ATM Security System Modeling Using Face Recognition with FaceNet and Haar Cascade

ATM or Automated Teller Machines are widely used by people nowadays. The existing conventional ATM is vulnerable to crimes because of the rapid technology development. A total of 270,000 reports have been reported regarding credit card fraud and this was the most reported form of identity theft in 2019. A secure and efficient ATM is needed to increase the overall experience, usability, and convenience of the transaction at the ATM. To provide better security for the conventional ATM, this paper proposed a face recognition system using FaceNet combined with Haar Cascade Classifier. Haar Cascade is used to detect the haar features based on the face features and stored to the FaceNet model as the face recognition model. The scale factor from the Haar Cascade is also modified a few times to get a better accuracy and processing time. With the proposed method, the Personal Identification Number (PIN) will be replaced by face recognition. Every time the customers insert their card, the system will detect and start to identify the face. If it does not match, the card will be blocked. This research uses the Face Recognition Dataset from Kaggle. By using this dataset, the proposed system achieves the highest accuracy with 90.93%.