Hongsheng Hu

Hongsheng Hu

School of Computer Science, Shanghai Jiao Tong University
AI Security and Privacy

I am an Associate Professor at School of Computer Science, Shanghai Jiao Tong University. Before joining Shanghai Jiao Tong University, From 2024 to 2026, I was a tenured Lecturer in the School of Information and Physical Sciences at the University of Newcastle, Australia. Prior to that, I was a Postdoctoral Research Fellow at CSIRO’s Data61. In 2022, I received my Ph.D. from the University of Auckland, New Zealand, under the supervision of Professor Xuyun Zhang, Professor Gillian Dobbie (Fellow of the Royal Society of New Zealand), and Professor Zoran Salcic (Fellow of the Royal Society of New Zealand).

My research focuses on trustworthy AI, with particular emphasis on AI privacy and LLM agent security. In particular, I am interested in understanding real privacy and security risks of deep models, developing methods to mitigate these risks, and designing secure and robust AI systems. My work has been published in top-tier venues such as IEEE S&P, NDSS, USENIX Security, ICLR, CVPR, NeurIPS, WWW, IJCAI, and AAAI.

Email Google Scholar

Prospective Postdocs, PhD, Master, and Graduate Intern Students

I am always looking for self-motivated talents interested in AI security and privacy, especially trustworthy AI, LLM agent security, model privacy, and robust AI systems. Please read the information below carefully before contacting me.

Postdoctoral Research Fellow

  • Applicants should have a strong research background aligned with AI Security, LLM, Agent, and a record of high-quality publications.
  • For information about postdoctoral salary and related support, please refer to the SJTU postdoctoral salary reference.

PhD and Master’s Students

  • For PhD and Master applicants, preference will be given to candidates who have completed a research internship in my lab at SJTU.

Please email me for the above positions. Your email should include: i) your CV and academic transcripts; ii) a brief description of your research interests and how they align with AI Security, LLM, Agent; and iii) your intended position, expected start date, and any relevant publications, research projects, or code repositories experience.

News

Two papers have been accepted to EMNLP 2026: Staying on Task: Defending Against Prompt Injection via Task Consistency Checking (Main Conference) and Latent Visual Sensitive Fingerprinting for Black-Box Vision–Language Model Tamper Detection (Findings). Congratulations to Feiyue and Chaoxiang!

We are so lucky to receive 🏆 Distinguished Artifact Awards at USENIX Security 2026 (CCF-A) for our paper CompLeak: Deep Learning Model Compression Exacerbates Privacy Leakage. Congratulations to the team!

Our paper DP2-RAG: An Efficient Full-Process Differential Privacy Implementation in Retrieval-Augmented Generation was accepted by IEEE Transactions on Information Forensics and Security (TIFS) (CCF-A). Congratulations to the team!

I was invited by the PC Chair to serve on the Program Committee for the IEEE Conference on Secure and Trustworthy Machine Learning (SaTML) 2027. Looking forward to your good work!

I was invited by the PC Chair to serve on the Program Committee for USENIX Security 2027 (CCF-A). Looking forward to your good work!

I joined School of Computer Science, Shanghai Jiao Tong University as an Associate Professor.