I’m planning to set up LUKS on an SSD. Many guides are suggesting using a simple key to set things up and then revoke it when everything is in place.
Given the wear leveling behavior on SSDs I am assuming a simple key might be able to unlock even beyond the revocation if a determined attacker has the disk. I don’t want someone to be able to put the disk in factory access mode and be able to brute force attempt their way to browser cookies and email accounts.
I’m going to ignore the suggestion about using a weak key to set up, but am I being overly paranoid? Am I being not paranoid enough and I should also not rely on revocation for a spinning rust disk?
I’d be very suspicious at what else this guide has for you if it advises to use “key” passphrase for a LUKS key.
With that said, most SSDs do encrypt data before it reaches the NAND. I don’t know what “factory access mode” is and whether that can get around it.
I think yeah, I will not be following that advice for sure. Just wondering at this point if someone should take extra precautions around SSD encryption. Like should one overwrite the whole drive if a key is leaked so that the odds of recovering any info from the chips is lessened? Or is revoking the leaked key sufficient?
It all depends on the threat model. If I were protecting against accidental data disclosure after decommissioning an SSD, then revoking would suffice. However if I were a journalist gathering data on some unsavory state subjects, I’d probably only ever use high-entropy keys.
Use cryptsetup and it should handle key creation for you. I’ve never heard this but about key revision. How are you supposed to use the disk if the key is revoked?
Hdd’s have bad block remapping sort of like ssd’s, so the same issues apply to both types of media.
Yes. Some guides suggest, say, “just use ‘key’ for now, we’ll replace it later.” I didn’t mention their step adding a stronger key, I guess I didn’t see that as an important part of the question.
Ive always used cryptsetup and never seen any intructions like how you are describing. I wonder if you have a different use case than I do? It seems like adding more complexity to start with one decryption method only to change it soon. Why?
