AI-Driven Autonomy in Cyber Conflict: Offense, Defence and Agentic Systems
CyCon 2026 — 18th International Conference on Cyber Conflict, CCDCOE
Academic Activities
Reviewing
Associate Editor — ICSR 2024, International Conference on Social Robotics
July – November 2024
Talks
Multi-Agent Reinforcement Learning
Cyber Alp Retreat, Davos · June 2025
Media
Feature: MIT Technology Review (German edition) – “Vergiftete Daten” (Data Poisoning) Issue 6/2026 | Author: Stefan Burkhardt — Featured on the growing threat of data poisoning to AI systems.
- Discussed the cat-and-mouse game between manipulation and detection, drawing on research at the Helmholtz Center for Information Security (CISPA) and the EU ELLIOT project on agentic, self-improving cybersecurity systems.
- Highlighted the security risks of “vibe coding,” the limits of agent-based safeguards, and the societal question of who decides which data AI models are trained on.
Interview: Deutschlandfunk (DLF) – Forschung Aktuell Computer und Kommunikation | March 2026 - Discussed the intersection of Generative AI and software security, highlighting the risks of “vibe coding” and AI-assisted development.
- Shared insights from research at CISPA and the ELLIOT Project regarding LLM-induced vulnerabilities, such as missing input validation and the use of outdated security standards.
- Listen to the full segment (German)
Teaching
Teaching Assistant – Machine Learning in Cybersecurity
Instructor: Mario Fritz | CISPA, 2025
- Leading tutorial sessions for the course Machine Learning in Cybersecurity
- Guiding students through exercises on ML applications in cybersecurity, including attacks, defenses, and AI for CTF challenges
- Supporting course logistics and student learning, both in-person and via Zoom