Abstract
dc:descriptionThe rising security threats in the Internet of Things correspond to the rapid proliferation of smart devices across homes, industries, and healthcare environments. Addressing these threats while maintaining low computational power and energy consumption has limited the range of viable solutions. Consequently, energy-efficient Machine Learning and Deep Learning techniques, capable of analyzing and classifying vast amounts of data in real-time, have been extensively researched and developed. In this context this thesis explores the application of a Graph Neural Network to analyze the structure of complex IoT networks and process network data from nodes and edges. The graph depiction of the network infrastructure reflects the communication among the interconnected entities, assisting in local identification of network breaches. We aim to create a sophisticated anomaly detection approach which identifies specific attack patterns based on their IP flow features, providing a robust and scalable solution in near-real time performance. A distributed attack detection powered by the Intelligent Agents is proposed to deploy the developed algorithm, leveraging the benefits offered by distributed systems, also minimizing power and bandwidth resource consumption. Evaluation of the proposed method is conducted through extensive lab experiments utilizing both online and simulated datasets, as well as real pilot testing. The aim is to assess the effectiveness of the approach in terms of detection accuracy and detection delay, and also to highlight the power consumption and bandwidth overhead across different attack scenarios and network infrastructure. The research conducted in this study was funded by the European Union as part of the SerIoT project. This project aimed to develop, implement and test a generic IoT framework by incorporating attack detection, mitigation, QoS and Energy-Aware Routing, advanced analytics and user-friendly data analytics with visual monitoring. The proposed advanced anomaly detection module was part of its overarching architecture, by contributing to the project's objective of improving network security and resilience against emerging threats in the IoT domain. The project was hosted at the Informatics and Telematics Institute (ITI) of CERTH, providing a strong research environment that facilitated its successful implementation.
Degree
thesis:*- Grantor dc:publisher
- Πανεπιστήμιο Μακεδονίας
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Protogerou, Aikaterini
- Πρωτογέρου, Αικατερίνη
Subjects
dc:subject × 16- Τεχνητή νοημοσύνη
- Πολυπρακτορικά συστήματα
- Νευρωνικά δίκτυα γράφων
- Ανίχνευση Επιθέσεων
- Διαδίκτυο των πραγμάτων
- Artificial Intelligence (AI)
- Multi-agent systems
- Graph neural networks
- Anomaly detection
- Internet of things
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Επιστήμες Μηχανικού και Τεχνολογία
- Αυτοματισμοί και Συστήματα ελέγχου
- Electrical Engineering, Electronic Engineering, Information Engineering
- Engineering and Technology
- Automation and control systems
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/59739