The fast growth of Internet of Things (IoT) systems has made them very susceptible to advanced cyber-attacks, and an intelligent and privacy-sustainable intrusion detection system is required. Conventional centralized intrusion detection frameworks have the drawbacks of data privacy threats, scale constraints, and a single point of failure, and conventional federated learning still faces the threat of malicious client membership and a lack of trust during model aggregation. To overcome these obstacles, this paper suggests a federated learning (B-FL) system based on a Blockchain to ensure safe and reliable intrusion detection in the distributed Internet of Things. The framework proposed is a combination of federated and blockchain-based trust management to guarantee decentralized collaborative model training and maintain data confidentiality. The use of smart contract-based verification tools and trust-weighted aggregation counteracts the adversarial threats, such as model poisoning, data manipulation, free-rider behavior, and Sybil attacks. Testing is performed on the CICIoT2023 dataset, which consists of traffic produced by 105 IoT devices in 33 different types of attacks, and it allows testing all the aspects of its work in terms of a real and heterogeneous network. Findings reveal that the proposed B-FL model has high detection rates, high convergence stability, and enhanced robustness as compared to traditional methods of centralized and federated intrusion detection. Another study, Receiver Operating Characteristic (ROC) analysis, supports the presence of excellent discriminative ability with respect to several classes of intrusion. Though the integration of the blockchain has a marginal increase in computing overhead, it benefits the system in terms of transparency, reliability, and aggregation security significantly. In general, the suggested framework offers a scalable, privacy-aware, and trust-conscious IoT intrusion detection system in the next generation to enable secure collaborative intelligence in dynamic and adversarial IoT environments such as mining and mineral-processing environments.
Kamran et al. (Thu,) studied this question.