What does it take to re-identify 100,000 “anonymous” users? In our latest study, we show that anonymized human mobility data isn’t as safe as it seems. By leveraging subtle patterns in location density, movement structure, and temporal activity, we were able to reverse-engineer real trajectories of users at a national scale in Japan.

Our findings demonstrate that even carefully anonymized datasets can retain enough signal to compromise user privacy. This work calls for a serious rethink of how we release and protect spatio-temporal data in the public domain.