Deepfake competition results announced: Emphasizing ethical technology use

The results of the INCLUSION•Conference on the Bund and Ant Group’ Global Multimedia Deepfake Detection Challenge 2024 are in, and provide some good news amid a proliferation of faked and manipulated content online.

More than 2,200 participants competed in the challenge over a two-month period. Participants hailed from China, the U.S., India, Australia, Japan, Vietnam and Indonesia, among 26 countries and regions around the world.

The competition was set up to evaluate the effectiveness and accuracy of deepfake detection tools in real-world scenarios, with a panel of experts also assessing the generalizability and practical applicability of each entry. The INCLUSION Conference on the Bund co-hosted the competition with Ant Digital Technologies. The total prize pool was 1,000,000 RMB (around US$137,000), and an award ceremony was held at the Bund Summit on Friday.

Inclusion Conference 2024 participant

A team of scholars from Hong Kong and Macau won the Image Track with 97.038 percent accuracy in deepfake detection across a variety of different image types and scenarios. The team was led by Dr. Wu Haiwei, a postdoctoral fellow at the City University of Hong Kong.

Solo developer Tang Yongwei won the Audia & Video Track of the competition, who developed algorithms to detect audio manipulation.

The Challenge’s committee applauds the dedication of all participants in the announcement and states that the participants demonstrated the potential of AI to address important challenges.

Recent studies, such as one published in August by iProov, indicate that deepfakes are already impacting businesses around the world. UN Secretary-General António Guterres suggested in his annual address this year that deepfakes “could have serious implications for democracy, peace and stability.”

Sharing ideas and code

Zoloz CTO William Yao played a key role in organizing the competition and assigning tasks. Zoloz is a subsidiary of Ant Group, and has been working on generative AI and deepfake detection since 2019. Yao says the company “incorporated our commercial experience in attack and defense into the competition design and dataset.”

Inclusion Conference 2024 William Yao, Zoloz CTO

Zoloz launched its biometric deepfake detection software in April, joining an emerging marketplace of tools to defend against fraudulent content.

In addition to Zoloz, Ant Group has invested in intellectual property around palm biometrics for retail purchases.

Ant Digital Technologies CEO Wenbiao Zhao emphasizes the opportunity provided by the challenge for researchers to exchange ideas.

Following the competition’s completion, the VisionRush team based at the Institute of Automation at the Chinese Academy of Sciences shared its competition models as open-source code on GitHub. The move is an attempt to lower the barrier to deepfake detection deployment.

“If one person’s torch is too small, we will pass the torch to everyone in the world,” says VisionRush’s Zhang Xinyi, who also won an award in the competition.

The open-source release by VisionRush created a stir in China, reaching the top of Weibo’s trending topics list.

“This competition provides researchers with a highly simulated industrial environment for practice, which helps integrate industry, academia, and research, and cultivate responsible practical talents,” says Joey Zhou Tianyi, deputy director of the Centre for Frontier AI Research at the Agency for Science, Technology and Research in Singapore.

The Global Multimedia Deepfake Detection Challenge 2024 is part of a growing trend of influential organizations bringing together AI experts and developers to tackle the threat of deepfakes.

This is a sponsored post. For information about advertising, please contact us. 

Article Topics

Ant Group  |  biometric liveness detection  |  biometrics  |  deepfake detection  |  deepfakes  |  ethics  |  INCLUSION Conference  |  responsible AI  |  VisionRush  |  Zoloz