Behavior-Based Interpretable Trust Management for IoT Systems

Establishing appropriate trust at the “right” level for the “right” application at the “right” time is important in the constantly evolving environment of the Internet of Things (IoT). Nevertheless, the process is challenging due to the lack of explainability and interpretability of the machine learning models. This paper presents a novel approach to managing IoT trust by employing explainable artificial intelligence (XAI) to connect complex algorithmic decisions with human understanding. Specifically, we propose a mutual information selection technique to determine the most significant behaviour-based features to identify trustworthy and untrustworthy behaviour in IoT device systems. Based on these behaviour-based features we develop a rule-based decision tree (DT) method to help enhance the explainability of our model. We evaluate our approach with a transformed UNSW NB 15 dataset, the results demonstrate improved user trust and system transparency. In addition, we did a comparative analysis of our XAI-behavioural-based trust management system (BB-TMS) with state-of-the-art methods, which demonstrated that our model surpasses competitors in terms of precision and interpretability. This highlights the effectiveness of integrating XAI with traditional machine learning approaches in the IoT domain.