Friday, May 11, 2007

NSA, state secrets, and bonding: A suggestion

A government agency does something that is arguably illegal and ends up in court. Their attorney claims, perhaps truthfully, that in order to defend the case he would have to reveal facts about the agency's activities that must, in the interest of national security, remain secret. He accordingly asks the judge to dismiss the charges.

This situation--I'm thinking of litigation over the National Security Agency's wiretapping and related activities--raises a real problem. One can imagine a similar problem arising in a patent infringement suit when the defendant argues that, in order to show it is not infringing the patent, it would have to reveal a valuable trade secret. Our usual mechanism for settling legal disputes requires both sides to have free access to the relevant information; in some cases giving them such access could result in very large costs.

There is, I think, a fairly straightforward way in which we might deal with such situations. It depends on two assumptions. The first is that there is some responsible person--the President, the head of NSA, the research director of the corporation sued for patent infringement--who already knows the secret. The second is that the secret will expire within some predictable length of time. If the NSA is using top secret high tech to data mine phone records, it is unlikely to still be using the same method ten years from now.

Under my system, when the attorney asks for dismissal on grounds of state secrets or the equivalent, his request must include a time limit after which he agrees that the information will no longer be secret; the longer the time limit, the less plausible his claim and so the less likely the judge is to accept it. Prior to dismissing the claim, the judge asks the relevant responsible person, the one who already knows the secret, to testify under oath that the information does not confirm the other side's case, that revealing it would not result in a near certain verdict of guilty. When the time limit on the secret has expired the court takes up the case again, determines whether the testimony was true, and imposes penalties for perjury if appropriate.

Before taking the testimony the judge informs the witness that perjury in this particular situation is a capital offense. "Now, Mr. President, would you like to reconsider your position?"

That is, of course, an implausibly extreme version of the proposal. More generally, in order for the case to be dismissed, someone with the information must testify as to the implications of the information for the case, with penalties for false testimony severe enough to make it unlikely that he will be willing to perjure himself. In effect, he is bonding his testimony--in the limiting case with his head.

Obviously there are some practical problems, beginning with the fact that the person who knows the secret is probably not a lawyer. He should, however, have access to lawyers; it should not be that difficult for him to determine what sort of facts would or not result in his side losing.

How do we currently deal with such situations? One approach is to provide the secret information to the judge but not to the opposing side. But without the usual mechanisms of an adversarial trial, the judge may have a hard time telling how true or how complete the information he is given is. If he later discovers that what he was told was deliberately misleading, he has the option of imposing penalties for contempt of court, giving at least a weak version of my proposal.


Charlie Martin said...

David, have you ever had any real exposure to classified information? It doesn't appear so: first because you're apparently unaware that classified information already has an expiration date (in almost all cases, no more than 30 years, and in a development completely uncovered by the media, the "most secretive administration in history" has unilaterally reversed previous policy, so that material is automatically declassified after 30 years unless a cse is made to maintain classification, rather than the old policy); and second, because (perhaps for purposes of hypothesis, rather than ignorance, but you can't tell from this) you make the rather peculiar argument about a lack of lwyers cleared for certain material. Believe me, if there's one thing the agencies don't lack, it's lawyers.

David Friedman said...

In response to Seneca--whose English is surprisingly good for an ancient Roman ...

1. I have had no real exposure to classified information.

2. The real point of my post wasn't about classified information--that was just the example that inspired it. The real point was to suggest an approach to dealing with situations where it was important to establish truth, but revealing the secret information needed to do so would be undesirable.

Anonymous said...

In cryptography, there exist zero disclosure proofs. By answering a series of challenge, it is possible to prove that some statement is true without revealing information.

Without using so much formalism, there are very simple cases of this. I can prove I made an anti-gravity device by some direct confirmations without revealing the inner working. I can prove that I can break some encryption schemes without showing how.

With the right formalism, it is even possible to prove that I have a proof of any mathematical statement without revealing the proof itself.

With a lot of work, maybe such a formalism could be brought to resolve disputes with zero disclosure.

Of course, the easiest way to do that is to use a trusted third party, and in many case it should be more than a sufficient guarantee. A private company could easily do that and provide claim verifications without disclosure. But the point is that theoretically (as long as one way function exist) a trusted third party is not even necessary.

Anonymous said...

Seneca, the issue is that when the classified information is revealed after 30 years, and it turns out that government officials were lying about it, absolutely nothing will happen to them. Government officials generally don't prosecute other government officials for perjury - ask any defense attorney, they will tell you of cases where a cop was caught perjuring himself in the courtroom, and nothing happened to him. Beyond that, a 30 year span means that most lawsuits and criminal laws are impossible because of statutes of limitation, and even if it is possible to get a case into court, it's generally impossible to prove it. After 30 years evidence is hard to find, memories have faded, etc.

So, there's no real downside to officials lying about classified material.

Arthur, those mathematical theorems you cite are only applicable within very narrow domains, not to the kind of evidence we're talking about. As for trusted third parties, the judge is supposed to be a trusted third party, but:

1) The judge likely owes his job to the boss of the people who are presenting the case for keeping the secret, and

2) the judge will not know whether claimed "facts" are really facts or could be easily disproven by the other side in the case - if they knew what they needed to disprove.


Anonymous said...

Two further objections: first, the government does not seem to have ever been short on people willing to lie on its behalf, especially when they are convinced that it is their patriotic duty to deceive the people. Second, it seems unlikely that appropriate follow-up will ever occur.

Let me suggest an alternate approach: rather than trusting an involved party to accept future penalties, trust an uninvolved party. The judge (perhaps with the aid of suitable amici curiae) selects a knowledgeable and reputable person or small group. This group, along with the judge, is sworn not to reveal the allegedly sensitive information. They see all the evidence from both sides and determine whether or not there is a probability of success in the action.

The best part is that we already have a similar system -- the Grand Jury.