Search

Recently, the massive development in the field of deep learning (DL) and artificial intelligence (AI)-aware tools have forced the requirement of caution when using several types of digital data. Serious security and privacy concerns have risen due to the significant advancements in the creation of the manipulated technique known as deepfakes. One avenue of deepfakes is to add and eliminate tumors from medical images. The inability of the automated systems to detect medical deepfakes can cause serious security and privacy problems resulting in an extensive burden on hospital assets or even loss of human life. To counter such effects a reliable deepfakes detector that can tackle the latest manipulation generation approaches is required. In the presented work, we attempt to solve the problem by introducing a DL method called the MedNet model to detect lung CT-Scan-based deepfakes samples. Descriptively, we have proposed a custom EfficientNetV2-B4 framework with extra added dense layers at the last of the network. To further increase the feature computation ability of the introduced approach, we proposed a spatial-channel attention mechanism to emphasize the altered areas of samples which result in improved classification performance. Extensive experimentation containing a standard dataset called the CT-GAN dataset is performed to show the efficiency of the presented work. We have attained an accuracy score of 85.49% which is showing the effectiveness of the presented work in reliably detecting the real samples of lung CT-Scan from the deepfake images. [ABSTRACT FROM AUTHOR]

Recently, the higher availability of multimedia content on social websites, together with lightweight deep learning (DL) empowered tools like Generative Adversarial Networks (GANs) has caused the generation of realistic deepfakes. Such fabricated data has the potential to spread disinformation, revenge porn, initiate monetary scams, and can result in adverse immoral and illegal societal issues, etc. Hence, the accurate identification of deepfakes is mandatory to discriminate between real and manipulated content. In this work, we have presented a DL-based approach namely a unified network for FaceSwap (FS) and Face-Reenactment (FR) Deepfakes Detection (AUFF-Net). More clearly, both the spatial and temporal information from the video samples are used to detect two types of visual manipulations i.e., FS and FR. For this reason, a novel DL framework namely the Inception-Swish-ResNet-v2 model is introduced as a feature extractor for computing the information at the spatial level. While the Bi-LSTM model is utilized to measure the temporal information. Additionally, 3 dense layers are included at the last of the model structure to suggest a discriminative group of the feature vector We performed extensive experimentation on a challenging dataset namely the FaceForensic++, and attainede average accuracy values of 99.21 %, and 98.32 % for FS, or FR, respectively. Furthermore, we introduced an explainability module to show the reliable keypoints selection capability of our technique. Moreover, we have performed a cross-dataset evaluation to show the generalization power of our approach. Both the qualitative and quantitative results have confirmed the effectiveness of the suggested approach for visual manipulation categorization under the occurrence of various adversarial attacks. • A novel AUFF-NET model is proposed for FaceSwap, and Face-Reenactment deepfakes detection. • Accurate detection of deepfakes due to the ability of the model to tackle the model over-fitting. • Improved model explainability power due to the reliable feature selection ability of model. • Robust model due to its ability to detect deepfakes in the presence of various adversarial attacks. • Perform well to unseen examples due to the better generalization power of the AUFF-NET model. [ABSTRACT FROM AUTHOR]