XAI in Intrusion Detection Systems for loT: A Survey of Models and Interpretability Approaches

With the explosive development of Internet of Things (IoT) networks, the potential security risks are also greatly magnified, and there is a very strong requirement for high performance IDSs. While ML and deep learning (DL) IDSs have led to high detection accuracy, their black-box nature restricts trust and transparency and it is not practical in real-life safety-critical as well as resource constrained IoT environments. Explainable Artificial Intelligence (XAI) has recently proved to be a feasible candidate to cope with these issues and achieve interpretable and transparent decision-making in IDSs. The contribution in this work is a systematic and structured review of XAI based intrusion detection methods for IoT networks. The paper provides a structured overview on the state-of-the-art of AI-based IDS models, both conceptually and technically: from modelling to explainability solutions, up to deployment architecture at edge, cloud and hybrid scope. Performance dimensions including detection accuracy, interpretability, latency and resource efficiency are compared. The study also identifies important challenges, such as adversarial robustness, scalability and the absence of standard evaluation frameworks. Lastly, this paper points to the emerging research challenges, focusing on light-weight, adaptive and inherently explainable IDS solutions that can come with trustworthy and scalable IoT security deployments.

کلیدواژه ها:

Explainable Artificial Intelligence ، Intrusion Detection Systems ، Internet of Things Security ، Machine Learning ، Deep Learning ، Network Security