The current state of the art in anonymization is Differential Privacy [1], and its application to location data is exactly as you suggest: quantize each point,

The current state of the art in anonymization is Differential Privacy [1], and its application to location data is exactly as you suggest: quantize each point, introduce noise, and then aggregate. The thing is that you can choose the amount of noise you want to introduce to trade off privacy vs. accuracy, making it not all-or-nothing. [1] https://en.wikipedia.org/wiki/Differential_privacy https://en.wikipedia.org/wiki/Differential_privacy