In an upcoming paper, Brice Canvel, Alain Hiltgen, Serge Vaudenay, and Martin Vuagnoux describe and demonstrate a timing-based attack on CBC ciphersuites in SSL and TLS. An active attacker may be able to use timing observations to distinguish between two different error cases: cipher padding errors and MAC verification errors. Over multiple connections this can leak sufficient information to make it possible to retrieve the plaintext of a common, fixed block. In order for an attack to be sucessful, an attacker must be able to act as a man-in-the-middle to intercept and modify multiple connections, which all involve a common fixed plaintext block (such as a password), and have good network conditions that allow small changes in timing to be reliably observed. A patch is available from the OpenSSL project
I know you're going to want to backport the fix, but please please please consider biting the bullet and figuring out how to just get to 0.9.6i. There have been ***lots*** of not-as-security-critical bugs fixed between b and i. The one biting me today is that the very cool http://www.washington.edu/pubcookie/ simply will not work on any RedHat system.
Could someone please provide some status update here? Other distributors have released their updated openssl packages weeks ago. Is RHL vulnerable and if so, when is the patch coming? Thanks.
We wanted to make sure that a fix for the critical sendmail vulnerability would be available on Monday so we adjusted the priority of the OpenSSL errata accordingly. We are working on updated OpenSSL packages and will make them available shortly.
Updated OpenSSL packages for Red Hat Linux are now available http://rhn.redhat.com/errata/RHSA-2003-062.html Leaving this bug open until packages for Red Hat Linux Advanced Server are also pushed.
Was actually fixed some time ago, see latest OpenSSL advisory http://rhn.redhat.com/errata/RHSA-2003-102.html