TruceNet: A CNN-Based Model for Accurate Classification of DeepFake Images

With the rise of GANs, the prevalence of Deepfakes has increased, resulting in significant privacy, security, and social concerns. To tackle this issue, we propose TruceNet, a novel approach that utilizes a CNN architecture to detect Deepfakes in images. Our method analyzes various image features to determine if manipulation has occurred. Through experiments on a comprehensive dataset of Deepfakes and non-Deepfakes, we evaluate TruceNet's effectiveness. Our CNN architecture achieves high detection accuracy, outperforming state-of-the-art methods. TruceNet exhibits robustness against common Deepfake manipulations such as face swapping, expression modification, and Neural Texture Synthesis. It consistently achieves an average accuracy of 94.2% in correctly classifying manipulated images, maintaining a precision of 95% across different manipulation scenarios. The research findings highlight TruceNet's significant contribution in mitigating Deepfake risks and enhancing digital media security and trustworthiness. By accurately detecting Deepfakes, TruceNet offers a reliable solution to combat the growing threat posed by manipulated images.