Communications of the ACM
Oregon's a Testing Ground for Amazon's Facial Recognition Policing, But What if Rekognition Gets it Wrong?
Opponents warn police reliance on Amazon's face-scanning Rekognition software could lead to wrongful arrests of individuals who only faintly resemble video images, as well as threatening citizens civil liberties and safety.
Credit: Stegerphoto/Getty Images
Police in Washington County, OR, have used a facial-recognition algorithm from Amazon to publicly test new surveillance techniques for more than a year, accelerating criminal investigations based on video footage, but also intensifying debate about the tool's legal ramifications.
Opponents warn police reliance on Amazon's face-scanning Rekognition software could lead to wrongful arrests of individuals who only faintly resemble video images, as well as threatening citizens’ civil liberties and safety.
Rekognition's advantages include easy activation, no need for significant technical infrastructure, and affordability.
Although Amazon advises police to use the algorithm's results only when matches have 99% confidence, Washington County deputies are not shown that metric when they employ Rekognition, receiving instead five possible matches for each search, even if the software's match confidence is much lower.
From The Washington Post
View Full Article - May Require Paid Subscription
Abstracts Copyright © 2019 SmithBucklin, Washington, DC, USA
No entries found