Computer scientists to FBI: don't require all our devices to have backdoors for spies


In an urgent, important blog post, computer scientist and security expert Ed Felten lays out the case against rules requiring manufacturers to put wiretapping backdoors in their communications tools. Since the early 1990s, manufacturers of telephone switching equipment have had to follow a US law called CALEA that says that phone switches have to have a deliberate back-door that cops can use to secretly listen in on phone calls without having to physically attach anything to them. This has already been a huge security problem — through much of the 1990s, AT&T's CALEA controls went through a Solaris machine that was thoroughly compromised by hackers, meaning that criminals could listen in on any call; during the 2005/6 Olympic bid, spies used the CALEA backdoors on the Greek phone company's switches to listen in on the highest levels of government.

But now, thanks to the widespread adoption of cryptographically secured messaging services, law enforcement is finding that its CALEA backdoors are of declining utility — it doesn't matter if you can intercept someone else's phone calls or network traffic if the data you're captured is unbreakably scrambled. In response, the FBI has floated the idea of "CALEA II": a mandate to put wiretapping capabilities in computers, phones, and software.

As Felten points out, this is a terrible idea. If your phone is designed to secretly record you or stream video, location data, and messages to an adverse party, and to stop you from discovering that it's doing this, it puts you at huge risk when that facility is hijacked by criminals. It doesn't matter if you trust the government not to abuse this power (though, for the record, I don't — especially since anything mandated by the US government would also be present in devices used in China, Belarus and Iran) — deliberately weakening device security makes you vulnerable to everyone, including the worst criminals:

Our report argues that mandating a virtual wiretap port in endpoint systems is harmful. The port makes it easier for attackers to capture the very same data that law enforcement wants. Intruders want to capture everything that happens on a compromised computer. They will be happy to see a built-in tool for capturing and extracting large amounts of audio, video, and text traffic. Better yet (for the intruder), the capability will be stealthy by design, making it difficult for the user to tell that anything is amiss.

Beyond this, the mandate would make it harder for users to understand, monitor, and fix their own systems—which is bad for security. If a system's design is too simple or its operation too transparent or too easy to monitor, then wiretaps will be evident. So a wiretappability mandate will push providers toward complex, obfuscated designs that are harder to secure and raise the total cost of building and operating the system.

Finally, our report argues that it will not be possible to block non-compliant implementations. Many of today's communication tools are open source, and there is no way to hide a capability within an open source code base, nor to prevent people from simply removing or disabling an undesired feature. Even closed source systems are routinely modified by users—as with jailbreaking of phones—and users will find ways to disable features they don't want. Criminals will want to disable these features. Ordinary users will also want to disable them, to mitigate their security risks.

Felten's remarks summarize a report [PDF] signed by 20 distinguished computer scientists criticizing the FBI's proposal. It's an important read — maybe the most important thing you'll read all month. If you can't trust your devices, you face enormous danger.

CALEA II: Risks of wiretap modifications to endpoints