There is a continuous and ubiquitous collection of precise, timestamped, geolocation data from apps and devices, being amassed by private firms. This data collection can be considered an ongoing “population survey” and can serve many of the same uses as traditional government travel surveys. Instead, these “location based services” (LBS) datasets are used for advertisement targeting, company analytics, and other means for private profit. This work is about leveraging LBS data to benefit the public from whom it is sourced.
However, the utility of these datasets must be balanced with the privacy of individuals within them.
This work focuses on protecting the privacy of individuals within location datasets. To do so we develop an approach with machine learning deep neural networks to produce synthetic data that has the same attributes as the original data in aggregate, but sufficiently varies at the individual level in order to protect user privacy.