Over the past years, clients have largely depended on and trusted Automated Teller Machines (ATMs) to fulfill their banking needs and control their accounts easily and quickly. Despite the significant advantages of ATMs, fraud has become a very high risk and danger. As it leads to controlling all clients' accounts. In this paper, the proposed framework is using the iris recognition technology combined with the one-time password (OTP) to detect and prevent the known as well as the unknown attacks on ATMs and provide a table of the attackers and the suspected attackers with a counter to take a preventive action with them. Our proposed preventive actions are: card withdrawal, flagging the identified iris as an attacker in the database, notifying the card owner with this suspicious behavior, reporting to the Central Bank of Egypt (CBE), and calling the police when an attacker's iris counts three capturing times, even if for a different card. Two case studies were attempted to achieve the highest accuracy, the first case was using the Chinese Academy of Sciences' Institute of Automation V1.0 (CASIA-IrisV1) dataset using the Cosine Distance. The second one was using the Indian Institute of Technology Delhi (IITD) dataset using k-Nearest Neighbors (KNN) and Histogram of Oriented Gradient (HOG) techniques together reaching 100% accuracy.