Advanced methods for deepfake detection and classification 2024_RTDA_DEIB_1

A deepfake is a type of (multi)media content manipulated in a hyper-realistic manner through the use of artificial intelligence (AI) and deep learning techniques. The goal of the research activity is to develop advanced forensic techniques to detect whether a (multi)media content has been totally (or partially) synthetically generated through the use of AI and deep learning. The techniques will exploit both advanced notions of signal processing and machine learning. The activities will involve the analysis of image, audio, and video content, with particular attention to processing in the time domain. For example, for video content, the most advanced analyses are based on frame-by-frame processing followed by simple fusion strategies. A more holistic approach will be developed that looks jointly at the inconsistency in spatial and temporal dimensions. The activity is in full compliance with the principles of Spoke 2 - PNRR PESERICS Funded by the European Union - NextGenerationEU.