by Jonathan Capra
I’ve been glad to have the opportunity to engage my local police officials and community policing groups I am involved with on the implications of bio-metrics on policing and privacy rights in the future. Foremost, this has involved facial recognition technologies as might be applied to officer-worn body cameras or other police surveillance technologies.
As I expressed my concerns with these emerging technologies, it became clear that the phrase ‘facial recognition’ didn’t have one distinct meaning. There were many different processes that got lumped into the heading of ‘facial recognition’. Each had different implications and levels of cause for concern.
It was for this reason that I decided that I would try to coin some new nomenclature in order to give more clarity to the conversations we have about this emerging tech. I wanted a concise way to distinctly explain the different applications of bio-metrics in police surveillance technologies:
- TIER I: Footage from an officer’s body camera is uploaded into a video server at the end of their shift. The footage is then analyzed to merely identify the presence of people and note what time in the duration of the video they appear. It might also count how many people are detected in the frame at one time. Eventually we might see it noting distinctive features such as hair or skin color. Lastly, it might use the same algorithm to detect the presence of vehicles, landmarks or animals.
- TIER II: An officer or their superior can scan to a specific moment in a video recording where the officer interacted with a person. They can then manually choose to have that person’s faced further analyzed against an outside database of headshots – perhaps one containing people with outstanding warrants. Like a visual Google search, the system would then reply back with a list of possible matches.
- TIER III: Anytime any person is in view of an officer’s camera, the footage is automatically cross-referenced with aforementioned databases of headshots. Again, this could be a database of folks with outstanding warrants… But it could also be people who owe child support, or are on a terrorist watch list, or just the entire database of DMV license photos. This automatic process could occur at the end of the officer’s shift when it is uploaded to a server, or it could occur in real-time such that the officer receives some alert the moment he or she is in view of a person.
I feel Tier 1 bio-metrics have very few implications that might concern us. What’s being created in this case are just intelligent ‘bookmarks’ that help the viewer scan to a certain point in the video to review. For instance, if an officer remembers having a conversation with three kids, maybe the system would indicate with a note that there were three people on camera at 2 hours and 20 minutes into his shift. This way the officer would know this is likely the incident they want to review and they know exactly how far to fast forward to view it. Maybe they want to review an encounter they had with someone with brown hair. If they talked to many individuals throughout their shift, they could save time by just fast forwarding to spots where brown hair is noted. Or perhaps, they could see where traffic stops vs. pedestrian encounters are in the video, if the system could recognize when an automobile is present. In my opinion, this is all very benign, because it doesn’t identify the people at all, nor does it really implicate them in any way. It’s just a pretty neat and helpful feature of the software.
I feel Tier 2 still does not have great concern, because it still can be considered a narrow, targeted search. It could prove to be a valuable tool in cases where there is reasonable cause to believe a crime was committed. To give an extreme example, if an officer had been wearing a body-worn camera with these kind of bio-metric capabilities during the Boston Marathon bombing that happened to catch a suspect’s face, we would definitely would have wanted the police to perform this type of facial database search. In our communities, we should be making important decisions about whether doing so should require a judge’s warrant. Personally I think this should always be the case, but at the very least, there should be very specific department policies about who can perform these searches and in what circumstances.
In my opinion, Tier 3 is the no-go level of bio-metrics. This level of facial recognition technology creates what we call a ‘perpetual lineup’. Any and every person would be considered a possible suspect at any time they were in public in view of an officer. At this level of efficiency, we start to tilt the balance between freedom and safety too far the other way. Tier 3 bio-metrics breaks the spirit of the Fourth Amendment to particularly describe who and what you are looking for. In a scenario like this where everyone is always processed, we create what we call a dragnet, which is contrary to life in a free society. This shouldn’t be allowed in America.
As we make these considerations as a community and as police departments in forming policy, we must also consider the designs and business interests of the vendors that provide this technology. We must not only decide what we as communities and police officers can and can’t do, but also what we are not willing to let the technology providers do. This is especially important considering how interoperable and networked these systems are able to be. While we may choose to not allow our officers to perform automatic searching, we must make sure that the systems don’t give access to other entities to make use of our data in this automatic fashion, through data-sharing we might not be fully aware of.
As we see tech manufacturers like Axon submitting patents for Tier 3 bio-metric systems, we cannot allow our future to be determined merely by what the police are able to do, but rather what they should do. It’s up to communities they serve to make these determinations.