Hacker News Reader

Comment by taeric | Hacker News Reader

The asymmetric point is surprisingly useful.

It makes me slightly weary, as those sorts of things can have unintended consequences. For example, there was a soc

al engineering exploit where someone would call into Amazon and add a credit card to an account they didn't own. They would then call right back and request a new email added to the account and use the freshly added credit card as authentication.

They could then use your Amazon account, or use the real card associated with the account as authentication for something else

Not clear how that has any relevance here. If the encryption is broken, pretty sure most encryption is busted.

Encryption is broken, just not in all scenarios, that's why security without threat model is meaningless.

My point was that if this model is broken, I'm not sure I would trust any other for storing data at rest, either.

Specifically, why would you still trust any other password manager?

Password managers that use authenticated encryption are not vulnerable to such attacks. Well, it's more a hypothetical attack, in practice you're more likely to get a keylogger.

I'm not sure what you mean by authenticated encryption. I'm assuming you just meant single key encryption.

My question is if those are really on a different set of math, as my understanding was that they were not, all told. If you can bust public/private key encryption, you can typically bust all encryption. Is that not necessarily the case?

Pass uses public key to encrypt files, an attacker needs to know only the public key to forge pass files, and that public key isn't secret, it's stored in plain, unencrypted, that's why you can create pass files without entering master password that protects private key which is not used to create pass files. That's the catch with asymmetric encryption.

But what is the threat vector? I have to pull the attacker's changes into repo. And... I would still have the old passwords.

How does this help an attacker get my passwords?

It's not a certainty of abuse, only uneasiness about technical feasibility. Historically abuse of forgery was clever and unpredictable, as a result design of cryptographic systems tries to prevent forgery when possible. A hypothesis: pass prints text from decrypted file to terminal, terminals have rich functionality, legacy features and wide attack surface, so text printed to terminal is an attack vector.

And convenient. The only minor road block I sometimes encounter is when a website has dumb rules for password. I have to generate a auto one. Then edit it manually to make it comply to « no @ in password » or whatever dumb stuff like that

Just fyi, `pass generate` has a --no-symbols option that should save you some time in the future.

How could have miss that? It’s right there in the doc.

Thank you.

Reply on news.ycombinator.com