A Context-Aware Hybrid Local Differential Privacy–Federated Learning Framework for Digital Footprint Protection

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).

Digital services increasingly depend on information generated from users’ online activities, yet the same traces that make these services useful can expose individuals to re-identification and inference. This study develops and evaluates a context-aware Hybrid Local Differential Privacy–Federated Learning (LDP-FL) approach for protecting digital footprints while retaining useful machine-learning performance. The work distinguishes digital footprints into active, passive and hybrid categories and assigns privacy protection according to feature sensitivity rather than applying a single noise level to every attribute. The proposed pipeline perturbs sensitive features locally with a Laplace mechanism, clips local gradients, and uses Federated Averaging to train a global model without centralising raw client data.