Welcome to My Homepage

👨‍💻 About Me

Pinji Chen is a Ph.D. student in the Institute for Network Sciences and Cyberspace at Tsinghua University. He works in the Network and Information Security (NISL) Lab, advised by Prof. Jianjun Chen and Haixin Duan.

His research interests include AI security and network security. He is particularly interested in exploring emerging technologies and identifying & mitigating their vulnerabilities. His research has been published at major security conferences, including USENIX Security 2024, NDSS 2025, Black Hat USA 2025, S&P 2026, and NDSS 2027.

🎓 Education

• Sep 2024 – Present
  Ph.D. Student, Tsinghua University

• Sep 2020 – Jun 2024
  B.E., Wuhan University

🏆 Selected Awards and Honors

• First-Class Scholarship, Tsinghua University, 2025
• Top Ten Outstanding Youth (10 students across the university), Wuhan University , 2024
• Person of the Year (榜样珞珈, 10 undergraduates across the university), Wuhan University, 2024
• National Scholarship (Top 1%), Ministry of Education of China, 2023
• Lei Jun Scholarship (Top 1%), Wuhan University, 2022
• First-Class Scholarship (Top 5%), Wuhan University, three times (2021,2022,2023)

💼 Internship Experience

• Aug 2026 – Present
  Tencent Qingyun Program (腾讯青云计划) — Research Intern at Tencent, focusing on agent security.

🚩 CTF & Vulnerability Discovery

Member of Tsinghua Redbud CTF Team.

Competitions

• Aug 2025 — 🥈Runner-up, DEF CON CTF 2025 (Finals of the world’s premier CTF)

• Sep 2026 — 🥇First Prize, West Lake Lunjian (西湖论剑)

• Sep 2025 — 🥈Runner-up, Tianwang Cup (天网杯, LLM Track)

• Sep 2024 — 🥈Runner-up, ByteDance ByteAI Challenge (¥10,000 prize)

Vulnerability Discovery

• Discovered multiple vulnerabilities in AI systems and network infrastructure of leading companies, including Google, Microsoft, OpenAI, NVIDIA, xAI, Cloudflare, Baidu, Tencent, Alibaba, ByteDance, Huawei, and Meituan, receiving official acknowledgments and bug bounties.
• Uncovered dozens of high-severity vulnerabilities in open-source ML frameworks, earning 20+ high-severity CVE / CNVD / CNNVD identifiers.

🏓 Extracurricular Achievements

National First-Class Athlete in Table Tennis (China)

• 2025 — 🥇Champion, Ma Yuehan Cup (Team), Tsinghua University
• 2020 — 🥈Runner-up, Men’s Team & Mixed Doubles, Hubei Provincial University Table Tennis Championship
• 2018 — 5th Place, Men’s Singles, National High School Table Tennis Championship
• 2018 — 🥇Champion (Team) & 🥈Runner-up (Singles), Hubei Provincial High School Table Tennis Championship
• 2017 — 🥇Champion, Men’s Singles, Beijing Middle School Table Tennis Championship
• 2014 — 🥇Champion, Men’s Singles, Beijing Primary School Table Tennis Championship
• 2013 — 🥇Champion (Team) & 🥈Runner-up (Singles), China Youth Innovation Cup Table Tennis Championship
• 2013 — 🥈Runner-up, Men’s Singles, Beijing Primary School Table Tennis Championship
• 2012 — 🥇Champion, Men’s Singles, Beijing Primary School Table Tennis Championship

📰 News

• Sep 2026 - Our work: Detecting Insecure Pickle Deserialization of Untrusted Data in Machine Learning Frameworks is accepted by NDSS 2027 🎉🎉🎉 [To appear]

• Sep 2026 - Our work: Host Collision: Traversing Internal Networks via Reverse-Proxy Routing Misconfigurations is accepted by NDSS 2027 🎉🎉🎉 [To appear]

• Aug 2026 - Joined Tencent via the Qingyun Program. Looking forward to the journey ahead! 🎉🎉🎉

• Mar 2026 - Our work: LLMThief: Evaluating Configuration Leaking Risks in Commercial LLM App Stores is accepted by S&P 2026 🎉🎉🎉 [Link]

• Aug 2025 - Our work: Identifying Logical Vulnerabilities in QUIC Implementations is accepted by NDSS 2026 🎉🎉🎉 [Link]

• Aug 2025 — Honored to be a Briefings Speaker at Black Hat USA 2025 🎉🎉🎉 Awesome experience in Las Vegas! Incredible event 👍👍👍 [Link]

• July 2025 — Our work: The Danger of Packet Length Leakage: Off-path TCP/IP Hijacking Attacks Against Wireless and Mobile Networks appeared at EuroS&P 2025 🎉🎉🎉 [Link]

• Feb 2025 — Our work: Cross-Origin Web Attacks via HTTP/2 Server Push and Signed HTTP Exchange appeared at NDSS 2025 🎉🎉🎉 [Link]

• Aug 2024 — Our work: More Simplicity for Trainers, More Opportunity for Attackers: Black-Box Attacks on Speaker Recognition Systems by Inferring Feature Extractor appeared at USENIX Security 2024 🎉🎉🎉 [Link]