Explainable AI based Improved Intrusion Detection System in Wireless Sensor Networks

To protect computer networks from unauthorized access by users, including insiders, the intrusion detection system (IDS) is used. It monitors network traffic to detect malicious activity and sends alerts immediately when observed. The various challenges posed by the wireless sensor network (WSN) environment require the development of machine learning-based IDS. Adaptive WSN data analysis and anomaly detection are possible with Machine Learning (ML) based IDS. This paper aims to improve the security and reliability of WSN using ML. WSN will generate high-dimensional data due to the increase in user base and network size. In this project, two datasets; WSN DS and KDD CUP 99 are used to detect and classify attacks using different ML algorithms. Furthermore, to improve the interpretability of attack classification, the proposed method also tests explainable AI (XAI) algorithms; LIME (local interpretation that can be interpreted independently of the model) and SHAP (SHapley supplementary interpretation).