Federated Swarm Intelligence for Privacy-Preserving Mobile Collaborative Learning

Counts differ because each database indexes a different set of publications. We treat OpenAlex as the canonical count; Google Scholar is not shown (no API, and crawling it violates its ToS).

The increase in mobile learning escalates the friction between data-led personalization and safeguarding sensitive, institutionally fragmented student data. Centralized analytics necessitates the aggregation of raw records, which conflicts with privacy regulation and mobile resource limits. A novel federated swarm-intelligence framework for mobile collaborative learning to train a student-outcome prediction model using simulated devices without any raw data transmitted from them. A three-tier structure where cloud, edge and mobile interfaces work together is proposed, along with a particle swarm optimization (PSO) controller for energy-aware and latency-aware client selection. Further, we deploy differential privacy via DP-SGD with a formal accountant and employ membership-inference audit. Federated training obtained in all simulations accuracy (AUC up to 0.91) equal to that of centralized while a PSO selection strategy achieved 40–46% reduction in communication rounds and 73% less energy use compromising accuracy only marginally. Differential privacy provides a certifiable guarantee (ε ≈ 1-8) at low cost.