Audio fingerprinting and AudioContext study

Having a quick surf around this morning on the bus, I came across this post on Techcrunch about a study from Princeton University into online trackers.

They were able to create a fingerprint of a machine using the AudioContext API that allowed for the tracking of machines.

As the article mentions, online privacy will be an ongoing race and there is a long tail of tools that blockers may not find or block. Is machine learning the way but then where does one strike the balance between blocking trackers and stopping sites or tools working?

It may be an education and information battle. This a use of AudioContext that I had previously not considered. May be it is something to look into further as part of my own research.