Fingerprints
In a guest post on Andrew Sullivan’s blog last week, I described two new technologies that track and mine some of the transaction data created by commercial interactions. I indicated that this, rather than government spies, is driving most of the loss of privacy in the contemporary era.
Here is a great example of crossover: cops using your shopping patterns to find where you have moved even if you’ve changed your identity and live on all cash. I’ve worked a lot with these datasets, and it wouldn’t be quite as simple as the author indicates, but this is very doable.
On the other hand, it’s always amusing how simple attacks with $20 of equipment that use physical mechanisms – demonstrating the dangers of viewing technology only at an abstract level is dangerous – can overcome massive security systems. Individuals can sometimes beat Big Brother.
Where can one get datasets with this sort of data?
— TW Andrews · Mar 3, 01:58 AM · #
Having read the article, I’m a bit skeptical that the sort of pattern identification described here is possible. I’ve done work along the same lines—biomarker identification in genomics and protiomics—and the signal-to-noise ratio is such that you need a very strong pattern to be confident to even justify further testing. Additionally, the problem of false positives is quite important in such cases.
Though, despite my skepticism, as my previous comment indicates, I’d be very interested in looking at datasets of this nature.
— TW Andrews · Mar 3, 02:23 AM · #
TW:
I didn’t want to go into detail in the post about it not being “as simple as the author indicates”, but there would be some difficulties and limitations.
Getting the datasets is the big difficulty. Every major chain (Safeway, Wal-Mart, etc.) has transaction logs that even on non-loyalty-card data can generate market baskets. You would have to get all the major chains to cooperate, which wouldn’t be easy.
The limitations are that (1) the guy, if he knew this, could do all of his shopping at non-chain c-stores without t-logs, and (2) even if his shopping patterns were sufficiently unique to overcome what you correctly state as the signal-to-noise problem, he could purposely modify them after moving to fool you.
— Jim Manzi · Mar 3, 01:32 PM · #
I like your idea of a market for private data. There was a suit some years ago on this subject – a guy sued a magazine for selling mailing lists including his name without his permission, claiming that just as celebrities control commercial use of their names he should be able to also. He lost, unfortunately.
I also agree that in terms of quantity commercial interactions dominate the loss of privacy, but there’s something to be said for quality also. When Safeway looks through my records, the object is clear – they want to sell me more toothpaste. I don’t even want to think about the object when Cheney listens to his recordings of my phone calls.
— Peter · Mar 4, 03:49 AM · #