Towards a Deep Investigation of Cybersecurity Issues in Cyber-Physical Systems Using a Novel Graph Neural Network-Based Threat Analysis Approach

Activity: Talk or presentation / LecturePresentation at a scientific conference / workshop

Description

Cyber-physical systems (CPS) are complex, comprising multiple physical and cyber system components that communicate to provide particular actions. The complexity of connectivity in such systems is based on multiple factors, including heterogeneous components, real-time analysis, and security challenges. Addressing security issues in CPS is a pivotal challenge in their design. An inapplicable set of security measures for the system components can lead to safety breaches, reduce system reliability, and affect overall system efficiency. This highlights the critical need to identify any existing security vulnerabilities in the system that could lead to such consequences. Threat analysis is one of the leading approaches that can support the identification of existing security vulnerabilities and the detection of potential cyber threats that could lead to cyberattacks. However, in complex systems such as CPS, this creates challenges for traditional threat analysis approaches in deeply investigating all connected components and related security measures to determine if any security issues exist. The utilization of Graph Neural Networks (GNNs) in the context of threat analysis is considered to be one of the more suitable Artificial Intelligence (AI) approaches for coping with challenges. GNNs are based on the graph structure for learning node representations in the dataset, defining multiple parameters for each node and edge. This results in a more complex network function capable of solving difficult problems, such as deeply investigating complex interconnections and relations between system components and identifying potential threats. Therefore, we introduce a novel threat analysis approach for modeling a complex set of interconnected nodes, each representing system-related components and assigning specific features representing security measures that need to be investigated. This approach is designed to identify potential cyber threats that could be propagated due to missing or insufficient security measures, which could lead to successful cyber attacks among multiple interconnected nodes. The outcomes of our proposed approach will support the continuous updating of the system design with an applicable set of security measures to keep cyber risk at an acceptable level. This will aid the system's security architecture in making informed decisions about reducing cyber risks based on the identified security issues.
Period17 Jan 202419 Jan 2024
Event titleHiPEAC (High Performance, Edge And Cloud computing) : ENHANCE: Enabling Technologies and Dependability in Cyber-Physical Systems
Event typeConference
LocationMunich, GermanyShow on map
Degree of RecognitionInternational

Research Field

  • Dependable Systems Engineering

Keywords

  • Cybersecurity
  • Cyber-Physical System
  • Graph Neural Network (GNN)