- PhD Student
I am a PhD student supervised by Prof. Robert Mullins and co-supervised by Dr. Yiren Zhao. I am funded by the Tazaki Cambridge Studentship and a member of Downing College. My research interest is in AI Security.
Biography
- Fall 2026: Recently I co-founded Chrysalia.
- Spring 2026: I was a part-time member of technical staff at AI Sequrity Company working on system security defenses against prompt injections (Dual-LLM), a startup founded by Ilia Shumailov and Yiren Zhao.
- Fall 2025: I was a student researcher at Google Deepmind working with the Synth-ID team.
- Summer 2025: I was a research intern at the Vector Institute under Prof. Nicolas Papernot.
- Fall 2024 ~ : I am currently a PhD student at the University of Cambridge.
- Summer 2024: I was a research intern in the System Security lab of TU Darmstadt under Prof. Ahmad-Reza Sadeghi.
- 2023-2024: I finished my MPhil in Advanced Computer Science under supervision of Prof. Ross Anderson, Prof. Robert Mullins and Dr. Ilia Shumailov.
- 2019-2023: I graduated with a BSc in Natural Sciences (Computer Science, Mathematics) from Durham University under supervision of Dr. Ehsan Toreini and Dr. Gagangeet Aujla. During this time I also studied abroad for one year from 2021-2022 at the Computer Science and Engineering Department of Seoul National University in South Korea.
Scholarships/Studentships/Grants
- Blue Dot Impact Rapid Grant (8/2026, ~10,000$ for work on system security defenses for Computer Use Agents, and prompt injection attacks on OpenClaw)
- Tazaki-Cambridge Studentship (10/2023 ~ )
- Scholarship of the German Academic Scholarship Foundation (6/2021 ~ 7/2024)
Teaching
- Data Science: Michaelmas 2024-25 for St. John's, Peterhouse and Girton (11 students)
- Algorithms 1: Lent 2024-25 for St. John's, Peterhouse and Girton (12 students)
- Algorithms 2: Lent 2024-25 for St. John's, Peterhouse and Girton (12 students)
Professional Activities
- Reviewer NeurIPS’25, NeurIPS’26, ICLR'27
- Women@CL Cambridge Volunteer
Publications
- Giulio Zingrillo, Hanna Foerster, Ilia Shumailov, Yiren Zhao, Robert Mullins. Securing Computer-Use Agents Against Branch Steering Attacks. Arxiv Preprint 10/2026
- Jonas Guan*, Tom Blanchard*, Hanna Foerster*, Hengrui Jia*, Gabriel Huang, Nicolas Papernot. AI Agents Enable Adaptive Computer Worms. Arxiv Preprint 06/2026
- Article website
- Featured in NY Times, Scientific American, Fortune, University of Toronto
- Hanna Foerster, Ilia Shumailov, Cheng Zhang, Yiren Zhao, Jamie Hayes, Robert Mullins. Quantamination: Dynamic Quantization Leaks Your Data Across the Batch. AI4GOOD & TAIGR Workshops @ICML'26
- Harsh Chaudhari, Ethan Rathbun, Hanna Foerster, Jamie Hayes, Matthew Jagielski, Milad Nasr, Ilia Shumailov, Alina Oprea. Thought-Transfer: Indirect Targeted Poisoning Attacks on Chain-of-Thought Reasoning Models. Arxiv Preprint 01/2026
- Hanna Foerster*, Tom Blanchard*, Kristina Nikolic, Ilia Shumailov, Cheng Zhang, Robert Mullins, Nicolas Papernot, Florian Tramèr, Yiren Zhao. CaMeLs Can Use Computers Too: System-Level Security for Computer Use Agents. The Fortieth Annual Conference on Neural Information Processing Systems (NeurIPS'26, AIWILD & CompLearn Workshops@ICML'26)
- Featured in the Cambridge Independent, Cambridge Festival Open Day
- Hanna Foerster, Ilia Shumailov, Yiren Zhao, Harsh Chaudhari, Jamie Hayes, Robert Mullins, Yarin Gal. Reasoning Introduces New Poisoning Attacks Yet Makes Them More Complicated. 4th IEEE Conference on Secure and Trustworthy Machine Learning (SaTML'26)
- Hanna Foerster, Sasha Behrouzi, Phillip Rieger, Murtuza Jadliwala, Ahmad-Reza Sadeghi. LightShed: Defeating Perturbation-based Image Copyright Protections. 34th USENIX Security Symposium (USENIX Security'25)
- Article website
- Featured in MIT Technology Review, on Cyber News, on The Register, on Fast Company, on the University of Cambridge website, etc.
- Cheng Zhang*, Hanna Foerster*, Robert Mullins, Yiren Zhao, Ilia Shumailov. Hardware and Software Platform Inference. International Conference on Machine Learning (ICML'25)
- Hanna Foerster, Robert Mullins, Ilia Shumailov, Jamie Hayes. Beyond Slow Signs in High-fidelity Model Extraction. Conference on Neural Information Processing Systems (NeurIPS'24)
Panels, Talks (excluding conferences where papers were presented)
- When Agents Become Weapons: the AI worm (London AI Safety & Security Meetup hosted by ARIA and Meta, 10/2026)
- When Agents Become Weapons (UK AISI, 08/2026)
- When Agents Become Weapons (Japanese AISI, 07/2026)
- When Agents Become Weapons (Japanese AI Security Research Centre, 07/2026)
- Agents as Targets and Agents as Weapons (Arm Cambridge, 06/2026)
- Can we Trust AI to Use Our Computers? (Cambridge Festival, 03/2026)
- Lightening Talk about Lightshed (Vector's 2025 ML Security & Privacy Workshop, 07/2025)
- Talk on Model Extraction Attacks (MCR Seminar Night Downing College, 01/2025)
- Talk on Model Extraction Attacks (NeurIPS meetup at Cambridge hosted by ai@cam, 12/2024)
- Student, PhD, Postdoc Panel Discussion (High-Tech Women 2024, 09/2024)
Contact
I respond to emails in English, German and Japanese.
日本語でのメールにも対応しております。

