EMAth: ECU Authentication via Electromagnetic Side Channels

We propose “EMAth: ECU Authentication via Electromagnetic Side Channels”, a novel approach to authenticate and monitor Engine/Electronic Control Units (ECUs) of (autonomous) vehicles to enhance their security and identify units with counterfeit components by analyzing the unintended electromagnetic (EM) signals. EM side channels are byproducts of legitimate program executions due to transistor switching. Because these switching activities are correlated with the program execution, emanated EM signals carry information about the state and/or authenticity of subject devices/programs. The proposed framework will utilize EM side-channels by generating verification methods, algorithms and machine learning-aided program models, and develop prototypes to prevent attackers from exploitation of ECU units. Based on the analysis, the prototype will authenticate the ECU units for both hardware and software at the acceptance stage and alert system users while in operation for the adversarial activities. Utilizing purpose-designed antennas, EMAth will collect emanated EM signals and reveal information embedded into them, i.e., spectral features, signal pattern, active frequency distribution, etc., regarding the state of ECUs after processing. EMAth will use golden samples, which are guaranteed not to be tampered with and working as expected, to form various hypotheses about the signal models of ECU operations. Having multiple hypotheses helps to cope with presence of interrupts, input/output activity, cache misses, branch miss-predictions, other events, etc., and will allow EMAth to maintain high accuracy and fidelity while authenticating ECUs of interest. These events significantly change the emanated EM signal pattern in a way that is seemingly random and can lead inaccurate outcomes. However, EMAth will be designed to account for these deviations and even exploit them for better monitoring. The basic concepts and initial feasibility of EMAth for IoT-class devices was demonstrated by the founders of Aether Argus Inc. in the DARPA LADS program. We will build upon that work to demonstrate the ability to authenticate and monitor ECUs of interest.