A Comprehensive Review on Federated and Incremental Learning in Spatial Crowdsourcing Systems

Abstract

Spatial Crowdsourcing (SC) coordinates workers and requesters through personal mobile devices to complete location‐dependent tasks in real time. The combination of continuous worker mobility, self‐interested participants, and real‐time service requirements creates a learning environment that is more adversarial, more dynamic, and more resource‐constrained than the settings addressed by existing federated learning (FL) surveys. FL preserves location data privacy but remains vulnerable to inference, poisoning, Byzantine, and Sybil attacks, whose consequences in SC directly affect task allocation, pricing, and worker safety. Incremental learning (IL) addresses the continuous concept drift caused by mobility and demand shifts, enabling model adaptation without full retraining on resource‐constrained devices. Despite their complementary roles, FL and IL have largely been surveyed in parallel, and their integration within SC has not been systematically addressed. This survey analyses 172 primary studies published between 2019 and 2026. It contributes: an SC‐focused assessment of FL and IL techniques identifying where non‐SC results do not transfer; a connected taxonomy linking SC‐specific threats to evaluated countermeasures and their trade‐offs; a conceptual FL‐IL framework that jointly addresses privacy, security, and continual adaptation; a review of available datasets and open‐source tooling with identified benchmark gaps; and a prioritized research agenda on SC‐specific assumptions and deployment conditions that require further investigation. The survey is intended to guide researchers and practitioners in developing secure, adaptive, and resource‐efficient SC systems. This article is categorized under: Technologies > Crowdsourcing Technologies > Machine Learning Commercial, Legal, and Ethical Issues > Security and Privacy