Movement data you can share without giving people away

Digital services on the web and in smart-city infrastructures increasingly run on behavioural and movement data. A handful of data points is often enough to single out an individual, so classical anonymization reaches its limits quickly — and the uncertainty this creates keeps useful data locked away.

SynTrace develops methods for anonymizing and generating movement data that combine strong privacy guarantees with the data quality real applications need. The core idea is to pair differential privacy with generative models, so that synthetic data reproduces the statistical structure of the original without permitting inferences about individuals. Public, non-personal data — anonymous page-visit counts, road networks, induction-loop readings — is used as conditional information to keep rare events realistic, which is where purely private generators tend to fail.

A second line of work protects trajectories while they are still being streamed: hybrid mechanisms perturb some events and replace others with synthetic segments, removing the link to a person at the interface between the party collecting data and the party processing it. The resulting methods are integrated into an existing smart-city data platform as an anonymization API and evaluated in realistic scenarios.

SynTrace continues the work of two predecessor projects, SynthiClick (web click traces) and PROPOLIS (smart-city analytics), and brings both strands together at the point where data changes hands.