Despite 2013's high-profile attacks on the RC4 algorithm in TLS, its usage is today (March 2015) still running at about 30% of all TLS traffic. This is attributable to the lack of practicality of the existing attacks, the desire to support legacy implementations, and resistance to change.
We provide new attacks against RC4 in TLS that are focussed on recovering user passwords, still the pre-eminent means of user authentication on the Web today.
Our attacks enhance the statistical techniques used in the previous attacks and exploit specific features of the password setting to produce attacks that are much closer to being practical. We report on extensive simulations that illustrate this. We obtain good success rates with 226 encryptions of the password. By contrast, the previous generation of attacks required around 234 encryptions to recover an HTTP session cookie.
We also report on two "proof of concept" implementations of the attacks for specific application layer protocols, namely BasicAuth and IMAP.
Our work validates the truism that attacks only get better with time, and makes the continued use of RC4 in TLS increasingly indefensible.
Download the technical paper.
Our research is supported by EPSRC, Google, Mozilla Foundation and White Ops.