Hacker News Reader

I've read something about a method used by online banking services some time ago: They tracked the way the customer moves their mouse, calculated their arm lenghts etc. and when the movements were suspiciously different, the system assumes it's a fraud and logs them out.

View on HN

I did similar project with Tensorflow. Mouse movements were captured then converted to images and DL model was trained to classify user or not user.

It can also classify classes of users I.e. new portal users are moving mouse differently from users who are familiar with portal.

To add - by itself it’s not a reliable indicator of yes/no.

But rather another risk scoring input to overall identity detection system.

This is actually clever. Thanks for sharing the technique!

Pleasure :) Here's writeup:

https://www.splunk.com/blog/2017/04/18/deep-learning-with-sp...

Reply on news.ycombinator.com