How can I be sure Lastpass really can't access my passwords?

Question

The recent, widely publicized security incident where millions of Linkedin were exposed reminded me to tighten up my password practices. I'm looking at several password managers now and I'm especially curious about Lastpass.

They write on their homepage:

LastPass is an evolved Host Proof hosted solution, which avoids the stated weakness of vulnerability to XSS as long as you're using the add-on. LastPass strongly believes in using local encryption, and locally created one way salted hashes to provide you with the best of both worlds for your sensitive information: Complete security, while still providing online accessibility and syncing capabilities. We've accomplished this by using 256-bit AES implemented in C++ and JavaScript (for the website) and exclusively encrypting and decrypting on your local PC. No one at LastPass can ever access your sensitive data. We've taken every step we can think of to ensure your security and privacy.

How can I be sure that the bolded part is true? Is the method they describe capable of actually doing what they promise, can it prevent them from accessing my passwords? And how could I verify that they're actually doing what they're promising and not transmitting my password in any form they can access to their servers?

Answer 1

There is a way to see if LastPass is doing what they're saying.

Use the Non-binary Chrome, Firefox, Opera, or Safari extension. This is 100% JavaScript and open in the sense that you can see it -- you can use network sniffing with a proxy (e.g. Paros) to see that the sensitive data is encrypted with AES-256-CBC from data generated from a key created with the number of rounds of PBKDF2-SHA256 you have setup on your account: http://helpdesk.lastpass.com/security-options/password-iterations-pbkdf2/ and this is done locally on your machine only.

Then simply don't update/upgrade your extension until you want to audit it again. You could also audit the way we interact with the binary extension to decide if you trust that.

That's a bit extreme for most people, but a number of people and organizations have audited LastPass and liked what they found. LastPass is always helpful to anyone wishing to audit, feel free to contact us if you'd like help.

LastPass knows that it's perfectly reasonable to trust but verify, and encourage you to do so. There's a reason we tell people to utilize the extensions rather than the website: the extensions can't change as easily as the website could thus making them more secure.

Source: I work for LastPass.

Answer 2

Some of these answers are pretty dated, but the subject is important enough that I think it merits revisiting.

LastPass assertion is that they offer a Zero-knowledge proof implementation - i.e. the encryption happens client-side (with the password being the key) and that they, presumably, cannot decrypt the data even if they wanted to. If they're served with a warrant or court order they'll be obliged under law to hand the data over, but it would still be in encrypted form, and then it's up to the respective investigators' supercomputers (or modest GPU array) to crack that. In this regard it's fundamentally no different than storing a KeePass DB in DropBox (which I've seen more times that I care to mention)

That being said....

LP have recently released the source for their CLI client: https://github.com/LastPass/lastpass-cli

It's now up to us to do the peered code-review, so as to validate their claims match up to what's delivered.

Most importantly interrogating the source to see how the DB is generated, encoded & encrypted, if it meets with best standards & practices (or better), weed out bugs (or "unintentional backdoors"), and if the product of the generation matches that generated by the closed black-box implementations - similar thought-process involved in compiling code from source & comparing the checksum against that generated against the binary.

An independent code-review & pen-test from a reputable organization is what's needed IMHO, and so puts it well beyond my own skill-set.

This is not an attempt to steal or otherwise reverse-engineer their UX secret-sauce, where they (rightly) add value & derive revenue from - I'm happy to throw money & clients their way for this, as them making good security simpler makes my live safer & easier - but rather a way for the security community to raise the bar & ensure that those that abide by Kerckhoffs' principle are rewarded for their commitment.

Answer 3

The only way you can verify it is to look at the code they're sending you every single time you access your passwords. In other words, you can't. However, if they ever sent malicious code to anyone, there's a risk that person would notice, and all it would take is one person presenting that malicious code and everyone would know about it.

Answer 4

Even if you trust a company, you can not fully trust the product/service they provide. A vulnerability in it may be discovered/exploited or the company itself may be compromised. Anytime you make your passwords potentially accessible to a party, consider it known by that party. In the case of the normal registration/login process of a website, you are dealing with people who don't actually value your password as they do not need it. As soon as a third party is involved, this assurance is now out the window.

If you must, use a local password safe like KeePass2 and store to removable media that you control.

Also if a company uses a phrase like "complete security", know that they are already trying to trick you.