This sounds like the beginnings of the silicon obfuscation strategy: hiding things in silicon that lie below these assumptions
It's possible you could fool the algorithm in some way, but you'd have pretty weird and severe constraints on the shapes of your hidden parts to hide their presence from low-angle projections. They're probably not actually hand-crafting geometric assumptions about chip structures or anything like that.
> Five individual logic cells were functionally disconnected from the rest— with no pathways that would allow them to influence the output— yet when the researcher disabled any one of them the chip lost its ability to discriminate the tones. Furthermore, the final program did not work reliably when it was loaded onto other FPGAs of the same type.
The article says they only image the metal layers, so I guess you can do anything as long as you don't visibly change the metal layers.