#7085: Integrate Cryptocat Browser Extension into Tor Browser Bundle --------------------------------------+------------------------------------- Reporter: kaepora | Owner: erinn Type: enhancement | Status: new Priority: normal | Milestone: TorBrowserBundle 2.2.x-stable Component: Tor bundles/installation | Version: Tor: unspecified Keywords: | Parent: Points: | Actualpoints: --------------------------------------+-------------------------------------
Comment(by mikeperry): Replying to [comment:27 kaepora]: > Replying to [comment:26 mikeperry]: > > Second, I am very concerned that there were XUL XSS bugs in the chat windows. To me, that's a bad sign. Ideally, I'd like to see something on your side (ie a tag in your bugtracker or some other document you wrote) that enumerates the patches that resulted from your first audit. > > The patches in which we fixed the audit bugs are enumerated (perhaps incompletely) in this [https://blog.crypto.cat/2012/11/security-update- our-first-full-audit/ blog post]. This is great. You should show this to whoever you contact about further audit. Bugs tend to come in groups of similar nature, and seeing the types of mistakes you made will help the second auditor find more of them. > > Third, while it does look like the audit was extremely thorough, I think I'd prefer a second one for this reason. XUL XSS is quite serious, and since you're writing a network-facing app with lots of user and network provided content, its critical that your code receives lots of this type of review. > > Very well. Who would you recommend to perform the second audit? If you can give me a preferred auditor or a list of auditors that the Tor Project would feel comfortable with, I have no problem getting in touch with them. A quick search for "Firefox extension XSS" and related queries turns up Roee Hay and Roberto Liverani as previous folks who have done similar audits. I don't know these people, but I suppose an email couldn't hurt. You should also send Dan Veditz an email (dveditz at mozilla). He is a very Tor-friendly Mozilla security engineer. He might be too busy to give you a full review, but I bet if you also mention that integration with Mozilla's Social API is something you're considering, he might be able to justify devoting his time for such a review. Or at least point you in the right direction for getting a better understanding of the security issues the Social API team has had to deal with. I bet they are similar in nature. I can also do a review later, but for the next month or two I will be extremely busy trying to get TBB working with Firefox 17ESR. I also think the people I mentioned might actually be better at it than me (except perhaps for spotting potential proxy bypass issues). My approach with Torbutton has mostly been "Zomg, don't ever interact with or display content window material in XUL." Sadly, even that failed me when implementing the content window JS hook injection. Luckily, Dan Veditz came to the rescue in that case :). > > Fourth, I guess I am mildly concerned about the crypto security. I don't believe it's impossible to do crypto with JS, but I would prefer it if the underlying primitive implementations also had a chance for review, especially since our inclusion of this addon would probably be seen as endorsement of its crypto and security by many. > > Our OTR implementation has been reviewed. If there is a specific type of further review you would wish to ask for, we can see it done. I think Nick is the person to best answer this. Or Ian himself. I also thought Moxie's observations on the website version were spot-on. Has he looked into the browser version? Things that come to my mind include: Is your implementation fully compatible with other OTR and MPOTR implementations? Verifying this to be the case can be one way to probe for implementation errors. One question academic researchers in need of racking up publication count will drool over is "Can content window JS extract side channel info about your crypto operations?" If the JS scheduling mechanism is predictable/discoverable and/or runs in the same thread, the answer here might actually be "Yes. Very Yes.". Unfortunately, the publication bias in academia may very well cause them to structure their experiments such that the answer is "Yes" anyway regardless of threading model or content window JS timer resolution... :/ -- Ticket URL: <https://trac.torproject.org/projects/tor/ticket/7085#comment:29> Tor Bug Tracker & Wiki <https://trac.torproject.org/> The Tor Project: anonymity online _______________________________________________ tor-bugs mailing list tor-bugs@lists.torproject.org https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-bugs