Even Artificial Neural Networks Can Have Exploitable ‘Backdoors’ Posted on August 25, 2017 by Malicious machine learning can hide nasty surprises. Read the original post by Tom Simonite