And you leave fingerprints on everything you touch. But, it's all we've got as far as hard evidence aside from testimony and camera recording.

Basically all we have is camera recordings because 'eye-witness' testimony is notoriously bad.

Machine Learning will put a stop to that there are several project right now working completely fabricate audio and video in realistic ways...

Yep check these guys out: https://www.theregister.co.uk/2017/04/24/voice_stealing_lyre...

we can already do video (pretty well, anyway) : http://www.graphics.stanford.edu/~niessner/thies2016face.htm...