SHANGHAI, Sept. 30, 2024 /PRNewswire/ -- The Global AI Defense Challenge 2024, co-hosted by the China Society of Image and Graphics (CSIG), the Cloud Security Alliance Greater China Region (CSA GCR), and Ant Group, launched on September 7, with a total prize pool of more than 75,000 USD.
The competition focuses on defense strategies for the applications of large models. Technical experts, white-hat hackers, and other participants are invited to join this competition in two tracks:
In Track 1, participants leverage various dynamic attack techniques, including target hijacking, scenario embedding, and logical nesting, to induce large models to generate problematic images, identify potential vulnerabilities, and enhance the security capabilities of large models. Participants joining Track 2 need to detect fake KYC(Know-Your-Customer) information in financial scenarios and address threats from deepfakes and AI-generated fake credentials.
This competition provides a dataset comprising more than 1 million fake credential samples. Participants need to develop and train models, evaluate their effectiveness using separate test datasets, and let their models to identify the possibility of data forgery.
The competition is being organized by Ant Security Lab, Alibaba Cloud, Tsinghua University, Shanghai Jiao Tong University, Zhejiang University, and other industry, academic, and research institutions, with the Shanghai Artificial Intelligence Laboratory serving as a technical advisor. The competition aims to facilitate collaboration between the academic and industry communities, exploring and addressing potential technical risks during the rapid development of large models and supporting the sustainable growth for the AI industry globally.
According to Wang Yaonan, an Academician of the Chinese Academy of Engineering and Chairman of the CSIG, the security and governance of AI systems are especially important in the era of large models, and potential risks in the software supply chain should be properly addressed. Nowadays, image and graphics technologies play a crucial role in various innovative applications of large models, leading to an increase in innovations in image recognition and security technologies. This competition is dedicated to exploring more innovative solutions to enhance the security of large models.
The judging panel for this competition comprises experts from both academic and industry communities. Wang Yaonan, and Yale Li, Chairman of the CSA GCR, are serving as the leading advisors for the competition. They are joined by nearly 30 renowned scholars from institutions including Tsinghua University, Shanghai Jiao Tong University, Shanghai Artificial Intelligence Laboratory, and the China Society of Image and Graphics to organize the competition and evaluate results.
Timeline (UTC+8):
09/06 - 10/20 Registration
09/18 - 10/31 Official Competition and Result Submission
10/31 - 11/02 Expert Evaluation
Early November (tentative) Results Announcement
Participants can register via the Alibaba Cloud Tianchi Big Data Intelligence platform, among others.
Global AI Defense Challenge
Track 1：Vaccination for Text-to-Image Generative Models
https://tianchi.aliyun.com/competition/entrance/532268?spm=a2c22.29524702.0.0.240b526etSlSFS&lang=en-us
Track 2：AI Identity Verification - Financial Certificate Tampering Detection
https://tianchi.aliyun.com/competition/entrance/532267?spm=a2c22.29524702.0.0.240b526etSlSFS&lang=en-us
