Blockchain-Powered intrusion detection: a decentralized approach to cyber threat defense

Traditional Intrusion Detection Systems (IDS) have the disadvantage of being centrally controlled, having high false alarm rates, and being inflexible to changes in cyber threats. To solve these problems, this paper presents SLBAQ-Sec, a blockchain-based and AI-assisted intrusion detection framework that combines SHAP-LIME-based recursive feature optimization (SLRFO), an attentive spatio-temporal intrusion detection network (AST-IDNet), and adaptive deep Q-learning-based smart contract security (ADQ-Sec). Distributed IDS nodes operate within a network and share validated intrusion alerts via a permissioned blockchain, ensuring tamper-proof logging and eliminating single points of failure. AST-IDNet is a hybrid CNN, BiLSTM, and attention-based model that learns spatial and temporal attack patterns, and ADQ-Sec enables adaptive threat mitigation. Experiments on the CIC-IDS-2017 dataset show the proposed framework achieves 0.95 accuracy, 0.94 precision, and 0.93 recall, outperforming CNN, RNN, BiLSTM, and transformer-based IDS models. These results highlight the effectiveness of combining blockchain and deep learning for real-time cyber defense.