Hector Martin

It allows to understand how countermeasures are working...

No it doesn't. The programming spec is irrelevant for stuff like silicon voltage monitors and metal layer meshes, which are the kinds of things which actually provide environmental security.

Also, you're using the "security by obscurity" argument, which, even if the security is actually good, has been thoroughly discredited in the security industry, and silicon isn't a special case. I hope I don't need to give you a lecture on this, it's common knowledge.

I'm just saying they are protecting their IPs, I'm not saying it's a good thing...

Fair. I don't think they're actually accomplishing anything but... :-)

What I find not fair in the discussion is to induce the idea that using pwned circuit as they are open, is a better idea (security wise) than secure circuits as they are closed... I think it's a fallacy...

It depends on your threat model. If you are more concerned about firmware flaws and design, it makes more sense to use an open IC with a solid FW design. If you absolutely must resist physical attacks, maybe not.

For example, at least old YubiKeys did not cryptographically wrap the PGP private keys with the user PIN, which is insane. Under some threat models, that makes them less secure than an STM32. And in fact they did have a bug where they weren't checking PINs at all.