Comment by bradleyjg | Hacker News Reader

Comment by bradleyjg | Hacker News Reader

Parent Full thread

From the "For our technical readers" paragraph, it's unclear what cryptography model you are going for. On the one hand you say that the private key (albeit password protected) is sent to the server after it is generated, but on the other you say that only public key and encrypted data are sent to the server. Is the model fundamentally the same as Lavabit's, which was so devastatingly critiqued a few weeks ago by Moxie Marlinspike? [1]

Separately, how do you deal with the fundamental weakness of browser based security?[2] You say that the software will be open source, but will there be any mechanism to verify that the served javascript libraries on any particular visit match the sources on github? Will there be any minimization / compilation that would make such verification difficult or impossible to accomplish?

[1] http://www.thoughtcrime.org/blog/lavabit-critique/

[2] see e.g. http://www.matasano.com/articles/javascript-cryptography/

Lavabit's model involved sending/receiving cleartext data to/from the server, and you had to trust that the server-side would not store that plaintext. In our model no unencrypted data, except for the public key, is ever sent to the server. All encryption is done inside your browser.

For your second question, all resources for the site, including javascript, are served over SSL. This will prevent about 99% of MITM tampering attacks. While we can't overcome the inherent weakness in SSL, or guarantee against compromise on the server-side, we are working on a browser extension which will check the code you receive from the server against what is in the repository, so you can at least be alerted of any inconsistencies.

And since the website is all open-source, we're going to make it simple to host it yourself, if you want to be extra-sure that it's not compromised.

EDIT: By "website" I'm referring to the frontend, in-browser code

> When you go to Cryptic all files that you receive will be received over SSL. This is to ensure that you’re getting the correct code, and not a version compromised by some attacker.

This is the Lavabit way, I believe. However, it could be solved using a browser extension.

Cryptic will have a browser extension that will automatically verify the received code against a signed hash (also easily verifiable) on the open source repo. It is fundamentally different than Lavabit since everything that interacts with raw data is open source and verifiable. Someone is going to notice if the website is tampered with.

Reply on news.ycombinator.com