AlertAura -Privacy preserving federated learning for women’s safety risk prediction

Women's safety remains a critical societal challenge, especially in urban and semi-urban environments where existing mobile safety applications largely rely on reactive mechanisms such as emergency alerts and panic buttons.While these solutions provide post-incident support, they fail to predict potential risks in advance and often require centralized collection of sensitive personal data, raising serious privacy and trust concerns.To address these limitations, this paper proposes AlertAura, a privacy-preserving artificial intelligence framework for women's safety risk prediction based on Federated Learning.AlertAura evaluates experiment using simulated federated clients demonstrate that the global model converges effectively over multiple learning rounds, achieving improved risk prediction performance without centralized data storage.The AI framework ensures the feasibility of deployment intelligent, privacy-aware safety systems in sensitive real-world applications.By combining proactive risk assessment with privacy-by-design principles, AlertAura represents a significant step toward ethical, trustworthy, and scalable women's safety technologies.This decentralized learning approach allows the system to continuously improve predictive accuracy while maintaining strong privacy guarantees.The proposed model analyzes locally extracted features related to movement behavior, temporal patterns, and environmental context to generate proactive safety risk scores.