Control-Implications-Based Side-Channel Monitoring for Embedded Systems

Monitoring of embedded-system anomalies which influence the regulation of physical-world processes is examined. Specifically, detection of such anomalies using remote measurements of the controlled process itself - which we refer to as a controlled-process side channel - is studied. A physics-guided anomaly detection algorithm is proposed, which decomposes the measurement signal into patterned and ambient responses, and exploits sparsity in both components. Detailed simulations of two controlled-process side channels on an autonomous underwater vehicle, namely a motion guidance system and a sonar transmitter, are undertaken to assess the methodology. Our preliminary findings suggest that controlled process side channels may be sufficient for monitoring anomalies, when traditional side channel signals are difficult to measure.