<<< Date Index >>>     <<< Thread Index >>>

Re: [WEB SECURITY] Universal XSS with PDF files: highly dangerous



RSnake wrote:
The point is - someone with shared IP is vulnerable ONLY to an attacker with the same IP. Which makes attacks much less generic and much more painful. Rock solid it ain't, but I think it's a pretty good band-aid until all (hmmm...) clients upgrade to Acrobat Reader 8.0.

-Amit

Sorry for responding late, I've been doing some consulting work.

After talking with some people on my blog I don't believe that is the
case (at least not in theory).  Let's say Alice has an account with
Bob's website.  Cathy is an attacker who owns a website that uses
anti-DNS pinning.
Of course anti-DNS pinning would work against my algorithm, but anti-DNS pinning is a larger problem, one that is out of scope here. I mean - so many things are broken when anti-DNS pinning is introduced... especially any IP-based security techniques. Anti-DNS pinning should be solved by browser vendors (if possible), regardless of the PDF problem. And at any rate, I feel that my algorithm makes the attack harder because it forces it to involve anti-DNS pinning.

Anyway, if you worry about the current anti-DNS pinning techniques, you may simply serve your PDF files in HTTPS only. I believe this will defeat the present day anti-DNS pinning techniques (in the sense that the user under anti-DNS pinning attack will get a certificate error before being served the PDF).

-Amit