Securing Industrial IoT with Cross-System Federated Learning for Malware Detection

The rapid advancement of Industrial Internet of Things (IIoT) technologies has exacerbated the threat of crosssystem malware in industrial systems. Conventional malware detection approaches, typically constrained to single-system environments, struggle to cope with malware mutations and crosssystem propagation. We propose a federated learning-driven cross-system malware detection framework FedCrossMal. This framework partitions heterogeneous software from different systems into shared and system-specific features, enabling collaborative learning of malware behaviors across systems. FedCrossMal empirical evaluations on real-world software samples spanning multiple systems demonstrate that the proposed framework significantly enhances cross-system malware detection while preserving the original system's high detection performance. This work provides a robust solution for mitigating cross-system malware threats in IIoT environments and offers foundational insights for future research.