Crypto attack puts digital sig hash on collision course

Cryptographers have found new chinks in a widely-used digital-signature algorithm that have serious consequences for applications that sign email, validate websites, and carry out dozens of other online authentication functions.

The researchers, from Macquarie University in Sydney, Australia, found a way to break the SHA-1 algorithm in significantly fewer tries than previously required. Although the hash function was previously believed to withstand attempts numbering 2⁶³, the researchers have been able to whittle that down to 2⁵², a number that puts practical attacks well within grasp of well-funded organizations.

Secure hashing algorithms are designed to reduce text or digital files to a unique series of letters and numbers that is often compared to the document's signature. The findings, which were published Wednesday here (PDF), mean it's easier to create what cryptographers call collisions in SHA-1, in which two different sources share the same the same output.

"I'm expecting that we'll start seeing SHA-1 collisions before the end of the year, if not sooner," said Paul Kocher, president and chief scientist at Cryptography Research, a San Francisco-based consultancy. "For applications that fail because of collisions, you need to be really worried."

Indeed, it was collisions in the now-disgraced MD5 algorithm that allowed independent researchers last year to poke holes in the internet's digital certificate system by forging counterfeit credentials needed to impersonate websites that rely on the security measure. The discovery sent certificate authorities such as VeriSign's RapidSSL scrambling to change the way they issue SSL, or secure sockets layer, certificates.

Other hashes, including SHA-0 and MD4, have also been demonstrated to be weak enough to generate collisions using relatively modest means. With the latest findings, that list is steadily growing.

"Now, SHA-1 has some of the same problems that MD5 has had creeping up on it for a while," Kocher said.

The 160-bit output of SHA-1 should in theory allow hashes to withstand 2⁸⁰ operations without generating a collision. That number came down to 2⁶⁹ and later to 2⁶³ as researchers discovered ways to exploit weaknesses in the algorithm. But until now, only the most elite government organizations had the computing power to do that.

Cryptographers have long recognized that advances in computing power and research into weaknesses of various hashing algorithms jeopardize the reliability of digital signatures. Last year, the National Institute of Standards and Technology commenced an open competition to create a new algorithm. The winning entry should be one that is able to withstand attack and able to generate hashes quickly and without consuming too many computer resources. It will offer hashes with as many as 512 bits, enough to (at least theoretically) withstand 2²⁵⁶ attempts to create a collision.

The new attack technique combines what's known as a non-linear differential path with a boomerang attack. It decreases the cost of a collision attack by a factor of more than 2,000 compared with previous methods. The paper has not yet been peer reviewed.

The method works only for collisions, in which an attacker is able to control both messages that are hashed. SHA-1 is still believed to resist what's known as pre-image attacks, in which an attacker tries to create one message that generates the same hash value as a pre-determined message. ®