Artificial rabbits optimization with transfer learning based deepfake detection model for biometric applications

Deepfake detection is a significant area of research in biometric applications, as it is essential to ensure the integrity and authenticity of biometric information. Biometric data, including fingerprint recognition, facial recognition, and voice recognition, are used extensively for identification and authentication, which makes it crucial to prevent and detect deepfake attacks. Deepfake is a manipulated digital media, for example, a video or image of a person can be replaced with a similarity of another person. A crucial way to deepfake detection in biometric applications is to use a machine learning (ML) algorithm, particularly deep learning (DL), that could learn to distinguish between fake and real biometric information. Hence, the study proposes an Artificial Rabbits Optimization with Transfer Learning Deepfake Detection for Biometric Applications (AROTL-DFDBA) technique. The AROTL-DFDBA technique intends to detect fake and original biometric data using the DL model. In the presented AROTL-DFDBA technique, modified DarkNet-53 model for the feature extraction process. Besides, the ARO method was applied for the optimum hyperparameter selection of the modified DarkNet-53 model. For deepfake detection, the Weighted Regularized Extreme Learning Machine (WR-ELM) technique is applied. The simulation outcomes of the AROTL-DFDBA method can be validated on the DeepFake dataset. The extensive simulation results signify better detection outcomes of the AROTL-DFDBA technique over other existing techniques with a maximum accuracy of 96.48%.