As generative AI rapidly advances, the misuse of technology has led to the proliferation of deepfakes, posing significant risks to society and individuals. Building a secure and reliable system for detecting and defending against deepfake content is now more critical than ever.
In response to the emerging security challenges posed by AI-generated forgeries, the International Joint Conference on Artificial Intelligence (IJCAI) 2025 will host a workshop and challenge focused on Deepfake Detection, Localization, and Explainability. The event aims to bring together leading global scholars and industry experts to tackle core technical challenges, including multimodal forgery, weakly supervised forgery localization, forgery explainability, and adversarial attacks and defenses against generative AI.
The challenge, organized by Ant Group Digital Technologies, will provide participants with a massive dataset of over 1.8 million samples to conduct practical exercises in attacking and defending against AI fraud risks, thereby boosting AI security research. The event is co-organized by several prestigious institutions, including the Agency for Science, Technology and Research (A*STAR) in Singapore, Nanyang Technological University, Tsinghua University, the Institute of Automation of the Chinese Academy of Sciences, Hefei University of Technology, the Anhui Provincial Key Laboratory of Digital Security, the University of Rochester, the University at Buffalo, and the University of Campinas.
The challenge will feature both image and audio-visual tracks, adopting a defense through competition approach to address the limitations of existing detection algorithms in accurately locating forged regions and identifying multimodal audio-visual collaborative forgeries. This initiative aims to enhance the level of content security in the AI era.
Notably, Ant Group Digital Technologies has opened up a multimodal deepfake dataset of over 1.8 million forged samples for the challenge, covering 88 types of forgery techniques. This effort will promote the establishment of verifiable,
Views: 0