Over on Spark, the latest show talks about privacy. There are several interesting segments, including one about how a group of folks were able to take pictures they snapped of folks on a campus, use facial recognition and matching software to compare it against photos in Facebook to determine a name, date of birth and state of birth, turn that into a partial social security number, and fetch out private information of 1 out of 3 folks.
So, think about that. This is a research project able to map individuals based on information that they’ve most likely posted about themselves to gain additional information that they probably didn’t think was public. Scary, huh?
Now think about the fact that this sort of software is still in it’s infancy and that it makes mistakes. Now think that with a population rapidly approaching seven billion, there’s probably one other person out there that happens to look quite a bit like you, or at least enough like you that you fit within the error margin of the categorization software. Let’s also think about the fact that some industries have been known to adjust prices based on perceived demographics such as local income levels (think car dealers, airline tickets, fashion, etc.). Imagine in a far more personalized shopping environment, what you’d pay if your unknown doppleganger were to suddenly win the lottery?
As one of the interviewees notes, it takes surprisingly little information in order to isolate to an individual, but the general eagerness of system users tends to push the “success” margin lower than it probably needs to be. If you match to 90% accuracy, you match to 10 out of 100, or 3,000,000 out of the population of the US. Most of those 3,000,000 probably don’t have outstanding warrants or are sex offenders or declared bankruptcy within the past 10 years. Hopefully, you’re outside of their error margin when traveling, attending your niece’s recital, or applying for a loan. After all, they paid a lot of money for that brand new system, and everyone knows that you can forge an ID pretty easily. Still, what are the odds?
And if you can’t accurately answer that question, you need to worry about your privacy.