INTERPRETABLE AI FOR CYBER THREAT DETECTION IN SMART SYSTEMS

Abstract

The rapid development of intelligent technologies, IoT infrastructures, and cloud services has led to an increasingly complex cybersecurity landscape. Because intelligent systems generate massive volumes of heterogeneous streaming data in real time, the challenges of accurately detecting threats in the cyber environment have increased significantly. Traditional intrusion detection systems used to identify cyber threats typically rely on static attack signatures and, therefore, have limited ability to detect new forms of cyber threats. This study presents an interpretable artificial intelligence framework for real-time cyber threat detection in streaming intelligent systems using machine learning models, combined with explainable AI methods such as SHAP and LIME to enhance detection capabilities. Security-related data streams and other sources were analyzed to identify potential anomalies and cyberattacks. Based on experimental evaluation, the proposed model, based on combined methodologies, such as hybrid explainable AI model, demonstrated superior performance compared to each of the individual machine learning models across several evaluation metrics, including accuracy, precision, recall, and F1 score. Overall, this work demonstrates how to leverage machine learning and explainable AI methods to improve trust, transparency, and the practical applicability of cybersecurity monitoring solutions in dynamic, intelligent environments.