Hi 👋, I’m Mingxin.
I am currently a first-year PhD student based in Beijing, currently affiliated with Peking University. I’m fortunate to be supervised by Dr. Xinggong Zhang.I received my bachelor’s degree from Huazhong University of Science and Technology (HUST) that built a solid foundation in computer science and research, and have previously conducted research collaborations with researchers from the Institute of Computing Technology, Chinese Academy of Sciences (ICT, CAS) and Harbin Institute of Technology (Shenzhen).
💡 Interests
- AI Security
- Security, privacy, and alignment issues of large language models
- Jailbreak, robustness, and misuse analysis
- Network Security
- Malicious traffic detection
- Encrypted traffic classification and website fingerprinting
- AI Agents More recently, I have been exploring AI agents, focusing on autonomous and tool-augmented LLM-based systems. I am interested in how agentic workflows and multi-agent collaboration can enable LLMs to solve complex, real-world tasks in a secure and controllable manner.
I am always happy to connect with people who share interests in AI, security, and related research topics. I believe this is a field driven by diverse ideas and perspectives, and I am eager to learn through discussions and collaborations. Feel free to reach out.
✈️ News
- 2025.09.12: Attended the 3rd CCF ChinaNet 2025 conference in Shenyang, China.
- 2025.09: Enrolled at Peking University and began my Ph.D. studies.
- 2025.08.07: Participated in APNet 2025, held in Shanghai, China. Thanks to Yunpeng Tan.
- 2025.06: Graduated from Huazhong University of Science and Technology — a truly wonderful journey.
📝 Selected Publications
Please refer to my Google Scholar for full list of publications.
🧑🔬 AI for Security
Vuldetectbench: Evaluating the deep capability of vulnerability detection with large language models
Yu Liu*, Lang Gao*, Mingxin Yang*, Yu Xie, Ping Chen, Xiaojin Zhang, Wei Chen (*: Joint first authors)
“A new benchmark, VulDetectBench, evaluates LLMs on program vulnerability detection and shows strong performance on coarse-grained tasks but poor capability in fine-grained vulnerability analysis.”
| Paper | Code |
AI Safety
Unveiling the vulnerability of private fine-tuning in split-based frameworks for large language models: A bidirectionally enhanced attack
🧑🔬 Network Security
Graph-Based Encrypted Malicious Traffic Detection Under Flow Distribution Drift With Flow Sampling
Yunpeng Tan, Qingyang Li, Mingxin Yang, Xinggong Zhang
“Propose a flow-sampling-based graph construction method to mitigate flow distribution drift in encrypted traffic detection, significantly improving GNN accuracy and efficiency.”
| Paper |
👨🔧 Applications
Linger: Intelligent Resume Parsing System
Yifei Li, Mingxin Yang, Min Yang, Siyu Jiang
“Under the guidance of Professor He Kun, our team proposed an Intelligent Resume Analysis System that leverages natural language processing and machine learning—integrating resume text recognition, large language models, and job–candidate matching algorithms—to efficiently and accurately parse resumes and address real-world recruitment challenges.”
| News |
🧐 Service
- 2025, Open Source Summer Campus Ambassador
💼 Experiences
- [10 / 2022 - 07 / 2024 ]
HUST (Supervisor: Prof. Kun He, topic: Jailbreak) - [07 / 2023 - 01 / 2024] HUST (Supervisor: Prof. Wei Chen, topic: Code Detection)
- [12 / 2023 - 12 / 2024]
Harbin Institute of Technology, Shenzhen (Supervisor: Guanzhong Chen, topic: Split-Learning in Large Language Model) - [03 / 2024 - 06 / 2024]
Institute of Computing Technology, Chinese Academy of Sciences (Supervisor: Prof. Fei Sun, topic: AI Safety)
💬 I am deeply grateful to all the mentors and collaborators who have guided and supported me along the way. Your encouragement, trust, and inspiration have made all the difference in my journey.
📖 Educations
08 / 2025 - Now : Ph.D. student, 
Peking University
09 / 2021 - 07 / 2025 : B.E., Huazhong University of Science and Technology