DEFCON18 — Las Vegas — A hardware hacking expert here at Defcon18 successfully faked several attendees’ cell phones into connecting to his phony GSM base station during a live demonstration that had initially raised concerns at the Federal Communications Commission (FCC).

Security researcher Chris Paget’s presentation here was aimed at demonstrating security weaknesses in the GSM protocol using a homegrown GSM base station running over ham-radio frequency. His so-called “IMSI Catcher” acted as a spoofed GSM tower and fake base station that could convince GSM handsets to connect to as the closest “tower” in proximity.

GSM technology is used in 80 percent of the world’s mobile phone calls, and has been the subject of previous security research poking holes in it. Paget said his intention was to demonstrate how the protocol is basically broken: “The main problem is that GSM is broken. You have 3G and all of these later protocols with problems for GSM that have been known for decades. It’s about time we move on,” Pager said in a press briefing yesterday prior to today’s demo.

Paget’s demo almost didn’t happen at all: it wasn’t until the late last night that he decided to go forward with the live demo of the hack after conferring again with Electronic Frontier Foundation attorneys after the FCC voiced its concerns that the demo might involve the unlawful interception of phone calls: “The response we got from the FCC is that [they couldn't] advise whether this is a good thing or bad thing, but here’s a long list of statutes you should read to make sure you’re not in violation,” Paget said yesterday. “It seemed almost a scare tactic to convince me not to go ahead.”

Paget was careful to issue warnings about his demo to attendees during his presentation today and that his demo was in no way for malicious purposes nor would it retain any data gathered from “owned” phones. His use of ham-radio frequency to carry the GSM signal got around any spectrum violation issues, he said.

He built the IMSI (International Mobile Subscribe Identity) Catcher, a phony GSM tower/base station, for about $1,500 using open-source technology, which he said is “a thousand times” cheaper than a similar commercial device used by providers. Aside from the device, the setup also used two directional antennas, and a Debian laptop running OpenBTS and Asterisk, an open source tool that turns a computer into a voice communications server. He used the device only to intercept and handle outgoing voice calls — which were sent via voice-over-IP — and not incoming calls nor data. SMS messaging would require getting caller ID information, which is difficult to obtain, he said.

“When the phone is looking for a signal, it looks for the strongest tower. This offers the best signal,” Paget said, even though it’s only 25 milliwatts.

The system only intercepts outbound calls, and callers whose phones connected to Paget’s phony tower would get a recorded message when trying to dial out. “When attached to my tower, your phone is [considered] off, so incoming calls go straight to your voicemail,” he said.

Overall, Paget captured anywhere from 17 to 30 phones at a time during the demo, even after configuring the base station to appear as an AT&T tower. The phones automatically defaulted to 2G because Paget’s base station is 2G. The base station could also be configured to disable encryption, he notes, as well as to target specific brands of phones to connect to it.

Paget destroyed the USB key that held any data gathered from the cellphones after the demo, so he didn’t know for sure the total number of phones that connected to it.

In previous tests, Paget found that iPhones most commonly connect to his fake GSM station.

He also discussed methods of speeding up the capture of cellphones during his presentation. The solution, he said, is to move to 3G. “3G and later is the solution … 3G authentication is much better,” Paget said. But that’s no small feat: the conversion would entail upgrading all phones, networks, and towers, he said.

Adding encryption could help protect phones from a malicious GSM interception attack, as well as using VoIP and noticing when the 3G icon is no longer present during a call, he said.

This isn’t the first time Paget has been at the center of controversy over his research. At Black Hat DC in 2007, his talk on cloning HID’s RFID-based proximity cards was pulled from the program at the eleventh hour after the RFID vendor threatened him with a patent lawsuit. “I had no choice,” said Paget, who was a researcher with IOActive at the time. “We were a very small company and we had to pull the talk … they threatened patent litigation, which is extremely expensive and can cost [millions]” even if it turns out the suit has no legs, he said.

The creators of the in-development technology say they’ll be able to crack GSM encryption with only about $1,000 worth of equipment.

Security researchers presenting recently at the Black Hat D.C. conference in Washington, D.C., demonstrated technology in development that they say will be able to greatly decrease the time and money required to decrypt, and therefore snoop on, phone and text message conversations taking place on GSM networks.

Many mobile operators worldwide use GSM networks, including T-Mobile and AT&T in the United States. The 64-bit encryption method used by GSM, known as A5/1, was first cracked in theory about 10 years ago, and researchers David Hulton and Steve, who declined to give his last name, said today that expensive equipment to help people crack the encryption has been available online for about 5 years.

Until now, however, it’s been prohibitively expensive for people to get their hands on this technology. If it works, the technology Hulton and Steve are developing should be able to crack GSM encryption in less than 30 minutes with about $1,000 worth of equipment, or in about 30 seconds with $100,000 worth of equipment. The technology could potentially be helpful to law enforcement investigators, but could also be taken advantage of by malicious hackers. Hulton says he plans to commercialize the more expensive version of the technology.

Other hardware Hulton and Steve referenced uses two different techniques to snoop on GSM calls and can cost between $70,000 and $1 million. So-called “active” systems simulate a GSM base station and don’t rely on encryption because they trick phones into connecting to the GSM network through them. Other, so-called “passive” systems snoop on the traffic and are far more expensive.

Hutton and Steve’s technology relies on the use of an array of devices known as field programmable gate arrays to first create a table of all the possible encryption keys — in this case 288 quadrillion — and then decrypt each of those over the course of three months. The resulting tables of keys could then be used by software to decrypt GSM communications, which first have to be intercepted using a receiver that can listen in on GSM frequencies.

During their talk, Hulton and Steve also discussed the vulnerabilities of mobile device SIM cards, noting that GSM networks broadcast SIM cards’ unique IDs in unencrypted text, which can tell attackers or law enforcement what kind of phone someone is using. The GSM network also can tell snoopers how far a phone is from a base station, within 200 meters of error. They noted that SIM cards run Java Virtual Machines that operators have access to, and suggested that it could be possible for malicious attackers to install applications on user’s phones without them ever knowing, potentially rerouting traffic to a third party who listens in to phone conversations.

The GSM Association, a trade group representing more than 700 GSM operators, said it could not comment on the specific claims Hulton and Steve are making. However, spokesman David Pringle said in an e-mailed statement that while researchers have showed how A5/1 could be compromised in theory, none of their academic papers have led to “practical attack capability that can be used on live, commercial GSM networks.” He also noted that more advanced encryption is beginning to be deployed for GSM networks and that other networks, including 3G networks, don’t use A5/1.