Privacy-Preserving Federated Incremental Learning for Spatial Crowdsourcing: A Survey of Challenges and Methods

Spatial Crowdsourcing (SC) is an emerging crowdsourcing paradigm that allocates tasks based on real-time user locations to perform spatiotemporal relevant tasks. However, relying on continuously updated location and personal preferences data introduces significant privacy and security challenges. Federated Learning (FL) and Incremental Learning (IL) present innovative solutions to address these challenges by enabling adaptive learning while preserving user privacy. FL allows for collaborative model training without sharing raw data with central servers, and IL facilitates continuous model updates without requiring retraining. Despite FL's widespread use in SC, IL integration remains unexplored. This survey examines the potential of combining IL with FL in SC systems, highlighting the benefits of this integration, including enhanced adaptability and privacy preservation. We review existing methodologies, discuss challenges, and aim for future research directions to advance the privacy-preservation of SC systems.