Distributed verification of security events in IIoT networks using low-latency blockchain consensus assisted by artificial intelligence

This is the authors' abstract. We don't add key points for this paper.

Security in Industrial Internet of Things (IIoT) networks faces structural limitations when threats are generated concurrently by distributed, heterogeneous nodes. Current systems, whether local or centralized, even when employing advanced artificial intelligence models, produce alerts whose overall validity cannot be verified. At the same time, existing blockchain solutions are primarily used as passive logging mechanisms, introducing latencies incompatible with industrial requirements. This work addresses this gap through a distributed security event verification architecture that integrates AI-based detection with a blockchain consensus mechanism explicitly optimized for low latency. The AI models generate events enriched with a continuous suspicion score, which are evaluated by a set of independent validators and resolved using Byzantine-fault-tolerant, permissioned consensus. The experimental evaluation is performed using widely adopted open IIoT datasets, supplemented by a dataset designed for model validation and consensus processes under controlled load and disagreement conditions. The results show that consensus maintains average latencies below 100 ms and within the 95th percentile range under operational loads, with validation rates exceeding 700 events per second before saturation. Additionally, the event acceptance rate remains stable despite increased workloads, demonstrating that consensus does not amplify detection errors.