Re:The Secretive Business of Facial-Recognition Software in ...
by tartoran on Monday October 22, 2018 @10:57PM
(#18280410)
It's actually a hard problem that people who do this encounter and seems impossible to get right. We actually can tell if a friend is walking in the distance by looking at gait+clothes but could make mistkes as well, it's not 100% accurate. For surveillance reasons it's much easier to get 100 accuracy with just tagging people with microchips and that'a that. Why would we want to excel at gait-fingerprinting people?? That's a scary thing to want to perfect. Very Science fictiony and hopefully not prevalent in the future
The first thing I would do is gather a bunch of data for people walking trying to mess up their gait. Different types of shoes, pebbles in shoes etc. Then include that in the model or even just look for that as suspicious itself.
You're not giving enough credit to the capability of the people that do this kind of work. It is not simple to avoid and the vast majority of people would not even go this far.
How many people communicate entirely over encryption today?