Replies

There is some merit to this idea, although everything said about its potential for becoming a civil liberties nightmare is true.

Several such systems exist today, and to some degree they work. If someone were to steal your credit card without your knowledge, by the second or third transaction -- and maybe even the first -- a neural net system that has been watching you for a long time would likely notice "out of character behavior" for you and bring these transactions to the attention of a human. This is in place today; it watches over an incredibly large data stream of everyday credit card transactions that no human -- or even team of humans -- could keep up with.

There is nobody 'spying on you' here, it is just a machine, which promptly forgets 99.999% of what it looks at. It's really watching for aggregates and patterns. Could someone who works at the vendor that provides this system peek at what you do? Yes, but there is such a torrent of the stuff that you're basically anonymous in there.

The point is, this is not "profiling." There is no profile of what a credit card fraudster looks like. The expectation is that within this huge data stream of real-time credit card transaction data, all but a minuscule number of transactions are legitimate. What's more, people don't like to be hassled, so the system has to have a very low "false positive" rate, or the banks won't use it. In spite of those challenges, the damned thing works, and works pretty well.

Even though it may sound like pie-in-the-sky to examine a huge data stream to find one bad guy in a sea of good ones, this problem is solvable... it has been solved in some areas.

This debate is better carried on in terms of the privacy threats and the potential for abuse, than over speculative claims that it "won't work." It may very well work.

It may very well work.

Possibly. Unless the gov't is doing the work themselves. They have aleays been horrible at such systems. Consider fraud in any gov't health program or programs like food stamps. The fraud is huge.

If private firms do the work it depends on who gets the contracts for such projects. This could be something someone like snORACLE gets their fingers in and it gets blown up so big it never works. But, if outfits like Choicepoint and others get involved this idea could get interesting.

One thing you are correct about. Building various real-time data analysis systems looking for negative patterns is not that big of a deal. And if they are constructed correctly there are no real privacy issues because the system does not care about Joe Blow. Not that there is much in the way of electronic privacy anyway. That cat left the bag years ago. And the only way to put it back in is to outlaw electricity. I don't see that happening.