Abstract-This paper introduces an intelligent surveillance system aimed to track CCTV footage in real-time at public locations, like malls and train stations, in order to identify suspicious activity. The system consists of several parts, such as relationship modeling for understanding contextual relationships, object detection trained on the frames of suspicious activity. The technology compares relationships found in live video with pre-defined suspicious relationships to identify possible threats and triggering alerts. Security personnel are given a visual tool for effective threat identification when a graphical representation is generated, emphasizing relationships linked to suspicious activities. By integrating object recognition, relationship modeling, and graph-based representation for efficient anomaly detection in crowded environments, the suggested framework promises to transform public safety.