Recently, the surge in the popularity of photography has been driven by the widespread use of cameras. Photographs, which are crucial to our daily lives, are packed with information and can often be altered to reveal even more. While many tools exist to enhance photo quality, they are frequently employed to edit images, leading to the spread of misleading information. This trend has contributed to the increased prevalence of photo forgeries that is a significant problem today. Traditionally, images have been manipulated through techniques such as copy-move or image-splicing. Machine learning techniques is used to detect these types of forgeries. Recently, Generative Adversarial Networks are used to create a new kind of forgery known as deep-fake images, which are more harmful due to their realistic appearance. Various approaches for detecting deep-fake images generated by GANs, such as CNN, CAT-NET, and Buster-NET, have been analyzed in this study. A lightweight model combining Progressive GAN and DCGAN has been proposed, which performs more quickly than current techniques. Promising experimental results have been achieved, with an overall accuracy of 91.63%.