Enhancing IDS Through Multi-Objective Optimization: A Focus on Attack Graph under Worst-Parent Attack

This study focuses on the evolution of network security models for RPL based LLNs under worst-parent attack. The multi-objective optimization approach integrates additional metrics: i) The amount of change in the maximum path length within the graph (∆L) and ii) the amount of change in the number of nodes connected to the attacker node(∆W) alongside the conventional accuracy criterion. The analysis of these metrics plays a crucial role in the development of Intrusion Detection Systems (IDS), particularly through the examination of attack graphs. Attack graphs are small, distinct network substructures used to model potential attacks on the network. Analyzing these graphs is considered essential for identifying security vulnerabilities in the network and determining the spread and impact of attacks. The findings of the study demonstrate a progressive improvement in model performance across tasks, starting from a single-objective task and advancing towards multi-objective tasks. This improvement signifies the effectiveness of integrating additional metrics in enhancing predictive accuracy and understanding network structure and resilience.