Guest Lecture: On the security challenges in deep learning

Date: 21 February 2025

A research talk on "On the security challenges in deep learning" was held on February 21, 2025, from 9:30 to 11:00 am. The purpose of this session was to provide the audience with an understanding of machine learning security topics and to discuss various attacks and defenses. The talk began with a welcome address by Dr. Lee Wai Kong, who introduced the invited speaker, Dr. Stjepan Picek from Radboud University, Netherlands. There are 30 participants on-site and 89 online, with some participants joining the session from Korea and the UK. Dr. Picek started the presentation with a brief introduction to the security aspects of machine learning. He then discussed sponge attacks, defenses against adversarial examples, and strategies to counter backdoor attacks. Throughout the talk, Dr. Picek emphasized the challenges arising from collaborative learning settings and the implications of using generative AI. He concluded with an overview of open questions and the challenges facing this rapidly evolving field. The session ended with a sincere thank you to Dr. Stjepan Picek and the presentation of a souvenir as a token of appreciation for his valuable contribution to the event.