Amazon has taken a lot of heat from much of the tech press over its various facial recognition efforts. Most recently, this has come in the form of articles raising alarm bells about a patent application from its Ring subsidiary about using facial recognition in its doorbells to identify suspicious people. The problem is, those articles miss the point that all the major players in camera-equipped doorbells, starting with Google’s Nest, are already implementing facial recognition in their doorbells and other devices. So, as the saying goes, “the cat is out of the bag.”
What actually makes Ring’s application interesting, and possibly even more spooky, is that it involves using composite images taken from multiple devices — possibly from a variety of different properties — to recognize a face. If you thought Amazon’s flagship Rekognition product was flawed, this approach, if it is ever implemented, has even more potential to go awry.
Familiar Faces Is the First Step to Suspicious Faces
Current facial recognition offerings marketed to home users focus on identifying individuals who can then be tagged as “familiar.” Family members, friends, and regular delivery people are obvious candidates. Whether or not everyone in your family thinks it’s cool to have a complete record of their comings and goings, this seems like a fairly benign way to filter out alerts so that you only get a text or push notification when someone who isn’t on your list approaches your door. However, it isn’t a very big leap to go from that to flagging suspicious faces. For example, cameras at our house captured video of prowlers skulking up to our door and looking around late one night. If we’d had the option, I would probably have been happy to mark them as “suspicious” and get an alert if the camera saw them again.
Ring’s Neighborhoods Are Powerful, Even Without Facial Recognition
We don’t have Ring cameras at home, but several neighbors do. The night the prowlers were on our property, they also scouted out several houses with Ring doorbells. Those neighbors shared the videos instantly via Ring’s Neighborhood feature, which is actually the single most unique and clever thing about Ring’s doorbell. It provides a great way for neighbors to alert each other about suspicious activity. Videos can also be sent directly to local law enforcement. That already happens today. Adding facial recognition that compares individuals with a “suspicious” list seems like a logical and probable next step unless some type of regulation intervenes (for example, Illinois privacy law precludes the use of Ring’s facial recognition system).
Ring’s Patent Is Both Spooky and Clever
From experience, I know that each individual security camera recording may only show certain portions of a person’s face, or their clothing, or how they walk. Looking at footage from several cameras can help to put together a more complete picture. Ring’s patent application describes taking that to the next level, where images from different cameras on different properties can be combined and facial recognition applied to them in aggregate.
Now, I’m really skeptical that this is going to be practical, at least any time soon. But whether they are using facial recognition or gait recognition, there is no question that Amazon, Google, and others will network their home security cameras and pool information if they’re allowed to. There is also no question that consumers will purchase the capability if it is offered. So the big question will be how and whether other venues take steps like Illinois to limit its use.
Russian Scientists Arrested for Mining Cryptocurrency at Nuclear Facility
Several Russian scientists working in one of the country's most secure research facilities thought they could use the in-house supercomputer to mine some coins.
Facing Epyc Onslaught, Intel Will Aggressively Compete in Servers
Intel expects AMD to come gunning for its servers later this year and there's no way Intel gives up market share without a fight.
Amazon Tried to Sell ICE Its Faulty Facial Recognition Tech
Amazon was quite eager to sell its Rekognition facial recognition software to ICE this year, despite third-party tests showing its accuracy problems when used to identify non-white, non-male individuals.
ICE Used Facial Recognition to Scan Millions of Driver’s License Photos
According to a new report, ICE has sought access to millions of driver's license photos in order to search for undocumented immigrants with facial recognition technology. At least three states complied with this futuristic dragnet.