Are You Being Searched?

They know when you are sleeping, they know when you're awake, they know if you've been bad or good: They're the National Security Agency, and as The New York Times reported this Christmas Eve, they've been conducting analysis of telecommunications on a scale far beyond that of the targeted program of eavesdropping on domestic-to-international communications revealed earlier this month. Both programs remain shrouded in secrecy, but there's at least some reason to think that, under the logic of a Supreme Court ruling issued earlier this year, it's the more expansive one that will meet fewer constitutional obstacles.

The defenses offered thus far of warrantless wiretaps on U.S.-to-foreign communications of persons with "links" to al-Qaeda have been, if not quite tortured, then at least subject to coercive tactics. Consider, for instance, the argument that Congress' authorization of military force to pursue terrorists in the wake of 9/11 gave President Bush the power to authorize such surveillance—even though Attorney General Alberto Gonzales has said the administration declined to seek such authority explicitly for fear of being turned down. The argument turns on a strained analogy to a 2004 Supreme Court ruling holding that the authorization of force included the authority to detain captured combatants, a fairly obvious natural concomitant of war, even though it did not explicitly mention "detention." The alternative would be to conclude, ludicrously, that Congress intended a "take no prisoners" War on Terror, in which enemies must either be released or shot on the spot. But administration apologists—take the hairpin curve in this logic slowly or you may crash—have parsed the ruling as entailing that Congress therefore endorsed anything short of putting a bullet in a suspected terrorist's brainpan.

Prominent conservative blogger John Hinderaker turns his gaze on the Fourth Amendment's stipulation that governmental searches be "reasonable" and asks:

Is it reasonable for the administration to do all it can to identify the people who are communicating with known terrorists overseas, via the terrorists' cell phones and computers, and to learn what terrorist plots are being hatched by those persons?

Which is a fine rhetorical question, but one no more obviously helpful or relevant than its domestic equivalent: Is it reasonable for police to recover murder weapons and stolen goods? The function of judicial oversight is to make reasonably sure that the government is reasonably sure it's only doing such reasonable things.

Somewhat more plausibly, George Washington University law professor Orin Kerr suggests that domestic-to-foreign taps may fall under the "border search" exception to the Fourth Amendment, which permits warrantless searches of persons and parcels as they exit or leave the territorial United States, primarily for the purpose of keeping out contraband. That reasoning has the unsettling implication that we may routinely make ourselves subject to such searches without knowing it: In a world of global e-mail forwarding, international cell phones, and Voice over Internet Protocol technology that's indifferent to geography, digital border-crossing, unlike its physical equivalent, is easy to do accidentally. Even the NSA seems to have had trouble keeping its borders straight: On at least one occasion, it inadvertently tapped a purely domestic communication, one party to which was using an international phone number within the U.S.

What of the NSA's mass analysis program, then? It's hard to say much with certainty without knowing the specifics of what data is mined, and how, from the nation's telecommunications hubs and radio spectrum—and the agency isn't talking. But if the courts give a broad reading to a case handed down by the Supreme Court earlier this year, dealing with the seemingly disparate question of when drug-sniffing dogs may be used at traffic stops, law enforcement's sifting through our bits may not even count as a "search" under the meaning of the Fourth Amendment.

Back in 2001, in Kyllo v. United States, the Supreme Court considered how to treat the use of non-intrusive search technologies (in that case, a thermal imaging device used to hunt for marijuana grow-lights) under the Fourth Amendment. Justice Antonin Scalia spelled out what appeared to be a strict, clear rule:

[O]btaining by sense-enhancing technology any information regarding the interior of the home that could not otherwise have been obtained without physical "intrusion into a constitutionally protected area,"…constitutes a search—at least where (as here) the technology in question is not in general public use.

It's under that ruling that some news reports have suggested that government scanning of mosques for radiation may run afoul of the Fourth Amendment. But this year, in Illinois v. Caballes, the Court added an important caveat: The Kyllo rule would not apply if a search (like a dog's sniff) was tailored to reveal illegal activity without disclosing any other private information. And they held that to be the case even though, of course, even the best narcotics-sniffing dog is not infallible.

The New York Times report suggests that the NSA is, at the least, keeping tabs on "transactional" data about massive numbers of communcation—who is contacting whom, where, and at what times—in search of telltale paterns that might raise red flags. Yet the monitoring may well be more extensive: The agency's super-secret ECHELON program supposedly has the capacity to feed enormous streams of voice and data communciations through supercomputers using voice- and text-recognition technology to flag suspcious phrases—perhaps the names of terrorist operatives whose names have not been widely publicized by the media, and would likely be known only to terror operatives. If the NSA's filter algorithm were good enough, could it rifle through our correspondence with the legal impunity afforded a dog's schnozz?

Before the courts answer in the affirmative, they may want to give a moment's thought to British cleric and probability theory pioneer Thomas Bayes. Bayes' insight—now codified in Bayes' Rule—was that new information about how likely a proposition is to be true(whether it's "you have a disease" or "this person is a terrorist") has to be viewed in light of the independent, "unconditional" probability of its truth. In other words, if you test positive for a rare disease, and you know that the test is 95 percent accurate, that doesn't mean you've got a 95 percent chance of having the disease, because you have to factor in the low independent probability of having it.

We can pick some numbers to illustrate the point. Imagine, implausibly, that one in 10,000 people using communication technology is a terrorist. And imagine, equally implausibly, that the NSA's artifical intelligence is 99 percent accurate: It misses only 1 percent of genuine terrorist conversations, and mistakenly flags only 1 percent of non-terrorist conversations. For every million conversations it monitors, then, it will "alert" on 99 real terrorist communications and 9,999 innocent ones. And as Justice Souter noted in his dissent in Caballes, a drug dog need not have an accuracy rate anywhere near as high as 99 percent to be considered "reliable"—and, presumably, outside Fourth Amendment scrutiny. Such a filtering program might be restricted by statute, but, on this logic, it would not be subject to constitutional review.

The full implications of Caballes have not yet begun to be seriously tested. But if courts are not careful, a dubious dog-sniff ruling could leave us all on a very short leash.