Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
This paper introduces an active Bayesian inference framework to detect and mitigate sensor false data injection attacks in cyber-physical systems. The framework models perception pipelines as bipartite graphs within a Bayesian network to infer compromised sensors based on anomaly detector alerts. An active probing strategy, using system nonlinearities, maximizes the distinguishability between attack hypotheses, while selectively disabling compromised sensors. Experiments on an inverted pendulum demonstrate that the proposed method outperforms outlier-robust and prediction-based baselines, particularly under prolonged attacks.
Inverted pendulums can withstand prolonged sensor attacks if you actively probe their weaknesses and selectively disable suspect sensors.
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.