Unusual behavior identification in ATM using YOLO attenuation modelling

The rising popularity of new ATMs has made them prime targets for criminals so reliable security systems must identify unusual ATM user interactions. The research develops an automated detection system based on YOLOv5 deep learning model which uses attenuation methods for real-time security event monitoring at ATMs. The video surveillance analysis tool uses a combination of computer vision and machine learning technologies to detect skimming events directly and identify forced withdrawals and unauthorized entries and loitering actions while operating at speed using YOLOv5 architecture that performs background attenuation for removing nonessential environmental details to improve recognition abilities. A large collection of security incident data consisting of real-world events and artificially produced data supports the system during training processes enabling optimizations for different operational environments. The model-based research reached excellent precision and recall levels while effectively cutting down traditional method-related inaccuracies. Processing systems with real-time capabilities provide security staff with instant alerts that let them trigger automated responses in a timely manner. The framework introduces innovative observations to ATM security through its design of anomaly detection frameworks. Researchers plan to combine reinforcement learning technologies with their model to improve complex operational scenario performance through sensor data fusion capabilities.