alt Hacker News>The correct solution is to use free software cryptography on top of these things.
If implemented correctly, "hardware" encryption can be more secure than "free software cryptography". The most obvious benefit is protection against brute-attacks, which you can't meaningfully implement with software. Not to mention added protection of biometric authentication as a second factor, compared to typing a password into a possibly untrusted machine.
Replies
you're not gonna brute force an AES256 key
if you have a bad password that's a user issue
> If implemented correctly, "hardware" encryption can be more secure than "free software cryptography".
Well you can use open-source / free software cryptography to encrypt and decrypt using a secret key which you do not need to remember and which is safely stored behind a secure element.
And there are quite a few people doing just that.
It's immensely more secure than trusting Lenovo or Microsoft or whatever company.
And you can have your secret on the secure element protected by biometric auth too (some do offer just that).
And there are devices like that with a tiny attack surface that accept biometric (like your fingerprint) to unlock the secret.
To get the best of both world you trust the HSM with one thing: the secret, and that's it. And you use open-source software and protocols (like GPG and OpenSSH) with that secret.
“Correctly” is a big caveat. Almost no HW implementation is auditable in any way.
My baseline assumption is that even the good commercial products are only secure from a non-determined attacker, not TLAs. Anything safe from TLAs is what’s being sold only to them.