On December 13th, 2024 at 4.30 pm the seminar titled "Image Forensics in the Era of AI: Challenges and Solutions" will take place at Politecnico di Milano, 25.0.2 Room (Building 25).

The rapid advancement of artificial intelligence (AI) has enabled the creation of highly realistic synthetic images commonly known as deepfakes, posing significant challenges for the field of image forensics. Among the most pressing issues, there are the detection and analysis of deepfake images, particularly those depicting human faces, which have been weaponized for misinformation, identity theft, and other malicious purposes. Deepfake human faces are not the only emerging concern, as synthetic biomedical images manipulated or completely generated by AI risk to compromise the integrity of scientific research and diagnostic processes.
In recent years, the forensic community has developed numerous detection methods to identify synthetically generated content. Despite these advancements, several critical challenges remain. For instance, AI-based detectors often lack interpretability, making it difficult to understand their decision-making processes. Additionally, these systems frequently struggle to maintain robustness when strong post-processing techniques are applied to the images. Most concerning is recent evidence showing that neural compression methods, such as those implemented in emerging standards like JPEG AI, can severely undermine the ability of detectors to distinguish between real and synthetic content.
This seminar will offer a detailed examination of these evolving threats and introduce state-of-the-art forensic methodologies aimed at addressing them. Particular emphasis will be placed on the ongoing research and innovations at the Image and Sound Processing Lab of Politecnico di Milano, showcasing efforts to advance the field of image forensics in the face of rapidly evolving challenges.

This talk is part of the imaging seminar series organized within the EMJM program.