Augmenting ATM Security: Real-Time Facial Emotion Recognition as an Additional Safeguard

Motion detection and facial recognition fields have undergone thorough research to enhance surveillance services in locations that need accurate motion detection and human identification. With the continuous improvement of criminal strategies, it is essential for surveillance technologies likewise to progress accordingly. In the present literature, MD&FR stands out as a dependable surveillance technology. Although there is potential to integrate MD&FR technologies into existing surveillance systems (SS), the present systems have only implemented MD-based and FR-based SS separately.This research investigates integrating Facial Expression Recognition (FER) technology to bolster the security of Automated Teller Machines (ATMs). FER technology holds potential as an additional security layer, capable of detecting and responding to emotional cues in realtime. Through a thorough exploration, three different models—LSTM, CNN, and CapsNet—were evaluated for their effectiveness in classifying facial emotions. Results indicate that the CNN model outperformed LSTM and CapsNet, exhibiting higher accuracy, precision, recall, and F1 score. These findings underscore the CNN model's potential to enhance ATM security by accurately recognizing facial expressions. However, further research is needed to address challenges and refine FER models for optimal performance in real-world scenarios. This study contributes valuable insights into leveraging FER technology to bolster ATM security, ultimately enhancing user safety and mitigating fraudulent activities.