Photo of Philipp Normann

Philipp Normann

PhD student in the Security and Artificial Intelligence Lab (SAIL) at TU Wien, advised by Daniel Arp

About

I study how to measure, explain and improve what AI agents can do in cybersecurity. A central question in my work is whether a measured success reflects a real security effect, and whether it still holds when the setup changes. To answer this, I rely on outcomes that can be verified independently of the agent. Such outcomes can also serve as a training signal. In recent work, we used them to train a small open-weight model that runs locally. It solves most scenarios of an interactive security benchmark at a much lower cost than a large commercial model.

I joined TU Wien in 2025, and my position is funded by the WWTF project BREADS. Before that, I worked for seven years as a data scientist at OTTO in Hamburg, where I built real-time recommender and fraud detection systems. I also co-authored the company's first papers at ACM RecSys and ran its first Kaggle competition. I hold an MSc in IT Security and a BSc in Computer Science from FH Wedel.

News

Publications

Security and AI

Recommender systems

Datasets

Teaching

Supervision

Talks

Service

Subreviewer: USENIX Security (2026, 2027), ACM CCS (2026), ACSAC (2026), IEEE SaTML (2026, 2027), RAID (2025), DIMVA (2026), ACM AISec (2025, 2026), DIGHUM (2025).

Outreach: Hamburg coordinator of meetup.ai, the largest AI meetup in Hamburg with more than 2,000 members, 2019–2025.