I am Zheng Fang, from Wuhan University.
My research focuses on Audio Adversarial Attack and Large Language Model (LLM) Security.
Recently, I have been exploring audio-LLM security and agent security.

From June to August 2024, I worked as an LLM Red Team Intern at Alibaba Group, where I evaluated and stress-tested large language models and contributed to developing red-team methodologies for model safety.

📧 zhengfang618@whu.edu.cn
🎓 Google Scholar Profile

Research Interests

• Audio Adversarial Attack
• Large Language Model Security
• Audio-LLM Security and Agent Security

First-Author Publications

1. Zero-Query Adversarial Attack on Black-box Automatic Speech Recognition Systems
   ACM CCS 2024 (Big4). Zheng Fang, Tao Wang, Lingchen Zhao, Yunjie Ge, Shenyi Zhang, Qi Li, Bowen Li, Chao Shen, Qian Wang.

2. Selective Masking Adversarial Attack on Automatic Speech Recognition Systems
   IEEE ICME 2025. Zheng Fang, Shenyi Zhang, Tao Wang, Bowen Li, Lingchen Zhao, Zhangyi Wang.

Co-authored Publications

1. JBShield: Defending Large Language Models from Jailbreak Attacks through Activated Concept Analysis and Manipulation
   USENIX Security 2025 (Big4). Shenyi Zhang, Yuchen Zhai, Keyan Guo, Hongxin Hu, Shengnan Guo, Zheng Fang, Lingchen Zhao, Chao Shen, Cong Wang, Qian Wang.

2. CuckooAttack: Towards Practical Backdoor Attack against Automatic Speech Recognition Systems
   IEEE Transactions on Dependable and Secure Computing (TDSC), 2025.
   Bowen Li, Yunjie Ge, Zheng Fang, Tao Wang, Lingchen Zhao, Quan Lu.

3. Hijacking Attacks against Neural Networks by Analyzing Training Data
   USENIX Security 2024 (Big4).
   Yunjie Ge, Qian Wang, Huayang Huang, Qi Li, Cong Wang, Chao Shen, Lingchen Zhao, Peipei Jiang, Zheng Fang, Shenyi Zhang.

4. Palette: Physically-Realizable Backdoor Attacks Against Video Recognition Models
   IEEE Transactions on Dependable and Secure Computing (TDSC), 2024.
   Xueluan Gong, Zheng Fang, Bowen Li, Tao Wang, Yanjiao Chen, Qian Wang.