You are on page 1of 5

The 9th IEEE International Conference on Dependable Systems, Services and Technologies, DESSERT’2018

24-27 May, 2018, Kyiv, Ukraine

Availability Models for Healthcare IoT Systems:


Classification and Research Considering Attacks
on Vulnerabilities
Anastasiia Strielkina, Vyacheslav Kharchenko, Dmytro Uzun
Computer Systems, Networks and Cybersecurity Department, National Aerospace University “KhAI”,
17, Chkalova Street, Kharkiv, Ukraine, 61070,
a.strielkina@csn.khai.edu, v.kharchenko@csn.khai.edu, d.uzun@csn.khai.edu, https://csn.khai.edu/

Abstract—Networked medical and healthcare devices systems are characterized by a large number of failures
and their applications are already creating an Internet of due to the dynamism, multicomponence and
Medical Things which is aimed at better health monitoring multilevelness. For reducing this issues the
and preventive care. But the new concepts and applying of fragmentedness of the models being developed should be
new technologies bring certain risks including failures of
devices, infrastructure which may lead to the worst
used in some cases to describe repeated parts of models
outcome. This paper presents an approach to develop a which have similar structure and differ only values of
Markov models set for a healthcare IoT infrastructure that some parameters. It concerns fragmentedness caused by
allows taking into account safety and security issues. In changing of design faults and attacked vulnerabilities
addition, it is presented a case study with a Markov model number and the corresponding failure rates.
considering attacks on vulnerabilities of the healthcare IoT
system. B. Related Works Analysis
For today there are a lot of papers that describe
Keywords—attack; failure; Internet of Things; opportunities and benefits of healthcare IoT and medicine
healthcare; Markov model; security; vulnerability and at the same time they emphasize on issues of this
technology as the security and dependability.
I. INTRODUCTION The authors of [4] presented all healthcare IoT trends,
solutions, platforms, services and applications. They
A. Motivation outlined main problems during development and using of
The Internet of Things (IoT) has an important such devices related mostly to standardization and
technical, social and economic significance. According to regulatory issues. In addition, that paper analyzed
the preliminary forecasts [1], [2], about 50 billion devices healthcare IoT security and privacy features, including
will be connected to the Internet and the IoT market will requirements, threat models, and attack taxonomies and
reach about $1.7 trillion by 2020. proposed an intelligent collaborative security model to
The IoT can significantly improve the existing minimize security risk.
healthcare system as well as other domains. Combining A healthcare IoT infrastructure with a brief
the capabilities of the Internet of Things with medical description of each component was presented in [5]. The
equipment can greatly improve the quality and constituent components of the healthcare IoT
effectiveness of healthcare services and create better infrastructure are: Cloud, end-node (patient’s device),
conditions for patients who need constant medical and medical or healthcare providers and communication
healthcare supervision and / or preventive intervention. channels between: device and Cloud, and healthcare
One of the most widely used IoT devices is an insulin provider and Cloud. The authors gave basic cases with a
pump. According to the statistics of the World Healthcare few models of healthcare IoT system based on the
Organization presented in 2016 [3], about 8.5% of the queueing theory. The models describe streams of the
world population had diabetes in 2014. The networked requests and attacks on vulnerabilities and procedure of
insulin pump can be placed in an inconspicuous place recovery by restarting and eliminating of one and / or two
under the patient’s clothes, so a patient can carry out and vulnerabilities.
control the injection of insulin with a special console or The existing standards and regulatory acts in the field
smartphone. of healthcare [6-8] provide different approaches for risk
Nevertheless, with all the benefits of using such management to determine safety and security of medical
networked devices, the security and dependability risks and healthcare devices.
are increasing. Thus, the dependability and security Attacks on vulnerabilities of IoT based systems can
assessment of such systems is a complex process. Such be simulated using Markov’s modelling. In [9] was

978-1-5386-5903-8/18/$31.00 ©2018 IEEE 58

Authorized licensed use limited to: Manchester Metropolitan University. Downloaded on January 21,2021 at 16:01:52 UTC from IEEE Xplore. Restrictions apply.
presented and explained how Markov modelling can be M RFV is a model that takes into account all possible
I RD
used to evaluate the reliability of the complex systems
parallel redundant system. In [10] was shown that states of the the healthcare IoT system (functional states,
dependability consists of many measures (as reliability, vulnerabilities and reliability issues).
availability, safety, performability and security and it’s F – is the cardinality of a functional states set
attributes). Authors presented a good state-of-the-art how 
F = SF0 ,SF1 ,...,SFm . 
Markov models can be applied to the dependability and
When F=1 the set is a degenerate (e.g., only request
security analysis.
function). This case can be described using Markov
A Markov model, which takes into account the
process with discrete states and continuous time i.e. the
technical conditions of typical network components of the 10
IoT-based smart business center was presented in [11]. birth–death process (see Fig. 2 for M 00 , where n is a
The authors emphasized that the analysing dependability number of requests in the queue incoming from different
IoT system must take into account the reliability and end IoT users, λ01, λ12, λn-1n are intensities of service
security of the system at several levels IoT architecture. requests flow, μ10, μ21, μnn-1 are intensities of processing
In [12] the metric-based approach based on the flow) and in more details it was described in [5].
analysis of Markov chain’s stiffness, decomposability,
sparsity and fragmentedness was shown. According to the
authors the fragmentedness can be achieved dividing the
model into N fragments (they have the similar structure
but different parameters). Such approach may help to
present interconnection between different parts of system. Figure 2. A scheme of “Birth-death” process for the healthcare IoT
As a result of the foregoing the task of Markov behavior
modelling of healthcare IoT systems is in demand and
relevant. When F > 1 the model takes into account different modes
of operation of healthcare IoT infrastructure (e.g., readout,
C. Goals and Structure request, monitoring, sleep mode, injection (for insulin pump),
In this paper, it is proposed to develop a Markov F0
etc.). This case is depicted on Fig. 3 for M 00 .
models set for the healthcare IoT infrastructure that allows
taking into account the specificity of end user devices,
communication channels, technologies of data flows and
safety and security issues of these components.
The paper is structured as follows. The section II
presents a brief description of models sets for the
healthcare IoT system based on Markov process approach.
The section III presents a case study with a Markov model Figure 3. Graph describing the functional states of the healthcare IoT
infrastructure
considering attacks on vulnerabilities of the healthcare IoT
system followed by concluding remarks. The section IV V – is the cardinality of the security vulnerabilities set
concludes and describes future research directions.
II. DESCRIPTION OF MODELS SETS FOR THE

V = SV0 ,SV1 ,...,SVp ,  where p is a number of

HEALTHCARE IOT SYSTEM BEHAVIOR vulnerabilities. When V = 0 the set is a degenerate; when
V  1 the modes of attack reflection and eliminating of
For the description formalization of the healthcare IoT attacks are taking into account. As the vulnerabilities are
system behavior, it is proposed to apply the set-theoretic eliminated the number of attacks decreases.
model that would adequately represent both the structure
RI – is the cardinality of the degradation levels of the
of the system and the cause-effect relationships between
its components. Set-theoretical model that describes a set 
healthcare IoT infrastructure set RI = S RI ,S RI ,...,S RI
0 1 k
,
of models behavior of the healthcare IoT infrastructure is i.e. ability to serve applications from device nodes. When
presented in Fig. 1. RI = 0 failures are not taken into account, i.e. top level –
10
M 00 F0
M 00 nondegrading infrastructure. For the case M 10 RI 0 Markov

M 10 M 010RD M 10 M RFI00 M 0FR0D M RFI 0RD model is presented in Fig. 4 (where λ0’01, λ1’11, … are
RI 0 RI RD
intensities of infrastructure faults, μ’010, μ’212, … are
1V FV
M 00 M 00 intensities of recovering). RD – is the cardinality of the
devices (for this paper networked insulin pump) failures
M 1RVI 0 M 01VRD M 1RVI RD M RFV M 0FV M RFV
I0 RD I RD
 
set RD = S RD ,S RD ,...,S RD , and when RD = 0 failures
0 1 l

Figure 1. Model range for healthcare IoT systems are also not taken into account. For the case M 010RD

59

Authorized licensed use limited to: Manchester Metropolitan University. Downloaded on January 21,2021 at 16:01:52 UTC from IEEE Xplore. Restrictions apply.
Markov model is presented in Fig. 5 (λ0’’01, λ1’’11, … are papers presented attacks’ targets, weaknesses, and
intensities of end-deviced faults, μ’’010, μ’’212, … are technique of the security attacks. The main categories of
intensities of recovering). IoT attacks are aimed for control, data, controllers (end-
nodes) and networks. Attacks on data are very devastating
in the healthcare field due to the physician–patient
privilege and a patient privacy and confidentiality. Attacks
on control involve imply an intruder's intention to gain
access to the management of both the entire healthcare
IoT system and individual components. Attacks on
controllers are aimed at end-nodes (patients’ devices) to
gain access to control them and make a physical damage.
Attacks on networks are aimed to sniffing out, copying the
confidential information or any other data flowing in the
networks.
After analyzing classification of attacks according the
main aims and focus is presented in Fig. 6.
Figure 4. Graph describing faults of the healthcare IoT
infrastructure

Figure 6. Classification of healthcare IoT attacks

Such attacks on vulnerabilities can prevent the devices


Figure 5. Graph describing faults of the devices (end nodes) in and infrastructure to communicate correctly and without
healthcare IoT infrastructure
failures.
Based on the developed models in the context of models B. Markov Model Considering Attacks on Components
behavior set of the healthcare IoT infrastructure it can be Vulnerabilities of Healthcare IoT System
concluded that the use of such models is promising for
According to the approach described above in the
research, analysis and improvement of the healthcare IoT 1V
infrastructure on account of multilevelness decomposition, Section II for the case M 00 , a simplified Markov’s graph
fragmentedness, inheritance and convolution. is presented below. The model does not take into account
eliminating of vulnerabilities and design faults. The failure
III. CASE STUDY: A M ARKOV MODEL CONSIDERING and/or attack, recovery and/or repulse rates are constant.
ATTACKS ON VULNERABILITIES OF HEALTHCARE IOT Fig. 7 shows a Markov’s graph of the main components
SYSTEM functioning of the healthcare IoT system during attacks, λ -
the failure and/or attack rate, µ - the recovery and/or
A. Analysis of the Origin Nature and Classification of reflection rate. Thereby, the basic states of the healthcare IoT
Possible Cyber Attacks system are: 1 - normal condition (upstate) system; 2 – traffic
Since attacks are the possible consequences of the analysis attack; 3 – spoofing attack; 4 – cloning attack; 5 –
threat implementation the existing vulnerabilities. unauthorized access to the network or database; 6 – failure
Therefore it is necessary to consider the attack as a of the network; 7 – failure due the data leakage; 8 – man-in-
malicious action affecting the healthcare IoT system's the-middle attack; 9 – DoS/DDoS attack; 10 – failure due
performance. the loss of control; 11 – attacks on software (i.e., viruses,
In [13] mentioned that about 250 cyber attacks were worms, Trojan horses, spyware, adware, etc.); 12 – phishing
targeted on health sector (only publicly disclosed attack; 13 – malicious scripts injection attacks; 14 – social
incidents) in 2016-2017. There are several types of attacks engineering; 15 – failure or controllers (hardware, end-
on IoT that were discussed in many papers [14]–[16]. The nodes); 16 – tampering of the end-nodes attack; 17 –
security issues of insulin pump in the context of cyber- jamming attack; 18 – malicious node injection attack; 19 –
security systems were shown in [17]. The authors of these

60

Authorized licensed use limited to: Manchester Metropolitan University. Downloaded on January 21,2021 at 16:01:52 UTC from IEEE Xplore. Restrictions apply.
physical damage; 20 – complete failure of healthcare IoT
system.

Figure 8. Dependence of the changing of the availability function


depending on the changing λ15 rate

Figure 7. Markov’s graph of attacks on healthcare IoT


infrastructure
Figure 9. Dependence of the changing of the availability function
Such Markov model for the healthcare IoT system can depending on the changing λ19 rate
be divided into four levels: the first level is the upstate (state
1), the second level are the states state the transition to which
occurred due to attacks (states 2-5, 8, 9, 11-14, 16-19), the
third level implies states with different failures (states 6, 7,
10, 15); the fourth level is a failure of a whole system (state
20). So, there is only one up-state in the considering system.
After solving the system of Kolmogorov-Chapmen
equations, it is possible to obtain the availability function
value of the healthcare IoT system, the number of system
failures due to cyber attacks.
To solve a system of linear Kolmogorov differential
equations it is necessary to collect and analyse statistics on
failures and attacks on the healthcare IoT infrastructure. The
data for model simulation was taken from [18], [19] (for the Figure 10. Dependence of the changing of the availability function
healthcare cyber attacks [20] – [22]) and experts’ depending on the changing λ111 rate
assessments.
The obtained steady-state probabilities for the considered
Markov model are:
P1(t) = 0.9183149; P2(t) = 3.43523e-05;
P3(t) = 2.292919e-05; P4(t)= 0.0001693283;
P5(t) = 0.0002168167; P6(t) = 0.006018265;
P7(t) = 0.04055902; P8(t) = 4.878616e-05;
P9(t) = 0.0001007445; P10(t) = 0.01365683;
P11(t) = 3.79046e-05; P12(t) = 0.0001206337;
P13(t) = 1.514862e-05; P14(t) = 0.001278129;
P15(t) = 0.001488987; P16(t) = 0.0006177497;
P17(t) = 0.0104252; P18(t) = 0.0005000198;
P19(t) = 0.005830752; P20(t) = 0.0005435141. Figure 11. Dependence of the changing of the availability function
Fig. 8 – 11 show the dependence of the changing of the depending on the changing λ114 rate
availability function depending on the changing of the
different types of attacks on the healthcare IoT systems rates.

61

Authorized licensed use limited to: Manchester Metropolitan University. Downloaded on January 21,2021 at 16:01:52 UTC from IEEE Xplore. Restrictions apply.
The analysis of the obtained results shows that the [3] Global report on diabetes, World Healthcare Organization, 2016,
88 p.
biggest influence on the change in the availability function
[4] S.M.R. Islam, D. Kwak, Md.H. Kabir, M. Hossain, K.-S. Kwak,
has the λ19 rate – DoS/DDoS attack (one of the most "The Internet of Things for Health Care: A Comprehensive
frequent and destructive attacks), next is the λ114 rate – Survey", IEEE Access, vol. 3, pp. 678-707, 2015. DOI:
social engineering (i.e., human factor), next the rates λ15 – 10.1109/ACCESS.2015.2437951.
[5] A. Strielkina, D. Uzun, V. Kharchenko, “Modelling of healthcare
unauthorized access to the network or database and λ111 –
IoT using the queueing theory,” in 9th IEEE International
attacks on software (i.e., viruses, worms, Trojan horses, Conference on Intelligent Data Acquisition and Advanced
spyware, adware, etc.), that are confirmed by statistical Computing Systems: Technology and Applications (IDAACS), vol.
data. 2, 2017, pp. 849-852. DOI: 10.1109/IDAACS.2017.8095207.
[6] “Applying Human Factors and Usability Engineering to Medical
IV. CONCLUSIONS AND FUTURE WORK Devices: Guidance for Industry and Food and Drug
Administration Staff,” U.S. Department of Health and Human
Integrating sensors, devices and people into a single Services Food and Drug Administration Center for Devices and
network, IoT technologies enable new opportunities for Radiological Health Office of Device Evaluation, 2016.
[7] Guidance for Industry: Q9 Quality Risk Management, U.S.
interaction between a person and a machine, software and Department of Health and Human Services Food and Drug
equipment. However, with such great potential IoT leads Administration Center for Drug Evaluation and Research Center
to a variety of threats. for Biologics Evaluation and Research, 2006.
In this paper, a Markov models set for the healthcare [8] ISO 14971:2007 Medical devices -- Application of risk
management to medical devices, 2007.
IoT infrastructure that allows taking into account the [9] K. S. Trivedi and D. Selvamuthu, "Markov modeling in
specificity of end user devices, communication channels, reliability," in Encyclopedia of Quantitative Risk Analysis and
technologies of data flows and safety and security issues Assessment. John Wiley & Sons, Ltd, 2008. DOI:
of these components has been developed. One of the 10.1002/9781118445112.stat03635.
[10] D.M. Nicol, W.H. Sanders, and K.S. Trivedi, "Model-based
Markov models considering attacks on vulnerabilities of evaluation: From dependability to security", IEEE Transactions on
the healthcare IoT infrastructure has been developed and Dependable and Secure Computing, vol. 01, no. 1, pp. 48-65,
researched to implement the suggested approach. 2004. DOI: 10.1109/TDSC.2004.11.
Next steps of research will be dedicated to [11] V. Kharchenko, M. Kolisnyk, I. Piskachova, and N. Bardis,
“Markov model of the Smart Business Center wired network
development of safety and security models for healthcare considering attacks on software and hardware components,” in
IoT systems and combining results of this paper and future International journal of computers and communications, vol. 10,
developed models taking into account both the reliability, 2016, pp 113-119.
safety and security requirements and issues. [12] V. Butenko, V. Kharchenko, E. Odarushenko, and D. Butenko,
“Metric-based approach and tool for modeling the I&C system
ACKNOWLEDGMENT using Markov chains,” in Proceedings of ICONE-23 23rd
International Conference on Nuclear Engineering, Chiba, Japan,
This paper results from the Erasmus+ programme 2015, p. 9. DOI: 10.1299/jsmeicone.2015.23._ICONE23-1_279.
educational project ALIOT «Internet of Things: Emerging [13] McAfee Labs Threat Report, December 2017, 13 p.
[14] M. Abomhara, and G. M. Kien, “Cyber Security and the Internet
Curriculum for Industry and Human Applications» of Things: Vulnerabilities,Threats, Intruders and Attacks,” in J.
(reference number 573818-EPP-1-2016-1-UK-EPPKA2- Cyber Secur. Mobil, vol. 4, no. 1, 2014, pp. 65-88.
CBHE-JP, web-site http://aliot.eu.org) in which the DOI: 10.13052/jcsm2245-1439.414.
appropriate course is developed (ITM4 - IoT for health [15] M. Farooq, M. Waseem, A. Khairi, and S. Mazhar, “A Critical
Analysis on the Security Concerns of Internet of Things (loT)”, in
systems) within its framework, we have developed International Journal of Computer Applications, vol. 111, 2015, 6
modules related to IoT systems modelling. The authors p. DOI: 10.5120/19547-1280.
would like to thank colleagues on this project, within the [16] M. Nawir et al., "Internet of Things (IoT): Taxonomy of security
framework of which the results of this work were attacks", 3rd International Conference on Electronic Design, pp.
321-326, 2016. DOI:10.1109/ICED.2016.7804660.
discussed. The authors also would like to show deep [17] A. Humayed, J. Lin, F. Li and B. Luo, "Cyber-Physical Systems
gratitude to colleagues from Department of Computer Security—A Survey", IEEE Internet of Things Journal, vol. 4, no.
Systems, Networks and Cybersecurity, National 6, pp. 1802-1831, 2017. DOI: 10.1109/JIOT.2017.2703172.
Aerospace University «KhAI» for their patient guidance, [18] “Kaspersky security bulletin: overall statistics for 2017,”
Kaspersky Lab, 2017, 29 p.
enthusiastic encouragement and useful critiques of this [19] “The State of Industrial Cybersecurity 2017,” The Business
paper. Advantage Group Limited, 2017, 23 p.
[20] G. Bell, and M. Ebert, “Health care and cyber security: ncreasing
REFERENCES Threats Require Increased Capabilities,” KPMG International,
[1] D. Lund, C. MacGillivray, V. Turner, and M. Morales, USA, 2015, 7 p.
“Worldwide and Regional Internet of Things (IoT) 2014–2020 [21] A. Le Bris, W. El Asri, “State of cybersecurity & cyber threats in
healthcare organizations,” Essec Business School, 2016, 12 p.
Forecast: A Virtuous Circle of Proven Value and Demand,” IDC,
Framingham, MA, 2014. [22] “The state of cybersecurity in healthcare organizations in 2016,”
[2] “Understanding the Internet of Things (IoT)”, GSM Assosiation, Ponemone Institute, 2016, 32 p.
2014, 14 p.

62

Authorized licensed use limited to: Manchester Metropolitan University. Downloaded on January 21,2021 at 16:01:52 UTC from IEEE Xplore. Restrictions apply.

You might also like