Keynotes/Talks

The IAPR Computation Forensics Technical Committee is announcing a new webinar Backdooring Deep Neural Networks: from threats to opportunities by Dr. Benedetta Tondi on 11 May 2023 at (9 am in Z├╝rich, Paris and Rome, 8 am in London, UK).

Title: Backdooring Deep Neural Networks: from threats to opportunities

Abstract: Deep Neural Networks (DNNs) have been shown to be vulnerable to backdoor attacks, hiding a backdoor behaviour or functionality into the target DNN by poisoning a portion of the training data. A backdoored model behaves normally on standard test inputs, while it predicts a target class when the test sample contains a so-called trigger signal. As such, backdoor attacks are very subtle attacks raising severe security concerns in real-world applications in all the fields, from computer vision to natural language processing, and several defence methods have been proposed.

In this talk, we will identify different threat models under which the backdoor attacks and defences operate, where different requirements have to be satisfied by backdoor attacks and defence methods. Distinction is made depending on the control that the attacker has on the training process. Understanding the threat scenario is a necessary step for a proper security analysis and to develop effective defences against this threat. Then, in the second part of the talk, we will discover that backdoor attacks have also benign uses (not all bad comes to hurt). In particular, backdoor attacks have been exploited for watermarking DNNs with much success, thus turning the backdoor attack threat to an opportunity.

Bio of the speaker: Benedetta Tondi received the master degree (cum laude) in Electronics and Communications Engineering at the University of Siena, Siena, Italy, in 2012, and the PhD degree in 2016 with a thesis on the “Theoretical Foundations of Adversarial Detection and Applications to Multimedia Forensics”. She is currently Assistant Professor at the Department of Information Engineering and Mathematics,
University of Siena. She is a member of the IEEE Young Professionals and IEEE Signal Processing Society and a member of the National Inter- University Consortium for Telecommunications (CNIT).
Her research interest focuses on the application of Information theory and Game theory concepts to forensics and counter-forensics analysis, and on adversarial signal processing. More recently, her research is focused on adversarial machine learning and on the use of deep learning techniques for forensics and security. She has been member of the Technical Program Committee, Organizing Committee and Area Chair of several Eurasip, ACM and IEEE Workshops and International Conferences. She has been Technical Program Chair at ACM IH&MMSEC 2022.
She has been organizer of several special sessions at Eurasip and IEEE workshops and Guest Editor of several special issues. She is recipient of the Best Student Paper Award at the IEEE International Workshop on Information Forensics and Security (WIFS) 2014, the Best Paper Award at the IEEE International Workshop on Information Forensics and Security (WIFS) 2015 and at the The Ninth International Conferences on Advances in Multimedia (MMEDIA) 2017. She is winner of the 2017 GTTI PhD Award for the best PhD Theses defended at an Italian University in the areas of Communications Technologies (Signal Processing, Digital Communications, Networking).She currently serves as an Associate Editor for the EURASIP Journal of Information Security and for the IEEE Transactions on Information Forensics and Security. She is also a member of the Information Forensics and Security (IFS) Technical Committee of the IEEE Signal Processing Society.

 The link to join the talk is here: meet.google.com/gva-mbsr-swj

This page summarizes all the keynotes and presentations related to the IAPR TC6 topics: