Advancing Cyber Resilience for Autonomous Systems with Novel AI-based Intrusion Prevention Model

Abstract

Autonomous systems rely on intricate algorithms and are networked, which makes them more susceptible to cyber-attacks. The contexts of traditional intrusion prevention systems (IPS) are frequently difficult with the ever-changing nature of cyber threats. To overcome these limitations, we propose the Gannet optimized-mutated k-nearest Neighbour (GO-MKNN) approach as a novel customized intrusion prevention paradigm for self-governing systems. To improve detection accuracy and adaptability, the GO-MKNN algorithm combines the optimization powers of Gannet algorithms with the durability of the MKNN approach. Initially, this study obtained a dataset from CICIDS2017 CAN-intrusion, which was utilized for automobile attacks, and suggested designing additional IDS for the CAN system to train our suggested model. Following dataset collection, data cleaning and normalization were performed. Python was utilized to simulate our proposed method. The suggested method's effectiveness was evaluated in terms of Precision (%), Accuracy (%), F1-score (%) and recall (%). The experimental findings of the research may contribute to the development of a strong framework for intrusion detection that would guarantee the dependability and safety of autonomous vehicles.