Can They Hear You Now?
How the FBI eavesdrops on Internet phone calls (and why it sometimes can't).
The Federal Communications Commission * and the Justice Department are at loggerheads over a new problem in the war on terror: how to listen in on Internet phone calls. Thanks to the blistering growth of VoIP—Voice over Internet Protocol—services, which have been adopted by approximately 10 million people worldwide so far, law enforcement officials now worry that wiretapping may one day become technically obsolete. If traditional phone lines go the way of the horse and carriage, will the FBI still be able to listen in on Internet phone calls? How would it go about tapping one? Is it even possible?
I contacted three of the leading VoIP providers in the United States—Time Warner Cable, Vonage, and Skype—to ask them how they would comply with a court order to permit a wiretap. As it turns out, the Justice Department has good reason to worry. Depending on the provider, tapping a VoIP call can be either tricky or impossible.
For Jeffrey Citron, the CEO of Vonage, the critical problem is this: The 1994 law that dictates how telecoms must cooperate with the feds (it's known as CALEA) stipulates that government agents can listen in on phone calls only in real time. They are not permitted to record calls and play them back later to check for incriminating information. But as Citron explained it, on Vonage's system, it is technically impossible (for now) to listen in on a live phone call.
Here's why: A VoIP call transforms your voice into digital bits, then segments them into separate packets of data that are routed through the Internet and reassembled upon arrival at the other end. From an old-fashioned perspective, there is no actual "sound" passing through the Internet at any time—the PC or other device you use to place the VoIP call digitizes your voice in your home. Of course, a huge amount of regular phone traffic is also segmented into digital packets at some point, but such calls are digitized and then reconverted into sound waves far deeper into the telephone system, at points outside private homes. Law enforcement can therefore listen in on your line within the telephone system itself; the technology to do this is already embedded in the phone company's switches.
In theory, Vonage could comply with a tap request by making a copy of the call in real time and streaming that call to a law enforcement agent. But that tack would violate CALEA, since Vonage would still be making a copy of the original call. The alternative, Citron says, is for Vonage to modify its VoIP system so that its digital routers include analog-friendly wires capable of producing a real-time sound wave. These could then be linked to a law enforcement agency, permitting simultaneous listening-in. Citron says making the shift would cost Vonage a few million dollars—before taking any action, he's awaiting further regulatory instructions from the FCC. The company has already complied with between 10 and 100 requests from various government agencies for general information (including call records and billing history), but to date, he has yet to receive a single request for a live tap into a Vonage call.
Time Warner Cable, which has announced that it will make VoIP available to all its digital cable markets by the end of the year, would have a much easier time wiretapping live phone calls. That's because Time Warner owns the underlying infrastructure its VoIP service relies on. So while Vonage could offer government agents access only to the handful of routers it uses to direct its calls over the wider Internet, Time Warner can offer them direct access to the cables, routers, and switches over which its VoIP calls travel. It could, in theory, open a live channel for law enforcement at the place where Time Warner's cable modem signals are routed onto the wider, public Internet. This switch, known as the Cable Modem Termination System, is a natural junction where a company like Cisco, which already builds CMTS hardware, could easily and cheaply add in CALEA-compliant technology.
Why, then, couldn't the feds tap any VoIP call by listening in on the line at the CMTS? Because some VoIP calls are routed, digitized, or encrypted in ways that law enforcement can't decipher. Skype, which now boasts 7 million users, specializes in such encryption. The company's system is designed to thwart potential eavesdroppers, legal and otherwise. The difference begins with how the networks are designed: Both Time Warner and Vonage offer VoIP services that run through centralized networks. For instance, when I place a call through Vonage, it starts by going to a centralized Vonage computer, which in turn looks up the phone number I am dialing and routes the call over to the traditional phone system. This is a classic instance of a "hub and spoke" network. But Skype, built by the same people who brought us Kazaa, is a totally distributed peer-to-peer network, with no centralized routing computers. (That's possible in part because Skype calls can only be sent and received by computers—you can't call a friend with an analog phone.) As a result, the company's network looks more like a tangled spider web, and the packets that make up your voice in a Skype call are sent through myriad routes to their destination. Part of the brilliance of the Skype software is that it has learned to use desktop PCs as "supernodes," each sharing some of the load needed to route Skype calls quickly to their destination. From the caller's perspective, this is all invisible: The call just works.
Since it's exceedingly difficult to follow the path that a Skype call makes through the network, law enforcement agents would be hard-pressed to figure out where to place a tap. But even if they could, the company has built in such strong encryption that it's all but mathematically impossible with today's best computer technology to decode the scrambled bits into a conversation. Here's how Skype explained it: "Skype uses AES (Advanced Encryption Standard)—also known as Rijndel—which is also used by U.S. government organizations to protect sensitive information. Skype uses 256-bit encryption, which has a total of 1.1 x 1077 possible keys, in order to actively encrypt the data in each Skype call or instant message." The point of all this mumbo-jumbo is that Skype uses an encryption algorithm * known as 256-bit AES. The National Institute of Science and Technology states that it would take a computer using present-day technology "approximately 149 thousand-billion (149 trillion) years to crack a 128-bit AES key." And that's for the 128-bit version; Skype uses the more "secure" 256-bit standard. Since computers have a way of quickly getting more powerful, the institute forecasts that "AES has the potential to remain secure well beyond twenty years."
Moreover, Skype says, the company does not keep the encryption "keys" that are used to encode each Skype transmission—each one is generated and then discarded by the computer that initiates the call. So government agents couldn't force Skype to turn over the keys needed to decrypt a call either.
Last Thursday the FCC held an open hearing on the future of VoIP telecommunications. In a 4-1 decision, FCC commissioners, supported by Chairman Michael Powell, voted that a VoIP provider called Free World Dialup should not be subject to the same regulations as traditional phone companies—including the particulars of CALEA compliance. Instead, the FCC decided to put off the issue, stating that it would initiate a proceeding "to address the technical issues associated with law-enforcement access to Internet-enabled service" and "identify the wiretapping capabilities required." One commissioner, Michael J. Copps strongly dissented, calling the postponement "reckless."