The first artificial intelligence safety competition officially launched

In order to accelerate the construction of an AI security innovation environment, on July 13, the 2022 AI Security Competition with the theme of "Building AI Security and Sharing an Intelligent Future" was officially launched.As a special competition of the Zhongguancun Forum International Frontier Technology Innovation Competition, this competition focuses on the security challenges faced by the AI ​​industry, aiming to solve security problems through competition, cultivate high-level AI security teams, and at the same time strengthen the AI ​​governance system and securityEvaluation capacity building to help the high-quality development of the artificial intelligence industry.

This competition is guided by Beijing Municipal Science and Technology Commission, Zhongguancun Science and Technology Park Management Committee, Beijing Municipal Bureau of Economy and Information Technology, Zhongguancun National Laboratory, National Industrial Information Security Development Research Center, Tsinghua University Artificial Intelligence Research Institute,Co-hosted by Laizhihui Technology Co., Ltd. and co-organized by the Artificial Intelligence Working Committee of China Electronic Information Industry Federation, the competition is authoritative, academic and professional.

This competition brings together top experts in the industry. Five academicians, Zhang Xi, academician of the Chinese Academy of Sciences, Ni Guangnan, academician of the Chinese Academy of Engineering, Shen Changxiang, academician of the Chinese Academy of Engineering, Yu Quan, academician of the Chinese Academy of Engineering, and Wu Jianping, academician of the Chinese Academy of Engineering, were invited to serve as the organizing committee of this competition.Chairman, build the highest-level competitive exchange platform in the field of artificial intelligence security.

This competition is the first competition in China with the theme of artificial intelligence safety. It sets competition topics in the fields of face recognition safety, deep forgery content management, automatic driving safety, etc., combining the actual needs of industrial applications with the most cutting-edge technologiesThe integration of security issues will provide an important reference for the evolution of AI capabilities.The topics of the competition are as follows:

Deep forgery security competition: At present, the abuse of forging audio and video using generative synthesis algorithms is increasing, and tracing the source by analyzing the similarity of forged audio and video has become an effective response.This challenge simulates the spread of deepfake videos on the Internet. By submitting a traceability algorithm for the forgery method, the contestants identify deepfake videos with the same generation method, and finally rank according to the accuracy rate.

Autonomous driving safety competition: Self-driving vehicles are prone to misjudgment in extreme scenarios, and at the same time face algorithm security threats, which may be maliciously attacked.This challenge simulates adversarial attacks in autonomous driving scenarios, and provides an online simulated autonomous driving environment. By submitting adversarial patches, players conduct attack tests on mainstream autonomous driving target detection models, and finally rank according to the attack success rate.

Face recognition security competition: The face recognition system is vulnerable to adversarial sample attacks, that is, by making small changes to the face image, it can deceive the recognition system to make a wrong judgment.This challenge simulates an adversarial attack in a face comparison scenario. In a black-box scenario where the details of the model are unknown, contestants submit adversarial patches to conduct an attack test on the model, and finally rank according to the attack success rate.

This competition has a generous prize pool. At the same time, the top ten teams in the finals of each track will receive a certificate of honor issued by the competition.

• Champion of the finals: RMB 50,000 (single track bonus)

• The runner-up in the finals: RMB 20,000 (single track bonus)

• Third place in the finals: 10,000 RMB (single track bonus)

At present, all the work of the competition is ready, and we sincerely invite algorithm developers from domestic universities, research institutes, and technology enterprises to participate in this offensive and defensive competition related to future security (registration method: log on to the official website of the competition: www..zgc-aisc.com), strengthen technical research in the frontier field of artificial intelligence security, and escort the development of artificial intelligence industry!