CISPA Helmholtz Center for Information Security

How can game theory help to safeguard software security in the AI era? CISPA researcher Christoph Landolt and his colleagues have created a bi-level Stackelberg game to calculate how the release of new frontier models could be sequenced strategically. More here: cispa.de/en/oracles-gambit

How can AI search across different data sources to find relevant information? In this episode of CISPA TL;DR, Philipp Christmann explores how search across heterogeneous sources works, and how it relates to search-oriented conversational AI. cispa.de/tldr-51 Episode in German.🇩🇪

Our CISPA Inaugural Lectures continue!🎓 Rayna Dimitrova and Krikamol Muandet celebrated their debut as newly appointed tenured Faculty, presenting research on reliable systems and AI reasoning under uncertainty. Congratulations to you both.

As one of the world’s leading centers for cybersecurity, CISPA was part of an official delegation of the @helmholtz.de Association to Japan. The visit provided a valuable platform for engaging with the Japanese research community, exchanging perspectives and exploring areas of mutual interest.

Watermarks are not equally robust! This is the key finding of a new study by CISPA researcher Michel Meintz from the SprintML Lab. Meintz investigated how well AI image watermarks survive the training of new generative models. Read the full article on our website: cispa.de/en/meintz-watermarks

Congratulations to CISPA-Faculty Dr. Rebekka Burkholz on being elected an ELLIS Scholar in “Geometric Deep Learning”. ELLIS connects leading AI researchers across Europe. We’re proud to see Rebekka represent CISPA and contribute to advancing AI research. 🇪🇺 @ellis.eu

Congrats to Dhekra Mahmoud, who has been recognized by the “L'Oréal-UNESCO For Women in Science – Young Talents France” program! We spoke to Dhekra about her research, her fascination for cryptography and the importance of female role models in computer science: cispa.de/en/interview...

What happens when one AI agent tries to steer an entire group? 🤖 CISPA researcher Samira Abedini and her colleagues show how stubbornness, trust, and network structure can turn collaboration between AI agents into a security risk. More: cispa.de/en/ai-agents

👋 Welcome to CISPA! We are very pleased to welcome four new Faculty to CISPA: Dr. @mboehme.bsky.social, Dr. @ggarbacea.bsky.social, Dr. Pierre Edward Meyer, and Dr. @maximilianewindl.bsky.social. It’s great to have you here—here’s to everything we’ll accomplish together. 🚀

🔐 WordPress plugins are a major attack surface — and traditional web fuzzers can struggle with their complex execution paths. CISPA researcher Matteo Leonelli and his colleagues developed THEMIS, a context-aware grey-box fuzzer tailored to WordPress plugins. Read the full story: cispa.de/en/themis

Congratulations to @krikamol.bsky.social on becoming tenured CISPA-Faculty! 🎉 In our interview, he talks about “Rational Intelligence,” AI that knows the limits of its knowledge, and why tenure gives him the freedom to pursue ambitious research questions. 👉 Read now: cispa.de/en/krikamol-tenure

🔎In the latest episode of CISPA TL;DR, Samira Abedini explains how a model originally developed to describe human opinion formation can surprisingly well explain how LLM agents influence each other—and where to find multi-agent AI systems in the real world today. 🎧Tune in: cispa.de/en/tldr-50

What a week for usable security and privacy! 🔐✨ SOUPS 2026 brought the community together in Hannover for four days of inspiring research, lively discussions, new connections and plenty of exchange. We were proud to support SOUPS 2026 as a main sponsor. Thanks to everyone who made this possible.

Can general-purpose AI systems deliver on their promises without compromising privacy, security, and safety? CISPA-Faculty Dr. @ana-mariacretu.bsky.social centers her research on this question. Learn more about her research agenda at cispa.de/en/ana-maria-cretu

Next stop: Baltimore! 🇺🇸 From August 12–14, the security research community is coming together for the 35th USENIX Security Symposium in Baltimore—and CISPA researchers are part of the program with several papers. Find the full list of our USENIX Security 2026 papers here: cispa.de/en/usenix-2026

Ali Mustafa presents LEAKYLINKS, the first systematic study of how public URL scanning feeds can unintentionally expose sensitive access links. The study shows that privacy risks may emerge as an unintended consequence of combining otherwise beneficial security mechanisms. cispa.de/en/leaky-links

Paris, Vienna, Stockholm, Munich, Barcelona, Warsaw – and finally, home. The CISPA European Cybersecurity & AI Hackathon Championship wrapped up its first edition with a Grand Finale at CISPA. A journey worth watching. youtu.be/BXij1Vmycp4

Since 2025, Nina Altmaier has been strengthening CISPA as Head of Alumni Engagement. While completing her PhD, she began establishing our alumni program. Now, she is building a strong network that connects alumni, researchers, and the CISPA community. Read more: www.linkedin.com/pulse/5-ques...

Whether in the cinema🍿 or through exchange with the next generation of scientists🎓 – July showed the many different ways research at CISPA can be experienced. Thank you to everyone involved and all visitors who made these events possible.

After six regional qualifiers across Europe, the first CISPA European Cybersecurity & AI Hackathon Championship reached its conclusion in early July. 🎥 Watch some of the highlights from the championship final in the video below. youtu.be/RIb8MHICsPY

CISPA Researcher Warns Against One-Size-Fits-All Age Verification Online We spoke with Dr. Wouter Lueks about the findings of his recent study on age assurance technologies, their effectiveness, and trade-offs for privacy and civil liberties. 🔗 Read the full interview: cispa.de/en/interview-lueks

Can removing images of children from AI training datasets prevent text-to-image models from generating CSAM? New research by CISPA-Faculty Dr. @ana-mariacretu.bsky.social shows that dataset filtering provides only limited protection & can be circumvented. cispa.de/en/cretu-dataset-filtering

We're ready for ICML 2026! 🚀 From July 6–11, we'll be in Seoul at the 43rd International Conference on Machine Learning (ICML 2026), where our researchers will present numerous papers. Find an overview of all our ICML 2026 papers here: cispa.de/en/icml-2026

Congrats to @andreaszeller.bsky.social for receiving his 11th Impact Paper Award at FSE 2026 in Montréal today! He is honored for the SZZ-algorithm, which he presented together with Jacek Śliwerski and Thomas Zimmermann in 2005. Read our interview with Andreas at cispa.de/en/interview...

What a great week for the wireless security community🔒📡 CISPA had the pleasure of hosting the 19. @acm.org WiSec 2026 in Saarbrücken. Special thanks go to our CISPA Faculty, Mridula Singh and Nils Ole Tippenhauer, who served as the General Chairs of WiSec. We were delighted to welcome the community!

Team from Sweden Takes the Title at the First Hackathon Championship After 24 hours of coding and scientific creativity, the first-ever CISPA European Cybersecurity & AI Hackathon Championship has come to a successful close. Congratulations to the winning teams! cispa.de/en/hackathon...

Inside the Research Ethics Office at CISPA: In our latest interview, Thorsten Helfer and Jannik Zeiser talk about their academic roots in philosophy, their role and remit at the Research Ethics Office, and the challenges of applied ethics. Find the interview at cispa.de/en/interview...

How promising—and dangerous—is AI in Cybersecurity? This question was discussed in the workshop “AI in Cybersecurity: Offensive Capabilities, Impacts, and Response Strategies” organised by CISPA-Faculty Mario Fritz. Recordings of the talks are now online! Have a look: youtu.be/waqc5HA676U

CISPA now has its very own Research Ethics Review Board! It’s a resource for all our researchers and provides a structured process for the reflection of ethical challenges. Read more about how we approach research ethics at cispa.de/en/research-...