SERENA-IIoT: SEcure and REliable Networked Architecture for Industrial Internet of Things digital transformation

The SERENA-IIoT (SEcure and REliable Networked Architecture for Industrial Internet of Things digital transformation) project aimed at enhancing knowledge in the next-generation IIoT systems by introducing novel solutions concerning reliability and security. Crucial included: i) the collection, analysis, and protection of data from real datasets, facing heterogeneous and unstructured data formats; ii) near-real-time anomaly and attack detection to enable timely recognition and response. From an architectural viewpoint, SERENA-IIoT addressed the need for distributed and multi-layered IIoT architecture supporting data analytics and AI-based techniques towards the edge to efficiently collect and analyze the vast amount of generated data. Key outcomes include the RailRED framework for modeling complex Train Control Management Systems (TCMS), and the implementation of Fixed-Priority Preemptive Orchestration for serverless services. The use of virtualized architecture to protect and isolate on-device AI tasks, ensuring secure cross-layer interoperability, has been explored. Such innovations ensure that time-critical AI tasks are efficiently managed across the edge-to-cloud continuum, even under strict resource constraints. Concerning reliability, SERENA-IIoT aimed at timely identifying errors/failures, leveraging lightweight AI-based monitoring strategies. A core achievement was the development of the Open-FARI testbed, enabling federated anomaly detection, allowing for decentralized and collaborative monitoring in railway environments. A transformer-based log-driven testing was also defined, enabling the characterization of system behavior and the identification of anomalies from raw event logs through efficient deep-learning-based strategies for online learning. Finally, regarding security, SERENA-IIoT provided techniques for the supervision of activities in an IIoT system, managing access to resources, enforcing security policies, detecting attacks, and counteracting them. This was reinforced by the proposed SereWay and SereBench frameworks, which established benchmarking suites for the assessment of adversarial robustness and event classification accuracy in networked railway cyber-physical systems. The solutions developed in SERENA-IIoT have been prototyped to test their effectiveness through simulation-based experimentation, using real data traces.