COMMENTS OF THE ELECTRONIC PRIVACY INFORMATION CENTER

The unreliability of facial recognition as a law enforcement investigative tool does not merely stem from potential misidentifications by the facial recognition algorithm itself, but also stem from the various steps police typically take in the use of facial recognition for identification, including: 1) the selection of a probe photo, 2) the choice of database to use for the search, 3) the preproc. [...] The angle of the photo, the lighting, and the sharpness of the photo, among other things, can all have an impact on the accuracy of the search results, so much so that oversight and scientific bodies have issued standards for photos and vendors have made minimum photo quality recommendations.13 The database a probe image is run against affects the reliability of the results. [...] The unfamiliarity of the subject, low-quality images, the different angles of photos, the obfuscation of facial features, among other things, all contribute to the difficulty of identifying whether a probe photo and one of the photos from the candidate list are indeed a match. [...] It would destroy anonymity and put the control of identification in the hands of the government and further exacerbate the imbalance of power between the government and the people. [...] Agency promises and policies and even federal regulation are not enough to mitigate the risks of facial recognition technology The lack of strong federal regulation of facial recognition should be the death knell of the federal government’s current use of the technology.