Bank of Models: Sensor Attack Detection and Isolation in Industrial Control Systems

Attacks on sensor measurements can take the system to an unwanted state. The disadvantage of using a system model-based approach for attack detection is that it could not isolate which sensor was under attack. For example, if one of two sensors that are physically coupled is under attack, the attack would reflect in both. In this work, we propose an attack detection and isolation technique using a multi-model framework named Bank of Models (BoM) in which the same process will be represented by multiple system models. This technique can achieve higher accuracy for attack detection with low false alarm rates. We make extensive empirical performance evaluation on a realistic ICS testbed to demonstrate the viability of this technique.