Exploring AI Security Frontiers at the 12th Wu Wenjun AI Science & Technology Awards

Overview

The Wu Wenjun Artificial Intelligence Science & Technology Award, organized by the China Association for Artificial Intelligence, convened from May 6‑8, 2023 in Beijing and Suzhou. The event, themed “Scenario‑Driven • Digital‑Intelligent Nation,” featured a main forum, ten specialized forums, award ceremonies, launch events, and high‑level discussions aimed at shaping AI research, industry integration, and policy guidance.

Event Details

The conference adopted a “1+10+X” format, including one main forum, ten thematic forums, and a series of activities such as award presentations, startup ceremonies, report releases, and expert panels. It sought to showcase AI trends, gather top scientific resources, and promote the industrialization of AI breakthroughs.

Keynote and Forum Speakers

Lin Chenhao – Distinguished researcher at Xi'an Jiaotong University, expert in intelligent identity security and AI safety.

Shen Chao – Professor at Xi'an Jiaotong University, recipient of multiple national youth science awards, focuses on AI security and data‑driven cyberspace security.

Chen Kai – Researcher at the Institute of Information Engineering, Chinese Academy of Sciences; specializes in system security and AI safety.

Zhang Weiming – Professor at University of Science and Technology of China, renowned for work on deepfake detection and AI‑generated content security.

Ji Shouling – Researcher at Zhejiang University, leads the Trusted AI Research Center, focuses on AI and software/system security.

Liu Xianglong – Professor at Beihang University, expert in adversarial attacks, deep learning security, and evaluation platforms.

Yang Minghui – Security expert at OPPO Guangdong, specialist in data security, cryptography, and mobile AI trustworthiness.

Zhang Zhiqiang – Senior algorithm expert at Ant Group, leads graph learning and data‑intelligence applications for industry.

Report 1: Chen Kai – Adversarial Attacks and Defenses for AI Models

Abstract: Recent AI advances have introduced security threats such as adversarial examples and model backdoors, endangering applications like autonomous driving and facial‑payment systems. This talk covers physical‑world adversarial samples, model backdoors, fine‑grained model repair, defense strategies, model stealing, and watermark protection.

Report 2: Zhang Weiming – Active and Passive Defense Methods for Deepfakes

Abstract: Generative AI misuse, especially deepfake videos, poses economic fraud, misinformation, and national security risks. As deepfakes evolve toward high‑quality, adversarial content, detection becomes an asymmetric challenge. The presentation explores generation‑side defenses for deepfake videos and detection of AI‑generated text.

Report 3: Ji Shouling – FreeEagle: Detecting Complex Neutral Trojans in Data‑Free Cases

Abstract: (Image illustrating the FreeEagle system is included.) The work introduces FreeEagle, a data‑free approach for detecting sophisticated neutral trojans without requiring training data.

Report 4: Liu Xianglong – Adversarial Attacks and Security Evaluation of Deep Learning

Abstract: Deep learning is widely applied but faces adversarial samples and model backdoors across data, model, training, and deployment stages. The talk presents research on adversarial sample generation, backdoor detection, defense mechanisms, and introduces the “Zhongming” security evaluation platform and the “Wanxiang” physical simulation platform.

Report 5: Yang Minghui – Chip‑Cloud Collaboration for Trusted Mobile Intelligent Terminals

Abstract: In the era of ubiquitous connectivity, device security, privacy, and data protection become critical. Trustworthy mobile terminals require integrated mechanisms spanning system, identity, connection, application, behavior, and AI to ensure a secure user experience.

Report 6: Zhang Zhiqiang – Data‑Intelligence Practice in Technology‑Assisted Real‑World Applications

Abstract: The ATEC2022 competition, co‑organized by Ant Group and top universities, focused on “Technology Assisting Industry Development.” The presentation shares insights on applying data‑intelligence techniques to real‑world scenarios.