Key points are not available for this paper at this time.
While face recognition can be a useful tool, helping authorities to narrow leads and confirm a criminal suspect's identity, the media tended to highlight the technology's limitations in the Boston Marathon bombings investigation. Consequently, we were motivated to reexamine the efficacy of unconstrained face recognition using this incident as a case study. We chose two automated face recognition systems known for their superior unconstrained matching performance based on tests conducted by the US National Institute of Standards and Technology: NEC's NeoFace v3.1 and Google-owned Pittsburgh Pattern Recognition v5.2.2. To conduct the experiment, we added six reference images of the Tsarnaevs taken from press releases and news articles following their identification to a background database of 1 million mug shots. Against this database, we searched for matches of the five face images of the brothers extracted by the FBI from surveillance, smartphone, or point-and-shoot camera footage before their identification.
Klontz et al. (Fri,) studied this question.