Person re-identification, often called re-ID, is the process of matching the same person across different cameras, locations, or time periods. Unlike face recognition alone, re-ID can use clothing, body shape, gait, carried objects, and other visual cues when a face is unavailable or obscured. It is a core capability in modern surveillance and analytics pipelines.
In cyber security, re-identification matters because it turns scattered video feeds into a trackable identity graph. Defenders may use it for security monitoring, access control, or incident investigation, while attackers or privacy-invasive systems may use it to follow people without explicit consent. Re-ID systems are also a common target for adversarial machine learning: changes in clothing, occlusion, lighting, camera angle, or sensor quality can reduce matching accuracy. That makes robustness testing important, especially in camera-dense environments where surveillance claims often depend on how well the model handles real-world variation.