Abstract :

Artificial Intelligence (AI) is transforming the security, reliability, and user experience of banking technologies. Traditional ATM systems rely mostly on PINs and cards, which are increasingly susceptible to fraud, skimming, and unauthorized access. This project proposes a third-generation ATM enhanced with AI-based multimodal authentication including facial recognition, behavioral analytics, and anomaly detection, ensuring more secure and userfriendly transactions. By integrating deep learning models for real-time user verification and fraud detection, the system significantly reduces risk while improving transaction efficiency. Experimental results demonstrate high accuracy and improved security compared to conventional ATM methods. This research not only highlights the feasibility of AI integration in ATM systems but also provides a roadmap for scalable and robust future deployments. The proposed methodology supports realworld implementation with edge computing and cloud component