Email
Google Scholar
GitHub
LinkedIn01
Security & Privacy of Emerging AI Systems
AI Agents · Computer-Use Systems · World Models · 4D Models
Security and privacy risks in emerging AI systems with increasingly autonomous interaction and environmental understanding.
AI Security & Privacy
Hanrui Wang
Project Assistant Professor · Echizen Laboratory
National Institute of Informatics (NII), Tokyo, Japan
I am a Project Assistant Professor in the Echizen Laboratory at the National Institute of Informatics (NII), Tokyo, Japan. I received my Ph.D. in Computer Science from Monash University and my B.S. in Electronic Information Engineering from Northeastern University, China. Before entering academia, I worked in the IT industry and held a director-level position.
My research focuses on AI security and privacy, spanning adversarial machine learning, security of emerging AI systems, and AI evaluation and governance.
Research Highlights
02
Adversarial Machine Learning
Adversarial Robustness · Attack Transferability · Robustness Evaluation · Model Inversion
Understanding, evaluating, and exposing security and privacy risks in machine learning systems.
03
AI Evaluation & Governance
Scientific Peer Review · AI-Assisted Evaluation · AI Governance
Reliable and accountable use of AI in consequential evaluation and decision-making.
News
-
Our position paper “Peer Review Should Constrain Evaluative Authority” was accepted to NeurIPS 2026 (Position Track).
-
We are recruiting interns for the NII International Internship Program (2nd call, 2026). Eligible applicants are master’s or Ph.D. students at an institution with an MOU with NII. If you are interested and eligible, please contact me with your CV.
-
Our paper “DiffMI: Breaking Face Recognition Privacy via Diffusion-Driven Training-Free Model Inversion” was published in IEEE TIFS.
-
Our paper “An 8-Way Taxonomy for Multimodal Disinformation and Detection Benchmark” appeared at WWW 2026.
-
Our paper “GreedyPixel: Fine-Grained Black-Box Adversarial Attack via Greedy Algorithm” was published in IEEE TIFS.
-
Our paper “Iterative Window Mean Filter: Thwarting Diffusion-Based Adversarial Purification” was published in IEEE TDSC.