Reference-Free Spectral Analysis of EM Side-Channels for Always-on Hardware Trojan Detection
This work proposes the first reference-free detection method for always-on hardware Trojans (HTs), addressing the challenge of identifying such malicious circuits in the absence of a golden reference model. The approach leverages multi-window short-time Fourier transform (STFT) to extract time-frequency features from electromagnetic side-channel emissions and employs Gaussian mixture models (GMMs) to capture the statistical structure of circuit behavior. Detection is achieved by exploiting the distinct statistical stability patterns in the time-frequency spectra between Trojan-free and HT-infected circuits. Experimental validation on an AES-128 implementation demonstrates that the method accurately distinguishes circuits with always-on HTs from benign ones, significantly enhancing hardware security assurance in reference-free scenarios.