A Novel Multi-expert System in Conjunction with ML Models for Detecting DOS on Edge Computing

The rise of technologies such as AI and ML has created new advances in cybersecurity, especially in Intrusion Detection Systems (IDS). ML models outperform traditional methods in performance and scale, but their interpretability is a drawback in high-risk domains such as cybersecurity, where model reasoning is crucial. ML models may also inherit biases from training data, resulting in biased outputs emphasizing the need for transparency and justice in AI applications that will become Explainable AI. Expert Systems (ES), a successful early AI technique, capture expert information in a knowledge base for interpretability. Their effectiveness in many fields requires precise understanding, and then they can work well with other ML algorithms to create a new approach. This work introduces a two-layer Multi-Expert System (MES) architecture for edge computing DoS attack detection with an even number of experts, and they use a revolutionary voting process for optimal decision-making and transparent interpretation. The proposed system improves the interpretability of standard ML-based IDS models while preserving accuracy; then, it is validated by UNSW NB15 dataset experiments to show our higher DOS accuracy rate than other studies.