New GSM encryption cracked

An encryption algorithm designed to protect calls on GSM phones has been broken by three cryptographers using only a dual-core, Intel -based Dell Latitude PC running Linux.

In a just-released paper, the three researchers from the Weizmann Institute of Science in Rehovot, Israel, described a technique they developed called a "sandwich attack," which they used to derive the complete 128-bit key of the Kasumi encryption algorithm. Kasumi, otherwise known as A5/3, is the block cipher used to encrypt communications on 3G wireless networks.

Orr Dunkelman, one of the researchers, said the sandwich attack improves on research showing ho w the Kasumi cipher block could be theoretically attacked.

"What the research shows is you can reduce the amount of data and computing time you need" to extract the full 128-bit A5/3 encryption key, Dunkelman said. "Previously, if I had at my disposal all of the computing power in the world, it would have still taken me a year to find the key. Now it would take less than two hours on a single computer. We can all agree that's a bit disturbing."

The two other researchers involved in the report are Nathan Keller and Adi Shamir, who is one of the inventors of the RSA encryption algorithm.

About 1.2 billion handsets in use around the world are A5/3 ready, the researchers said, but only a few of the 800 or so wireless carriers globally have implemented the technology on their networks. Once adopted, A5/3 will become one of the most widely used cryptosystems in the world. "Its security will become one of the most important practical issues in cryptography," the researchers said.

According to Dunkleman, Kasumi was supposed to have been stronger than the current A5/1 encryption standard that is used to protect GSM telephony. Only last month, security researchers published a method for determining the complete A5/1 encryption key using specialised encryption cracking tables. The research showed how GSM phone conversations could be easily tapped using just a few thousand dollars' worth of hardware and software.

Concerns stemming from that research and others prompted the GSM Association to accelerate the transition to the new A5/3 algorithm. But the research shows that the encryption algorithm is considerably weaker than many might have imagined, Dunkleman said. "The security margin of the a5/3 is much lower than what is expected. We found out that it is not secure enough compared to the existing standard and compared to what was expected of it," he said.

The weakness stems from changes that were made to an encryption algorithm named Misty on which the A5/3 is based. In trying to make A5/3 a faster and more hardware-friendly version of Misty, the GSM Association appears to have weakened it. The attack described in the new paper exploits a "sequence of coincidences and lucky strikes when Misty was changed to Kasumi," the researchers said in their paper.

While the research exposes flaws in Kasumi, taking advantage of those flaws could be tricky, Dunkleman said. The so-called related key technique that was used to pull off the attack described in the paper is considerably harder to pull off in real life, Dunkleman said. So it's inlikely that anyone will be able to use the technique to listen in on 3G calls any time soon, he said.

Even so, the research is an important piece of work, said Bruce Schneier, a noted cryptographer and chief security technology officer at BT. "It does not have any immediate practical implementation because it is a related key attack," Schneier said. To pull off a related-key attack, an attacker needs to have access to the relationships between the plaintext and ciphertext, which can be hard to obtain in real life.

"But it is a really good math attack," which exposes Kasumi to be significantly weaker than the GSM Association may have intended when they modified Misty, he said. "This thing should be fixed, but there's no panic here,'' he said.