Citizen science and geoprivacy: Empirical analysis of location masking in large-scale crowdsensing networks

Crowdsensing that relies on individuals and communities to deploy low-cost sensors for data collection has become an increasingly important source of socio-environmental data. However, the locations collected through these networks also raise geoprivacy concerns. This study provides one of the first large-scale empirical examinations of location masking, a common privacy protection behavior where the reported sensor location is intentionally displaced from the true location, with the aim to address a current critical gap in understanding how such behaviors occur “in the wild” within large-scale, real-world crowdsensing networks. We leverage a large national dataset of PurpleAir sensors and apply ordinal logistic and mixed-effects models to examine how the immediate sensor placement and broader neighborhood characteristics shape location masking. Our results show that sensors placed indoors, in non-urban, or in non-residential areas tend to exhibit high levels of masking, meaning their reported locations deviate further from the true locations. In contrast, sensors in neighborhoods with high educational attainment, income levels, older populations, and larger proportions of non-white and Hispanic residents are associated with low levels of location masking. These findings indicate the importance of considering both the physical sensor placement and socio-spatial context in shaping privacy-related behaviors and suggest that such factors should be carefully considered when designing and promoting crowdsensing initiatives.