AI for Security
Machine learning and language models for software vulnerability detection, malicious application analysis, and security-related web content analysis.
About
I am a master’s student in Network and Information Security at Beijing University of Posts and Telecommunications, with expected graduation in June 2027. I am also a Research Assistant at The Chinese University of Hong Kong.
My research spans AI for Security and Security of AI, with work on learning-based vulnerability detection, language model robustness, and the security of AI agents.
I also collaborate on multimodal image retrieval and long-video reasoning. I received my bachelor’s degree in Information Security from Wuhan University of Science and Technology in June 2024.
Research interests
Security of AI
Security and robustness of language models and AI agents, including adversarial evaluation, deployment environments, cross-origin information use, and hallucination propagation in multi-agent systems.
Education
Sep 2024 –
Jun 2027 (expected)
Jun 2027 (expected)
Beijing University of Posts and Telecommunications
M.Eng. in Network and Information Security
Beijing, China
State Key Laboratory of Networking and Switching Technology
Research advisor: Tengfei Tu
Sep 2020 –
Jun 2024
Jun 2024
Wuhan University of Science and Technology
B.Eng. in Information Security
Wuhan, Hubei, China
School of Computer Science and Technology
GPA: 3.70 / 4.00
Research advisor: Tao Li
2017 – 2020
Hubei Xiaogan Senior Middle School
Senior secondary education
Xiaogan, Hubei, China
Research experience
May 2026 –
present
present
The Chinese University of Hong Kong
Research Assistant
Hong Kong SAR, China
Research advisor: Diwen Xue
Studying how content from different web origins influences browser-based AI agents. This ongoing project uses execution-trace analysis, controlled context masking, and action replay to distinguish individual and joint source effects.
Oct 2024 –
May 2026
May 2026
Tsinghua University
Research Assistant
Shenzhen, Guangdong, China
Shenzhen International Graduate School
Built agricultural multimodal data pipelines for dataset collection, image–text cleaning and conversion, augmentation, and multi-turn dialogue processing. Reproduced a LLaVA demo and refined video workflows for structured and unstructured information extraction.
Trained and evaluated Qwen-based models, including agricultural instruction tuning with Qwen2.5-VL and LLaMA-Factory, and experiments with Qwen3-4B. Compared prompting, LoRA and full fine-tuning, data proportions, and inference settings, and reproduced a fraud-detection evaluation pipeline.