Good point. Even though I have hardware knowledge and studied VLSI I'd be completely lost with the raw data. How I imagine this working - Huang and Applebaum b

Good point. Even though I have hardware knowledge and studied VLSI I'd be completely lost with the raw data. How I imagine this working - Huang and Applebaum both suggested similar - is that someone would publish "ground truth" templates of what hardware should look like. The validation would only flag suspicious differences (using "AI"/pattern-matching etc) and then those would definitely be escalated to an expert.