Attention mechanism with CNN for fingerprint and finger-knuckle based person authentication

Biometric authentication systems play an important role in maintaining security and privacy in all areas. This research paper introduces a method of authentication by combining unique characteristics of the CASIA-FingerprintV5 fingerprint and the PolyUKnuckleV1 knuckle images. To capture the textural features found in both types of biometrics, we used local binary pattern (LBP) features. We utilize a Convolutional Neural Network (CNN) to classify data generated by our proposed approach by concatenating the features of the fingerprint and the finger-knuckle to allow us to assess its effectiveness. The attention mechanism is implemented to focus on the most relevant features to prioritize the most informative part of the image. The results of our experiments indicate that the model achieved near-optimal performance, approaching its maximum potential. The outcome highlights the effectiveness of our fusion technique in enhancing the capabilities of biometric authentication systems, particularly when individual modalities may have limitations. The fusion of fingerprint and knuckle biometrics improves the resilience of the system against challenges such as sensor variations and noisy input data. This research shows promise for applications that require reliable biometric authentication, such as access control, identity verification, and forensic investigations.