will
be associated to it. Moreover, SRS maintains the data
consistency by Subversion server, tracking its
generation date, time and version. The tracking
operation also covers system files and folders, in
order to recover the edge server if its system is
attacked.
To detect the ransomware in the edge server, SRS
monitors the activity of the file system. The detection
logic will be explained in the next section.
3.3 Ransomware Detection
Ransomware attack typically aims to take control of
important files in a computing system by
cryptographic encryption. There are a number of file
types that ransomware is interested in, which are the
user files (e.g., docx, pptx, pdf, etc.) in home folder
and system files in the system folders. In addition,
other particular files and folders may be their targets
depending on their goal. Thus, in SRS, the detection
logic is designed to work at the file I/O level. It is
rule-based, monitoring the activity of local file
system. There are two rules in the logic, as listed
below.
Rule 1) File removal for over 30% of data files in
local file system; and
Rule 2) Change of file extension in any physical file
path.
In Rule 1), if data files are encrypted under
ransomware attack, their original file would be
deleted. If there are more than 30% of data files
deleted in a short period of time, it is assumed that
ransomware is taken place. This does not disturb the
normal operation of the edge server as the probability
of the deletion is very low. In Rule 2), if the file
extension of data files is renamed into something
unrecognizable in the system, it is also assumed to be
under ransomware attack. Figure 2 outlines the
decision making of the Detection logic.
LA
N
Recovery
Backup node
Edge server
Backend
server
Int
e
rn
e
t
IoT data
processed
data
IoT data
Self-recovery Service Securing Edge Server in IoT Network against Ransomware Attack
401
Figure 2: Ransomware Detection Logic.
When SRS finds an edge server under
ransomware attack, the recovery process will be
activated automatically and the system alert will be
sent out for ransomware removal.
3.4 Recovery Process
When ransomware attack is detected, the recovery
process will be activated instantly. A secure network
connection will be established to the backup node,
requesting for the IoT data for a certain period of
time. It is noted that the recovery process will not be
activated when the Internet connection is not
available. When the data is received, it will be sent to
the backend server after pre-processing.
4 EVALUATION
A simulation test has been conducted on SRS for
verifying its correctness. In the test, for verification
purpose, a standard computer running windows 7 SP2
operating system is used as the test platform. The core
hardware component of the computer is listed in
Table 1. It is noted that the computing power for an
edge server depends on the need of the IoT system. It
does not need to be a high-ended computer. A
raspberry pie device may be used as an edge server.
Table 1: Configuration of Simulation Test Platform.
Component Type
CPU i7 – 6700 HQ (8-core)
RAM 8 GB
Storage SSD
As a system service running in the background,
SRS takes system resources when it works. To ensure
that it does not degrade the system performance, the
test aims to verify the disk performance (I/O), CPU
performance, and the disk average transfer rate of the
system.
4.1 SRS Performance
(a) Read/Write Speed
(b) Writing Big File
Figure 3: Disk Performance.
In the simulation test, as can be seen in Figure 3(a),
the disk read/write speed is about
1077.6MB/1104.03MB per second when SRS is not
running. When it is running, the read/write speed
drops down to 1019MB/1036.01MB per second,
which is about 5.44%/6.16%. In the meantime, the
access (read/write) time increases, while SRS is
running, from 0.5ms/0.289ms to 0.519ms/0.296ms,
which is about 3.8%/2.42%.
To further test the disk performance, a binary file
of 1 GB in size was used in unzipping and writing
operations. (See Figure 3(b)) When SRS is not
running, it takes about 13.93 seconds in unzipping
(for testing the read speed). When SRS is running, it
takes about 14.92 seconds. There is around 1 second
time putting on the unzip operation by SRS which is
about 7%. On the other hand, it takes about 22.25
seconds in the writing operation when SRS is not
running. When it is running, it takes about 25.32
seconds. It is about 3 seconds (13.8%) of time putting
on the operation by SRS. The performance in disk
writing is slightly degraded.
Moreover, it revealed that the CPU performance
is slightly degraded. It takes around 0.2% (1.195% -
0.917%) of CPU time when SRS is running. The test
Detection Logic
1) File Removal (30%)
2) Change of File
Extension
Recovery Process
System Alert
IoTBDS 2020 - 5th International Conference on Internet of Things, Big Data and Security
402
Table 2: Summary of the Test.
SRS is NOT running SRS is running Result
Disk Read/Write Speed (MB per second)
1077.6MB/s,
1104.03MB/s
1019MB/s,
1036.01MBs
5.44%/6.16% DROP
Disk Access Time (Millisecond) 0.5ms/0.289ms 0.519ms/0.296ms 3.8%/2.42% INCREASE
UNZIP a 1GB File (Second) 13.93s 14.92s 7% INCREASE
Write a 1GB File (Second) 22.25s 25.32s 13.8% INCREASE
CPU Usage (%) 0.917% 1.195% 0.2% INCREASE
Disk Average Transfer Rate (Second) 1.051s 1.253s 0.2s INCREASE
also showed that the disk average transfer rate is
about 1.051 seconds when SRS is not running. When
it is running, the transfer rate is about 1.253 seconds,
which is about 0.2 seconds increase in the transfer
operation. Table 2 summarized the result of the test.
The result shows that SRS take insignificant
system resources from the system in terms of the disk
performance, CPU performance and disk average
transfer rate. However, it ensures the continuity of the
operation of edge server after ransomware attack.
5 CONCLUSIONS
Ransomware is arbitrary and destructive. Its harm to
IoT systems may create disastrous consequences. To
avoid the disaster happens, preventive measures are
commonly employed. SRS is one of the solutions that
eases the damages caused to edge servers by
ransomware. It ensures the continuity of the operation
of edge server after ransomware attack. It also ensures
that the data integrity and accuracy issues would not
be raised in IoT systems. In addition, SRS takes
insignificant system resources for its operation that
does not degrade the performance of the edge server.
REFERENCES
Fernández, C. M., Rodríguez, M. D., & Muñoz, B. R.
(2018). An Edge Computing Architecture in the
Internet of Things. In ISORC, 21st International
Symposium on Real-Time Distributed Computing,
IEEE, Singapore, 99-102. https://dx.doi.org/10.1109/
ISORC.2018.00021
Lopez, P. G., Montresor, A., Epema, D., Datta, A.,
Higashino, T., Iamnitchi, A., Barcellos, M., Felber, P.
& Riviere, E. (2015). Edge-Centric Computing: Vision
and Challenges. SIGCOMM Computer Communica-
tion Review, 45(5), 37–42. https://dx.doi.org/
10.1145/2831347.2831354
Shi, W., Cao, J., Zhang, Q., Li, Y. & Xu, L. (2016). Edge
Computing: Vision and Challenges. IEEE Internet
Things Journal, 3(5), 637–646. https://dx.doi.org/
10.1109/JIOT.2016.2579198
Statista. (2017). Edge Computing Market Size Forecast in
the United States From 2017 to 2025, by Segment (in
Million U.S. Dollars). Retrieved from https://www.
statista.com/statistics/909308/united- states-edge-
computing-market-size-segment/
Statista. (2020). Internet of Things (IoT) connected devices
installed base worldwide from 2015 to 2025. Retrieved
from https://www.statista.com/statistics/471264/iot-
number-of-connected-devices-worldwide/
Xiao, Y., Jia, Y., Liu, C., Cheng, X., Yu, J. & Lv, W.
(2019). Edge Computing Security: State of the Art and
Challenges. In Proceedings of the IEEE, 107(8), 1608-
1631. https://dx.doi.org/10.1109/JPROC.2019.29
18437
Alrowaily, M. & Lu, Z. (2018). Secure Edge Computing in
IoT Systems: Review and Case Studies. In SEC
IEEE/ACM Symposium on Edge Computing, Seattle,
WA, 440-444. https://dx.doi.org/10.1109/SEC.2018.
00060
Hafeez, I., Ding, A. Y., Suomalainen, L., & Tarkoma, S.
(2016). Demo Abstract: Securebox — A Platform to
Safeguard Network Edge. In SEC IEEE/ACM
Symposium on Edge Computing, Washington, DC, 117-
118. https://dx.doi.org/10.1109/SEC.2016.12
Self-recovery Service Securing Edge Server in IoT Network against Ransomware Attack
403
Caprolu, M., Pietro, R. D., Lombardi, F. & Raponi, S.
(2019). Edge Computing Perspectives: Architectures,
Technologies, and Open Security Issues. In EDGE
IEEE International Conference on Edge Computing,
Milan, Italy, 116-123. https://dx.doi.org/10.1109/
EDGE.2019.00035
Zhang, J., Chen, B., Zhao, Y., Cheng, X. & Hu, F. (2018).
Data Security and Privacy-Preserving in Edge
Computing Paradigm: Survey and Open Issues. IEEE
Access, 6, 18209-18237. https://dx.doi.org/10.1109/
ACCESS.2018.2820162
O'Gorman, G., & McDonald, G. (2012). Ransomware: A
Growing Menace. Symantec Corporation. Retrieved
from https://www.symantec.com/content/en/us/
enterprise/media/security_response/whitepapers/ranso
mware-a-growing-menace.pdf
Gazet, A. (2010). Comparative Analysis of Various
Ransomware Virii. Journal in Computer Virology, 6(1),
77-90.
Johansen, A. G. (2018). What is Ransomware and How to
Help Prevent Ransomware Attacks. Symantec.
Retrieved from https://us.norton.com/internetsecurity-
malware-ransomware-5-dos-and-donts.html
Adhikari, R. (2016). Gooligan Ransacks More than 1M
Android Accounts. TechNewsWorld. Retrieved from
https://www.technewsworld.com/story/84132.html
Anley, C. (2002). Advanced SQL Injection in SQL Server
Applications. CGI Security. Retrieved from https://
www.cgisecurity.com/lib/advanced_sql_injection.pdf
Martin, M. & Lam, M. S. (2008). Automatic Generation of
XSS and SQL Injection Attacks with Goal-Directed
Model Checking. In 17th Conference on Security
Symposium, Berkeley, CA, USA, 31-43.
McIntosh M., & Austel, P. (2005). XML Signature Element
Wrapping Attacks and Countermeasures. In Workshop
Secure Web Services, 20–27. https://dx.doi.org/
10.1145/1103022.1103026
Tse, R., Aguiari, D., Chou, K.-S., Giusto, D., Tang, S.-K.,
& Pau, G. (2018). Monitoring Cultural Heritage
Buildings via Low-Cost Edge Computing/Sensing
Platforms: The Biblioteca Joanina de Coimbra Case
Study. In GOODTECHS, 4th EAI International
Conference on Smart Objects and Technologies for
Social Good. ACM Press. New York, NY, 148-152.
https://dx.doi.org/10.1145/3284869.3284876
Tse, R., & Pau, G. (2016). Enabling Street-Level Pollution
and Exposure Measures. In MobiHealth, 6th ACM
International Workshop on Pervasive Wireless
Healthcare. ACM Press. New York, NY, 1–4.
https://dx.doi.org/10.1145/2944921.2944925
Aguiari, D., Delnevo, G., Monti, L., Ghini, V., Mirri, S.,
Salomoni, P., Pau, G., Im, M., Tse, R., Ekpanyapong,
M., & Battistini, R. (2018). Canarin II: Designing a
Smart e-Bike Eco-System. In CCNC, 15th IEEE Annual
Consumer Communications Networking Conference.
IEEE. https://dx.doi.org/10.1109/CCNC.2018.
8319221
IoTBDS 2020 - 5th International Conference on Internet of Things, Big Data and Security
404