Characterize the privacy-utility effects of combining sensing modalities

Characterize how combining multiple sensing technologies affects the privacy-utility trade-off of smart-city sensing systems, including whether multimodal combinations increase identity inference risk relative to individual sensors while preserving or improving utility for activity and attribute inference.

Background

MultiGait synchronizes recordings from eight sensing technologies and multiple perspectives, enabling research on multimodal fusion and cross-sensor recognition. Existing recognition systems generally rely on a single information source, such as video from one perspective.

The paper identifies the consequences of combining sensors as unresolved. Understanding these effects is necessary for selecting sensor combinations that achieve a required utility objective while minimizing biometric privacy risks, potentially in conjunction with anonymization methods deployed at the edge.

References

As current recognition systems mostly rely on a single information source (e.g., only video cameras from one perspective), it remains largely unknown how their combination affects the privacy-utility trade-off.

— MultiGait: A Multi-Sensor Multi-Perspective Multi-Session Biometric Inference Benchmark and its Dataset  (2609.01036 - Todt et al., 1 Sep 2026) in Section 6.1, Implications

Three-dimensional joint coordinates omit appearance information, and their identifiability relative to full-color video has not been tested here.

— Contactless gait-derived features as candidate digital biomarkers of depression severity in adolescents  (2610.18510 - Liu et al., 13 Oct 2026) in Discussion, paragraph beginning “A skeleton-based workflow allows repeated assessment.”