Congress Comes to 'Help' Run the Internet

OPINION: To the proposed Cybersecurity Act of 2009, add work in the House on a privacy act that could end up banning security functions by ISPs. Government regulation at its best is coming to the Internet.

The early life of the Internet has, perhaps, suffered from an excess of libertarian impulse, even from those who don't think of themselves as libertarians. Fear that the government would impede freedom of individuals on the Internet has led to opposition to just about any opening for law enforcement on it. Now the pendulum has swung and we're heading 180 degrees in the opposite direction. The new Democratic Congress seems determined to bring regulation to every part of the Internet it can find.

First we had the Cybersecurity Act of 2009, under consideration in the Senate, which envisions a vast expansion of federal, and specifically White House, authority over the security of large parts of the Internet and those who engage in the business of security on it. Now we have Rep. Rick Boucher and his Subcommittee on Communications, Technology and the Internet considering privacy on the Internet and making noises about restricting the use of DPI (Deep Packet Inspection).

Security people will recognize DPI as a very broadly applicable technique. In the most general sense, it means examining the data contents of a packet without being the intended recipient of that packet. If you read the news reports (like the one I linked to above), you get the impression that this is about new forms of targeted advertising in which ISPs use DPI to learn about the interests of subscribers; the example everyone uses is that you figure out someone is a dog owner and so you show that person ads for dog food. Sounds innocuous, even helpful, no? But some are creeped out by the possibility, and it's not hard to see why. If they can figure out I have a dog, what else can they figure out about me?

There's a lot more to DPI, of course, such as spam filtering: You can't do contextual analysis such as Bayesian analysis of spam without looking at the contents of packets. You can't do malware scanning without looking at the contents of packets. You can't do certain intrusion detection and prevention functions without doing the same. In other words, unless we disagree on the benefit of such functions, you can't do many security operations purely for the benefit of the consumer without doing DPI.

But that's not what it's about, you're saying, it's about the advertising stuff, right? Not according to Leslie Harris, president and CEO of the Center for Democracy and Technology, who testified before the subcommittee April 23. Harris' prepared remarks (PDF) state:

"It is important to stress at the outset that all applications of DPI raise serious privacy concerns because all applications of DPI begin with the interception and analysis of Internet traffic. Policymakers must carefully consider each use of DPI and balance the perceived benefit of its use against the risks to privacy and civil liberties, as well as to the Internet's character as an open platform. CDT believes that only rare uses of DPI will be acceptable after such a balancing. Today, DPI applications include management of network congestion, detection of network threats, content blocking for intellectual property protection and child safety, behavioral advertising, and government surveillance."

Harris reminds me, DPI is also useful for prioritizing data that is time or latency-sensitive. This is, at least arguably, for the benefit of the user. The CDT pays brief lip service to network maintenance and security, but clear sees that as a secondary concern. The CDT even objects to content filtering that customers might want to protect their children.