Privacy-protected statistics publication over social media user trajectory streams

摘要: Abstract An increasing amount of user location information is being generated due to the widespread use social network applications and ubiquitous adoption mobile wearable technologies. This data can be analysed identify precise trajectories individuals — where they went when were there. an obvious privacy issue, yet publication real-time aggregate over such streams provide valuable resources for researchers government agencies, e.g., in case pandemics it would very useful who might have come into contact with infected individual at a given time. Differential techniques become popular widely adopted address concerns. However, there are three key issues that limit application existing differential approaches trajectory data: (a) heterogeneous nature trajectories, (b) uniform sliding window mechanisms do not meet requirements (c) limited budgets impact on utility applied infinite streams. To tackle these problems, this paper proposes private stream statistics mechanism utilizingdifferential (DP-PSP). relieve heterogeneity issues, anchor point discovery (e.g., fixed locations like museums, parks, etc.) road segmenting proposed. We adaptive w-step approach allows users specify their own dynamic budget distribution optimize budget. preserve utility, we present multi-timestamp prediction models k-nearest neighbour selection perturbation algorithms reduce distortion induced through mechanism. Comprehensive experiments real-life location-based show DP-PSP provides boosts quality perturbed aggregation without compromising privacy.