Software Suite for "Hardware Trojans from Invisible Inversions: On the Trojanizability of Standard Cell Libraries"

This artifact accompanies our paper "Hardware Trojans from Invisible Inversions: On the Trojanizability of Standard Cell Libraries", published at IEEE S&P 2026. It contains the implementation of our via-position-based similarity metric and Trojan detection pipeline, along with preprocessed data enabling reproduction of the paper's main claims.

The artifact operates on the publicly available backside SEM image dataset of Puschner et al. (S&P 2023), covering four CMOS technology nodes (90 nm, 65 nm, 40 nm, and 28 nm). It includes methods for via extraction (including persistence-based detection), construction of cell-type representatives, pairwise similarity scoring of functionally distinct cell types, and Trojan detection. Preprocessed intermediate results are provided to facilitate artifact evaluation without requiring the full multi-day preprocessing pipeline from scratch; a proof-of-concept of the preprocessing is included in the Jupyter notebook.

The three main claims supported by the artifact are: (1) cells in the 28 nm technology node are substantially more similar than in 40 nm, 65 nm, and 90 nm technologies, (2) our via-position metric strongly outperforms baseline detection methods on the most similar cell pairs, and (3) our method detects all Trojans from the original experiment of Puschner et al., including in the 28 nm node where prior work reported false negatives.

Software, instructions, and further details can be found in the GitHub repository at the following URL: https://github.com/emsec/DAFT