Wednesday, May 24, 2006

Kevin Drum continues to stress the illegality of the NSA operations (the important issue) and stimulate me to argue that he shouldn't seriously consider the possibility that the program is effective.

I don't understand why you seem willing to believe that the NSA program might actually work. The NSA has powerful computers and a huge data base, but their guess that "flurries of calls to U.S. numbers placed immediately after the domestic caller received a call from Pakistan or Afghanistan." are indicative of an Al Qaeda operative is just a guess. There is, as far as I know, no evidence whatsoever which supports this hypothesis. In particular, since there are, as far as I know, no known al Qaeda sleepers in the USA the association of the pattern and al Qaeda sleepers is based on a sample of zero observations. The fact that after making that guess the NSA crunches terabytes does not reduce their reliance on the original guess. A calculation based on a bad assumption will not yield an accurate conclusion (this is a case of gigo). I really think that many people are awed by the power of computers and willing to believe that they can do the impossible, that is, accurately describe something which has never been observed.

Now, as to the case of Kevin Drum, I understand that you always write that the NSA approach *may* work, *might* be useful, and must be regulated by law *if* it works. Still, persistent interest in the implications of the assumption that their approach works, hints at a belief that there is a serious possibility that it does work.

I think that, by now, it is clear that the NSA approach is both illegal and ineffective. Someone who leaks to reporters has attempted to point to examples of useful applications of the program. Thus not all people in the know are restrained by respect for the secrecy of sources and methods. This is obvious by now. Exibit A was the Brooklyn Bridge Blowtorch Buffoon who really did consider destroying the bridge with a blowtorch. In that case there was a warrant authorising a phone tap. Obviously pattern analysis had nothing to do with that detection of a threat to the USA, since warrants are not issued based on illegally obtained data.

Also the FBI complains about "another investigation of Pizza Hut" (see the very first comment). The FBI will always complain about expending resources doing leg work for other organisations, but the evidence so far is that the NSA program is a total failure.

Dear Mr Drum

(in a comment on his blog I dared call him Kevin)

It seems to me obvious that terrorists will not create obvious patterns. If activity looks a lot like al Qaeda communications, al Qaeda operatives will refrain from that activity. That's game theory. The examples of useful pattern analysis (pedophiles, who is in the super bowl noted on this blog) are examples in which there is a large enough sample of the phenomenon of interest that the association of a pattern and the phenomenon is based on evidence not guess work. The NSA might have the worlds largest data set, but without a sample of al Qaeda sleepers in the USA, they can't possibly learn about patterns typical of al Qaeda sleepers in the USA.

No comments: