Today’s Colloquium on Information Technology & Society at NYU Law School featured talk by Jonathan Phillips of the National Institute of Standards and Technology (NIST) on recent developments in facial recognition technologies and algorithms. Some of the results suggested that having more than one image of a subject in the database to compare to improved the accuracy and reliability of the system (which makes sense).
My concern with such a finding is that if the makers and users of facial recognition systems see a value in having multiple pictures of me in their databases, then a new market might emerge for the collection and aggregation of my likeness.
In some cases, this might not be problematic. For example, if NYU wanted to use facial recognition systems to regulate who can enter the library, they could take 10 photos of me (from various angles, with various lighting conditions) when I have my official photo taken for my NYU identification card. I would likely allow this presuming that its purpose is to reduce false negatives when I try to enter the library.
But, what if the Times Square Business District or other places that rely on facial recognition for identification purposes (rather than NYU’s verification) want to have multiple images of people in their databases? The likely methods to get these images would seem to be (a) see if government agencies will provide images (the DMV, for example), (b) solicit images from other bodies who have my photo on file (NYU, my employer, my health club), (c) crawl the web for my image (from my website, my Friendster page, my Flickr account, or (d) rely on commerical data aggregation companies to do this for them.
Consider that last example. The need for a wide array of personal and transactional data about individuals has led to companies like Choicepoint to aggregate large amounts of public (and non-public) information about people. They mine through databases, web pages, and even send workers to government file rooms to digitize public records to capture information for their vast databases. Now imagine if there was a market for my image data as well as my transactional data. Would Choicepoint send workers to digitize yearbook photos? Aggregate still images from surveillance camera footage (Citibank has pictures of me every time I use their ATM machine with my debit card)? Systematically scrape the web for my likeness?
The idea that images of me could become part of the vast commercial data aggregation efforts currently underway is not something I have considered (nor had Phillips) – and its something I’m not particularly comfortable with either.