Attenuation YOLOv5-based detection of anomalous behavior in ATM

The extensive deployment of Automated Teller Machines in current times has led industries to develop state-of-the-art security models that track suspicious activities. The presented research details the development of an anomaly detection system for ATMs which operates in real-time by implementing the attenuation-enhanced YOLOv5 deep learning model. The combination of computer vision with machine learning techniques makes the system perform automatic threat assessments to identify skimming activities forceful ATM withdrawals unauthorized access events and loitering behaviors in surveillance feeds. YOLOv5 out-performs object detection yet its precision rises even more through attenuation mechanisms which remove environmental noise together with background clutter. The model receives training from a wide-ranging ATM anomaly database that includes both virtual and operational ATMs which ensures its capability to address diverse situations. The experimental finding shows that detection methods with YOLOv5 achieve better results in terms of precision and recall while producing less false positives. The system supports instantaneous processing and activates prompt security warnings through speed-enabled alerts. The research develops intelligent surveillance capability by implementing improved technology for ATM anomaly detection. Research into future development will combine reinforcement learning models with multiple sensing devices to enhance operational adaptability across various settings.