A Privacy-Preserving Cross-Platform Photo Intelligence and Travel Assistance System Using Federated Learning

Abstract

The exponential growth of personal digital photographs and the increasing concerns regarding data privacy have necessitated the development of intelligent photo management systems that operate with strong privacy guarantees. This paper presents a comprehensive privacy-preserving cross-platform photo intelligence and travel assistance system that combines on-device deep learning, end-to-end encryption, federated learning, and location-aware contextual services. The proposed system performs real-time face detection and clustering using YOLOv8 and MobileFaceNet, implements secure storage through AES-based encryption with XChaCha20-Poly1305, and enables collaborative model improvement without transmitting raw user data. Furthermore, the system integrates semantic search capabilities using MobileCLIP, automated journey organization through geo-spatial clustering, and travel assistance features that provide location-based contextual information while maintaining strict privacy preservation. The system has been deployed across Android, Windows, and Web platforms, achieving efficient real-time performance with latency below 100 milliseconds for face detection while maintaining zero-knowledge security guarantees. Experimental evaluation demonstrates that the proposed system achieves robust face detection accuracy of 95.2% and successful clustering precision of 93.8% while preserving complete user data privacy.