Mobile operators pooh-pooh universal phone-snooping plan
'What, us worry?'
Mobile operators have struck back at organizers of an open-source project that plans to crack the encryption used to protect cell phone calls, saying they are a long way from devising a practical attack.
"The theoretical compromise presented at the Black Hat conference requires the construction of a large look-up table of approximately 2 Terabytes - this is equivalent to the amount of data contained in a 20 kilometre high pile of books," the group, which represents almost 800 operators in 219 countries, said in a statement issued Friday. "In theory, someone with access to the data in such a table could use it to analyse an encrypted call and recover the encryption key."
The GSMA went on to say that even if such a table were built, the researchers still would need to build a complex radio receiver to process the raw radio data.
The vast majority of world's cell phone calls are protected by an algorithm known as A5/1 that has been in existence for more than a decade, said project leader Karsten Nohl, a cryptography expert and a researcher at the University of Virginia. Because it hails from the cold-war era when export laws prohibited the exportation of strong cryptography, the cipher is relatively trivial to break using a large number of networked computers.
More recently, cell phone makers have folded a newer cipher known as A5/3 into handsets to protect internet communications. Because its key is twice as long as A5/1, it's about a quintillion times harder to break, Nohl estimates. But despite the uncontested superiority of the newer algorithm, handset manufacturers still cling to the older one to protect voice calls.
"This is cold war stuff," Nohl told The Register in explaining why he spearheaded the project, which plans to use a distributed, peer-to-peer computing system to create the table needed to crack A5/1. "I don't see why I should be using something that was intentionally weakened during the cold war."
Indeed, in many respects, the GSMA's statement reads like a relic from the 1960s, when it was still fashionable to believe that arcane and proprietary technologies were sufficient to thwart determined hackers from breaking into protected systems.
"The complex knowledge required to develop such [signal-processing] software is subject to intellectual property rights, making it difficult to turn into a commercial product," it states.
The statement said the GSMA is phasing in A5/3, but provided no timetable or other details.
The GSM rainbow table project was announced at the recent Black Hat security conference in Las Vegas.
Nohl has already proved to be adept at defeating such security-through-obscurity defenses. In late 2007, he unveiled a practical attack on the world's most popular smartcard. For decades, Netherlands-based NXP Semiconductors relied on a proprietary algorithm to protect its widely used Mifare Classic wireless card.
"I'm puzzled by the GSMA's attempt to hide behind the alleged inability of hackers to snoop GSM traffic," Nohl wrote in an email to reporters. "This is 20 years old technology that ships in billions of handsets. The GSMA should take the hacker community and its current interest in GSM technology more serious." ®
On PC World's website, the going rate for a 1TB external drive is £70 or so. As others have pointed out, any 13year old knows there's no problem getting 2TB of storage these days, and nor is there any problem getting hold of someone else's "IP" (what do people store on 1TB drives anyway, except "someone else's IP"). Connecting it all together might take longer than an episode of The Simpons, but maybe not much longer.
GSMA, your PR people need to talk to their teenagers.
@John Savard - impractical?
You *never* have to do a linear search of a look-up table. The simplest sane algorithm is to order the data sequentially by key, and do a binary chop. You can get more speed by pre-ordering it into a search tree rather than a sorted list.
If you are searching disk-resident data, each access costs you a few milliseconds, so an efficient lookup into 2Tb will cost ~30mS if all the data is on disk, and ~10mS if you make constructive use of a couple of GB of RAM to cache the top of the tree. The question then becomes, do you have to do this so many times that the attack is impractically slow, or not?
Been there, done that!
It's called Echalon.