Active Bayesian Inference for Robust Control under Sensor False Data Injection Attacks
For cyber-physical systems, this work addresses the critical problem of maintaining reliable state estimation under sensor attacks, offering a practical method that outperforms existing approaches.
The paper presents a framework for detecting and recovering from sensor false data injection attacks in cyber-physical systems, using Bayesian inference and active probing. Experiments on an inverted pendulum show significant performance improvements over baselines, especially under prolonged attacks.
We present a framework for bridging the gap between sensor attack detection and recovery in cyber-physical systems. The proposed framework models modern-day, complex perception pipelines as bipartite graphs, which combined with anomaly detector alerts defines a Bayesian network for inferring compromised sensors. An active probing strategy exploits system nonlinearities to maximize distinguishability between attack hypotheses, while compromised sensors are selectively disabled to maintain reliable state estimation. We propose a threshold-based probing strategy and show its effectiveness via a simplified partially observable Markov decision process (POMDP) formulation. Experiments on an inverted pendulum under single and multi-sensor attacks show that our method significantly outperforms outlier-robust and prediction-based baselines, especially under prolonged attacks.