Online GIAN Course "Adversarial Signal Processing and Machine Learning with applications to Multimedia Forensics"

14 Feb 2022   -   18 Feb 2022
Electronics and Communication Engineering
Event Details Images

Security-oriented applications of signal processing are receiving increasing attention. Digital watermarking, steganography, multimedia forensics, biometrics, intrusion detection, network monitoring, are just a few. In all these cases, the presence of one or more adversaries aiming at making the system fail cannot be neglected. For each of the above fields, several attacks and counter-attacks have been developed, often following a typical cat & mouse loop wherein attacks and countermeasures are iteratively developed each time focusing on the latest developed solutions. A problem with such an approach is that it fails to provide a unifying view of the challenges that the application of signal processing tools in an adversarial setting poses. Worse than that, the security of the proposed solutions is hardly provable due to the lack of rigorous security models suited to capture the peculiarities of the addressed scenarios. The situation is even more critical when Machine Learning (ML) and Artificial Intelligence (AI) tools are involved. In fact, while the use of ML and AI tools can greatly boost the performance of security-oriented systems, their weakness to adversarial attacks can introduce into the system new security breaches thus compromising its security.