Fighting deepfakes by detecting GAN DCT anomalies

Synthetic multimedia content created through AI technologies, such as Generative Adversarial Networks (GAN), applied to human faces can bring serious social and political consequences in the private life of every person. State-of-the-art algorithms use deep neural networks to detect a fake content but unfortunately almost all approaches appear to be neither generalizable nor explainable. A new fast detection method able to discriminate Deepfake images with blazing speed and high precision is exposed. By employing Discrete Cosine Transform (DCT), anomalous frequencies in real and Deepfake datasets were analyzed. The \beta statistics inferred by the AC coefficients distribution have been the key to recognize GAN-engine generated data. The technique has been validated on pristine high quality faces synthesized by different GANs architectures. Experiments carried out show that the method is innovative, exceeds the state-of-the-art and also gives many insights in terms of explainability.