Abstract
dc:descriptionIn the digital era of smart economies, the Internet of Things (IoT) plays a significant role in Critical Infrastructures (CIs), providing several benefits such as improved productivity, efficient operation management and self-healing. However, this rapid evolution raises severe security and privacy issues, especially in critical environments. In particular, on the one hand, IoT introduces a set of new and heterogeneous technologies that rely on the vulnerable Internet model (Transmission Control Protocol/Internet Protocol (TCP/IP) networks). On the other hand, Industrial IoT (IIoT) environments are characterised by the presence of legacy systems that are prone to a wide range of security weaknesses and vulnerabilities. For instance, industrial communication protocols, such as Modbus/TCP, Distributed Network protocol 2 (DNP3) and IEC 60870-5-104 do not include any authentication and authorisation mechanisms. Therefore, cyberattackers are able to execute unauthorised activities, such as Man In The Middle (MITM) and Denial of Service (DoS) attacks. In addition, the IIoT-based CIs are an attractive target for a growing number of cyberattackers and Advanced Persistent Threats (APTs). Characteristic APT campaigns are Industroyer (also known as Crashoverride), Stuxnet and TRITON. Finally, it is worth mentioning that IIoT is characterised by constrained computing resources that do not allow the deployment of conventional security mechanisms.Based on the aforementioned remarks, it is evident that the presence of appropriate defensive mechanisms for the IoT paradigm is necessary. Therefore, the main goal of this PhD thesis is to investigate the security and privacy issues of IoT, providing effective security solutions. After the introductory chapter to this PhD thesis (Chapter 1), the security requirements, challenges and threats are examined in Chapter 2, paying special attention to the unique characteristics and constraints of the IoT entities. In particular, a new IoT threat taxonomy is provided, utilising the Common Attack Pattern Enumeration and Classification (CAPEC) system, while also the efficiency of the existing countermeasures is analysed. It is noteworthy that both academia and industry have provided sufficient solutions, such as the encryption and authorisation mechanisms of the IoT communication protocols. However, they cannot be adopted in IIoT ecosystems due to the role of Industrial Control Systems (ICS) and Supervisory Control and Data Acquisition (SCADA) systems. Consequently, the presence of Intrusion Detection and Prevention Systems (IDPS) is necessary.According to the countermeasure analysis of Chapter 2, next, Chapter 3 focuses on intrusion detection and mitigation mechanisms. In particular, a special emphasis is given to the architectural design and specifications of IDPS. In addition, the categories of IDPS are analysed, taking full advantage of Artificial Intelligence (AI) techniques. Moreover, mitigation and resilience mechanisms like honeypots and Software-Defined Networking (SDN) are discussed. Finally, this chapter analyses the role of Security Information and Event Management (SIEM) systems in IoT, paying special attention to the normalisation and correlation of the various security events.Next, in Chapter 4, a thorough analysis of the intrusion detection and mitigation mechanisms in the smart electrical grid (or Smart Grid (SG)) follows. In particular, SG is the largest IoT application, including multiple architectural elements, such as the Advanced Metering Infrastructure (AMI), ICS/SCADA systems, substations and synchrophasors. Therefore, for each of the previous SG elements, relevant intrusion detection and mitigation solutions (such as IDPS, honeypots and SIEM) are studied in terms of their architecture and detection performance. Based on this comparative analysis, the strengths and limitations of the existing solutions are further discussed, thus guiding the implementation of the proposed SDN-enabled SIEM in the following chapter.Subsequently, Chapter 5 presents the proposed SDN-enabled SIEM, which is composed of three IDPS, namely (a) Network Flow-based IDPS (NF-IDPS), (b) Host-based IDPS (H-IDPS) and (c) Visual-based IDPS (V-IDPS). First, NF-IDPS incorporates multiple ML and DL models that can discriminate cyberattacks and anomalies against a wide range of industrial communication protocols. Next, H-IDPS can recognise operational anomalies against four SG environments: (a) hydropower plant, (b) substation, (c) power plant and (d) smart home. Finally, V-IDPS can detect Modbus/TCP cyberattacks, using visual representations and Convolutional Neural Networks (CNNs). It is worth mentioning that the proposed V-IDPS follows a self-learning approach that can re-train the pre-trained CNN during the inference mode. The security events of the previous IDPS are normalised, correlated and mitigated by the Normalisation, Correlation and Mitigation Engine (NCME). For the correlation process, custom security rules are used, while mitigation relies on the SDN controller. In particular, NCME adopts a Reinforcement Learning (RL) agent, which guides the SDN controller about the appropriate mitigation actions. Finally, NCME includes a sophisticated honeypot deployment mechanism, which relies on a honeypot security game between two players: (a) attacker(s) and (b) defender(s). For the previous honeypot security game, the Nash Equilibrium (NE) is identified, while when NE is not available based on the parameters of the game, two alternative approaches are provided: (a) maxmin-based honeypot development and AI-powered honeypot deployment. Then, after the description of the proposed detection and mitigation solutions, Chapter 6 summarises the respective evaluation results. In particular, first, a concrete evaluation framework is defined, introducing (a) the evaluation environments, (b) the datasets, (c) comparative methods and (d) the evaluation metrics. Next, the evaluation results for each component of the proposed SDN-enabled SIEM are summarised and discussed, while Appendices M-N provide a detailed comparative analysis.Finally, Chapter 7 concludes this PhD thesis, providing also potential directions for future research work in this field. In particular, after discussing the key results studied and implemented in the context of this PhD programme, next, five research directions for future work are discussed: (a) Intrusion and Anomaly Detection using Federated Learning (FL), (b) Correlation Mechanisms using Association Learning, (c) RL-based Mitigation Strategies, (d) SDN-powered recovery mechanisms using Graph Neural Networks (GNN) and (e) Explainable AI (XAI) Techniques for AI Detection and Mitigation Models.
Degree
thesis:*- Grantor dc:publisher
- University of Western Macedonia
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Radoglou-Grammatikis, Panagiotis
- Ράδογλου-Γραμματίκης, Παναγιώτης
Subjects
dc:subject × 22- Ανίχνευση εισβολών
- Βιομηχανικά συστήματα ελέγχου
- Διαδίκτυο των πραγμάτων
- Ιδιωτικότητα
- Κυβερνοασφάλεια
- Πρόληψη Εισβολών
- Τεχνητή νοημοσύνη
- Τεχνολογία Καθοριζόμενη από Λογισμικό
- Artificial intelligence
- Cybersecurity
- Industrial control systems
- Internet of things (IoT)
- Intrusion detection
- Intrusion Prevention
- Privacy
- Software-defined networking (SDN)
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Επιστήμες Μηχανικού και Τεχνολογία
- Μηχανική και συστήματα επικοινωνιών, Τηλεπικοινωνίες
- Electrical Engineering, Electronic Engineering, Information Engineering
- Engineering and Technology
- Communication engineering and systems, Telecommunications
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/53552