Detecting the Undetectable: Deep Learning Model for Identification of Fake Images

Deep learning (DL) has brought big change in the field of image processing, allowing the generation of fake images, commonly known as Deepfake images. Because of the rapid advancements in generative models, it become difficult to differentiate between fake and real images. This research finds out the effectiveness of DL based Convolutional Neural Networks (CNNs) for the identification of deepfake images. In this study, the performance of three CNN models namely VGG16, VGG19, and InceptionV3 were employed, to evaluate their effectiveness in detecting fake images. The main objective of the study is to enhance detection accuracy by using robust feature extraction capabilities of these models and optimizing their performance through fine-tuning. Experimental results demonstrate that VGG16 provides the highest accuracy of 94% for detecting fake images. In this research, we utilize and compare the performance of three CNN models i.e., VGG16, VGG19, and Inception V3 to analyze their ability to detect fake images. The primary aim of the research is to improve the accuracy of detection using strong feature extraction abilities of these models and fine-tuning their performance. Experimental results indicate that VGG16 offers the best accuracy of 94% in detecting fake images.