I think for a while leading up to the recent session stealing hack, there has been a massive amount of positivity from Lemmy users around all kinds of new Lemmy apps, frontends, and tools that have been popping up lately.
Positivity is great, but please be aware that basically all of these things work by asking for complete access to your account. When you enter your Lemmy password into any third party tool, they are not just getting access to your session (which is what was stolen from some users during the recent hack), they also get the ability to generate more sessions in the future without your knowledge. This means that even if an admin resets all sessions and kicks all users out, anybody with your password can of course still take over your account!
This isn't to say that any current Lemmy app developers are for sure out to get you, but at this point, it's quite clear that there are malicious folks out there. Creating a Lemmy app seems like a completely easy vector to attack users right now, considering how trusting everybody has been. So please be careful about what code you run on your devices, and who you trust with your credentials!
Lol this is not how the hack worked. JWT cookies are encrypted. They don’t contain your password at all. There was no way to reverse engineer from your cookie to your password.
Correct me if I'm wrong, but Lemmys tokens have no expiration, right? So they are effectively username and password combined.
The decoding algorithm can change, which is exactly what happened, invalidating all previously generated tokens. They cannot be decoded to a password though since they are encrypted, meaning shared passwords wouldn’t be an issue (though you should use a password manager to not have this issue in the first place).
That's not how tokens work. Tokens are (usually short-lived) credentials that are passed to a client after it authenticated e.g. with username + password. But the password is never contained in a token. They are also not encrypted, but signed, meaning the Lemmy instance can verify that the plain-text token ist authentic. Also, not the decoding algorithm has changed, but the signing key was rotated, meaning all JWTs created before the signing key rotation would not look authentic anymore
Right. I was simplifying it by saying it was encrypted, but it’s just an auth token from the service after verifying that the user is authenticated.
And correct. I’m just an idiot and didn’t do research to explain it, even though I’ve used JWT tokens in my applications before. Hahaha. Thanks for the detailed correction!