Red Team Vs. Blue Team
summary
A public case study compares four real chips (90, 65, 40, 28 nm) by imaging backside layers with a scanning electron microscope (≈300 GB uncompressed per die) and running image-based matching against the GDS design files; the red team injects four additions and six replacements per chip in the design files, the blue team’s unoptimized algorithms run in hours while imaging takes 1–2 days. All added cells are found; replacements are detectable for 90/65/40 nm with a few hundred manual false positives, while 3 replacements in 28 nm remain undetected unless >70k false positives are tolerated. Data and code live in a GitHub repository.
AI-written summary · about
chapters (8)
- 0:00Introduction to talk
- 1:06Background on Hardware Trojans
- 5:39Chip manipulation by Red Team
- 7:25Blue Team imaging process
- 13:52Live detection demonstration
- 16:44Detection results and analysis
- 20:34Conclusion and future work
- 21:37Q&A and survey
related talks
29mRed Team, Blue Team Or White Cell? Trends In IT
42mRed vs. Bank: Surviving a Three Year Red Team
45mIntegrated Circuit Offensive Security
59mRed Teaming Large Language Models
22mRADOPA: Robustifying a Design Against Optical Probing Attacks
21mOpen Source IC Design and Hardware Reverse Engineering Or: How I Learned to Stop...
40mFantastic Red-Team Attacks and How to Find Them
23mExtraction of Secrets from 40nm CMOS Gate Dielectric Breakdown Antifuses by FIB...