Search papers, labs, and topics across Lattice.
This paper conducts the first comprehensive audit of memorization in mobility prediction models, revealing significant privacy risks associated with the retention of sensitive user trajectories from training data. By introducing a novel framework that quantifies memorization at various granularities, the authors systematically assess how user behavior and trajectory structure contribute to these risks. The results indicate that memorization patterns are prevalent and correlate with user regularity, highlighting the urgent need for privacy audits in this domain.
Pervasive memorization in mobility prediction models can expose sensitive user trajectories, necessitating urgent privacy audits.
Human mobility prediction models, which forecast the next location in a user's trajectory, are increasingly deployed in urban analytics, navigation, and personalized services. Yet, little is known about their potential to memorize and expose sensitive user trajectories from training data. While memorization has been extensively studied in language models, mobility prediction poses unique challenges: training sequences encode human behavior at various spatial and temporal scales, creating privacy risks at different granularities. In this paper, we conduct the first systematic audit of memorization in mobility prediction models. While prior work has shown that privacy leaks can arise from such models, we systematically assess and quantify memorization risks at scale. We identify key challenges, including the lack of a randomness space, the multi-scale structure of trajectories, and user-specific behavioral diversity. To address these challenges, we introduce a framework to quantify mobility memorization at different levels of granularity: individual locations, anchor pairs, and subtrajectory segments. We also develop user-grounded reference sets to assess how likely a model is to prefer training data over realistic alternatives. Our evaluation across multiple models and datasets reveals pervasive memorization patterns that correlate with user regularity and increase the risk of data extraction at inference time. Our findings call for mandatory privacy auditing in mobility prediction models.
