Currently, I am researcher at Shanghai AI Lab, working on AI agent systems for cybersecurity and AI agent security.
I earned my Ph.D. degree in 2025 from Zhejiang University, under the fortunate supervison of Prof. Prof. Zhongjie Ba. I was honored to work with Prof. Peng Cheng and Prof. Qinglong Wang. Before that, I obtained my bachelor’s degree from the College of Information Science & Electronic Engineering, Zhejiang University, China.
I’m always open to collaborations! Feel free to reach out at penghuang1999 AT gmail.com.
Research Interests: Agent for Cybersecurity, Agent Security, IoT Security. CV
🔥 News
- 2026.07: 🎉🎉 Our paper “MagStealth: A Stealth Command Injection Attack on Voice Assistants Via Magnetic-Driven Inner Loudspeakers” got accepted in IEEE TDSC.
- 2026.05: 🎉🎉 I was selected as the ACM MobiSys Rising Star 2026 for research on building systems to protect users’ voice privacy.
- 2025.04: 🎉🎉 Our paper “SecHeadset: A Practical Privacy Protection System for Real-time Voice Communication” got accepted in MobiSys 2025 (Acceptance Rate: 18.0%, 42/233).
- 2024.10: 🎉🎉 Our paper “Phoneme-Based Proactive Anti-Eavesdropping with Controlled Recording Privilege” got accepted in IEEE TDSC.
- 2023.07: 🎉🎉 Our paper “Text-CRS: A Generalized Certified Robustness Framework against Textual Adversarial Attacks” got accepted in IEEE S&P 2024.
- 2023.07: 🎉🎉 Our paper “ALIF: Low-Cost Adversarial Audio Attacks on Black-Box Speech Platforms Using Linguistic Features” got accepted in IEEE S&P 2024.
- 2022.12: 🎉🎉 Our paper “InfoMasker: Preventing Eavesdropping using Phoneme-Based Noise” got accepted in NDSS 2023.
📝 Publications
SecHeadset: A Practical Privacy Protection System for Real-time Voice Communication
Peng Huang, Kun Pan, Qinglong Wang, Peng Cheng, Li Lu, Zhongjie Ba, Kui Ren
Paper | Slides | Code | Project Page
- SecHeadset is an end-to-end, plug-and-play privacy protection system for real-time voice communication based on voice obfuscation, which is compatible with various communication service providers such as Telegram and Skype.
InfoMasker: Preventing Eavesdropping Using Phoneme-Based Noise
Peng Huang, Yao Wei, Peng Cheng, Zhongjie Ba, Feng Lin, Li Lu, Fan Zhang, Kui Ren
Paper | Slides | Code | arXiv | Project Page
- InfoMasker is a highly effective anti-eavesdropping system. It jams nearby microphones with human-inaudible ultrasound noise, which can effectively mislead state-of-the-art ASR systems and is robust against various speech enhancement methods.
- Text-CRS: A Generalized Certified Robustness Framework against Textual Adversarial Attacks, Xinyu Zhang, Hanbin Hong, Yuan Hong, Peng Huang, Binghui Wang, Zhongjie Ba, Kui Ren, IEEE S&P 2024. | [Paper] [Code]
- ALIF: Low-Cost Adversarial Audio Attacks on Black-Box Speech Platforms Using Linguistic Features, Peng Cheng, Yuwei Wang, Peng Huang, Zhongjie Ba, Xiaodong Lin, Feng Lin, Li Lu, Kui Ren, IEEE S&P 2024.
🎓 Education
- Ph.D. student, School of Cyber Science and Technology, Zhejiang University. Supervisor: Prof. Zhongjie Ba.
- B.Eng., College of Information Science & Electronic Engineering, Zhejiang University.