Research
Publications
Characterizing LLM Latent Space with Persistent Homology
Aideen Fay, Inés García-Redondo, Qiquan Wang, Haim Dubossarsky, Anthea Monod
Proceedings of the Fourteenth International Conference on Learning Representations (ICLR), 2026. Oral.
Get My Drift? Catching LLM Task Drift with Activation Deltas
Sahar Abdelnabi, Aideen Fay, Giovanni Cherubin, Ahmed Salem, Mario Fritz, Andrew Paverd
IEEE Conference on Secure and Trustworthy Machine Learning (SaTML), 2025, pp. 43–67.
Preprints
LLMail-Inject: A Dataset from a Realistic Adaptive Prompt Injection Challenge
Sahar Abdelnabi, Aideen Fay, Ahmed Salem, Egor Zverev, Kai-Chieh Liao, Chi-Huang Liu, Chun-Chih Kuo, Jannis Weigend, Danyael Manlangit, Alex Apostolov, Haris Umair, João Donato, Masayuki Kawakita, Athar Mahboob, Tran Huu Bach, Tsun-Han Chiang, Myeongjin Cho, Hajin Choi, Byeonghyeon Kim, Hyeonjin Lee, Benjamin Pannell, Conor McCauley, Mark Russinovich, Andrew Paverd, Giovanni Cherubin.
Preprint.
Datasets
LLMail-Inject from the large scale Microsoft adaptive prompt injection challenge. Hugging Face.
TaskTracker: LLM Activation Dataset for Indirect Prompt Injection and Task Drift. By request.
Media & Blogs
- Announcing the Winners of the Adaptive Prompt Injection Challenge (LLMail-Inject), Microsoft Security Response Center, 2025
https://www.microsoft.com/en-us/msrc/blog/2025/03/announcing-the-winners-of-the-adaptive-prompt-injection-challenge-llmail-inject - Announcing the Adaptive Prompt Injection Challenge (LLMail-Inject), Microsoft Security Response Center, 2024
https://www.microsoft.com/en-us/msrc/blog/2024/12/announcing-the-adaptive-prompt-injection-challenge-llmail-inject/ - Azure Kubernetes Service (AKS) Threat Hunting, Microsoft Security Response Center, 2023
https://www.microsoft.com/en-us/msrc/blog/2023/03/azure-kubernetes-service-aks-threat-hunting - Configuring Host-Level Audit Logging for AKS VMSS, Microsoft Security Response Center, 2023
https://www.microsoft.com/en-us/msrc/blog/2023/03/configuring-host-level-audit-logging-for-aks-vmss