Adaptive Aggregation for Distributed Industrial Control Systems Anomaly Detection

Industrial Control Systems (ICS) increasingly face sophisticated cyber attacks that individual facilities cannot detect in isolation. While collaborative anomaly detection could significantly improve threat identification, it requires sharing sensitive operational data between potentially competing organizations, raising privacy concerns. Moreover, ICS operate under strict real-time constraints where traditional privacy-preserving mechanisms introduce unacceptable latency. This paper presents an adaptive aggregation framework that dynamically balances privacy and detection latency based on operational context and threat levels. Our key insight is that privacy-performance trade-offs need not be static; by adapting parameters to current conditions, we achieve both strong privacy during normal operations and rapid detection during critical threats. Our work shows that adaptive mechanisms can transcend traditional privacy-performance trade-offs, enabling secure collaboration among industrial competitors while maintaining operational confidentiality and real-time requirements essential for critical infrastructure protection. The work presented here represents a crucial step toward practical, privacy-preserving collaborative security for critical infrastructure, demonstrating that the traditional trade-off between privacy and performance can be effectively managed through intelligent, context-aware adaptation.

Hali tarjima qilinmagan