The Guardian has a story about ‘Optic Nerve’, GCHQ’s operation intercepting and collectin frames from Yahoo webcam feeds. It contains a couple of choice quotes from the agency’s documents. The first, and perhaps most telling is:
“One of the greatest hindrances to exploiting video data is the fact that the vast majority of videos received have no intelligence value whatsoever, such as pornography, commercials, movie clips and family home movies.”
Bulk collection is, perhaps, leading to wasted effort and, perhaps, leading to a counter strategy – to flood the databases and servers with too much information, in a way like Hasan Elahi. Putting one’s life online as the ultimate alibi and extra information as a counter surveillance measure.
My favourite part of the article, though, is “it noted that current ‘naïve’ pornography detectors assessed the amount of flesh in any given shot, and so attracted lots of false positives by incorrectly tagging shots of people’s faces as pornography.”
The idea of a naïve pornography detector seems hilarious, but it also picks out a further problem with the mass of data collected – it’s not possible (or at least efficient) to trawl it manually so in the absence of truly accurate or intelligent algorithms it’s borderline meaningless. Again this brings to light a method for skirting surveillance – image creation for algorithms, to amplify the expected results. Spoofing data by talking to the processes observing us.