Reference-Free Spectral Method Detects Always-On Hardware Trojans via EM Side-Channel Analysis

Global: Reference-Free Spectral Method Detects Always-On Hardware Trojans via EM Side-Channel Analysis

Researchers Mahsa Tahghigh and Hassan Salmani introduced a reference‑free technique for identifying always‑on hardware Trojans using electromagnetic (EM) side‑channel data, as detailed in a paper submitted to arXiv on 28 January 2026.

Background and Challenge

Always‑on hardware Trojans represent a significant threat to the integrity of trusted microelectronics, yet conventional side‑channel detection approaches typically depend on a golden reference model of a Trojan‑free circuit, which is often unavailable in real‑world supply chains.

Methodology

The authors combined time‑frequency analysis via Short‑Time Fourier Transform (STFT) across multiple window sizes with Gaussian Mixture Models (GMMs) to characterize EM emissions without requiring a reference baseline.

Key Observations

Experimental results indicated that circuits without Trojans exhibit fluctuating statistical structures in the spectral domain, whereas always‑on Trojans produce persistent footprints manifested as fewer, more consistent mixture components.

Validation on AES‑128

The approach was evaluated on an AES‑128 implementation, demonstrating that the reference‑free model could successfully distinguish Trojan‑free from Trojan‑infected hardware, confirming feasibility without golden reference data.

Implications

By eliminating the need for a pre‑characterized reference, the technique could enhance security assessments in environments where hardware provenance is uncertain, potentially strengthening supply‑chain assurance.

Future Directions

The authors note that further testing on diverse architectures and under varied environmental noise conditions is required to generalize the method’s applicability.

Publication Details

The work is cataloged under arXiv identifier 2601.20163, classified in the Cryptography and Security (cs.CR) subject area.

This report is based on information from arXiv, licensed under Academic Preprint / Open Access. Based on the abstract of the research paper. Full text available via ArXiv.