Cognitive Surveillance Infrastructure for Secure Identity Tracking in Urban IoT Systems

This paper proposes a cognitive surveillance infrastructure for secure identity tracking in smart cities, emphasizing person-centric tracking and ethical use of IoT data. The system deploys edge devices (e.g., smart cameras) that perform local face detection and multi-feature encoding (deep neural embeddings, perceptual hash, and ORB keypoints). Each detection is tagged with spatial–temporal metadata (time, GPS coordinates, camera ID) and securely transmitted via encrypted FTPS to a central server. The server maintains a database of face features and executes long-term spatio-temporal matching to associate identities across cameras and time. We validate the approach on a public face dataset, demonstrating high recognition accuracy (over 90%) and substantial bandwidth savings compared to raw image transfer. Our evaluation reports face recognition performance, bandwidth usage, and latency under varying load, indicating scalability to city-scale deployments. The proposed design balances public safety objectives (e.g., missing-persons tracking) with privacy-preserving measures and complies with emerging ethical guidelines for smart-city surveillance.