Crypto attack against SSL outlined

Risk for email password exchange using OpenSSL, at least

Swiss security researchers have discovered an attack against implementations of the ubiquitous SSL protocol that could potentially compromise email passwords, though not ecommerce transactions.

The protocol itself has not been compromised and the weakness only applies to versions of OpenSSL prior to version 0.9.6i and 0.9.7a, according to early analysis.

Users of earlier versions of OpenSSL are strongly advised to upgrade.

At this point its unclear whether alternative implementations of SSL are at risk.

Credit card transaction secured using even earlier versions of OpenSSL are not at risk because of the mechanism of the attack.

Secure Sockets Layer (SSL), which is supported by all major Web browsers, is one of the most common security protocols in use on the Net. SSL, and its successor Transport Layer Security (TLS), manage the security of message transmission: which can be anything from the details of credit cards made during a ecommerce transactions to an Outlook client logging onto an email server.

In a paper researchers at the Security and Cryptography Laboratory of Swiss University (Lasec) EPFL demonstrate a timing-based attack on CBC cipher suites in SSL and TLS.

The attack assumes that multiple SSL or TLS connections involve a common fixed plaintext block, such as a password. Since credit cards numbers are normally sent to a secure server only once this particular attack has little or no chance of success.

When checking emails, using for example an Outlook Express 6.x client, using a secure connection passwords are sent periodically as email is checked. This leaves the door open for an attack.

The researchers at Lasec have demonstrated a form a man in the middle attack (using DNS spoofing) can be used to discover email passwords.

Essentially an attacker would substitute specifically made-up cipher text blocks in a legitimate communication and monitor the error messages an email server generates.

In this way, through cryptanalysis of the error messages, it is possible to glean clues on the make up of a legitimate password. Dictionary or brute force attacks may be used, as explained in greater detail in the researcher's paper.

The flaw with earlier versions of OpenSSL lies in the way error messages are constructed, a problem that doesn't apply to OpenSSL versions 0.9.6i and 0.9.7a. This point is explained by the OpenSSL project in much greater depth in an advisory published earlier this week. ®

External Links

Password Interception in a SSL/TLS Channel, paper by security researchers at Lasec
OpenSSL Security Advisory [19 February]: Timing-based attacks on SSL/TLS with CBC encryption

Related Stories

Admins slow to tackle SSL security risks
Slapper worm spanks Apache servers
Sun Crypto curves into open source project

Sponsored: Designing and building an open ITOA architecture