ARC@ORU The (Dark) Interplay of AI and cybersecurity
24 juni 2026 09:30 – 10:00 Visual Lab, ARC and Remotely
ARC inbjuder till ett forskningsseminarium med Alberto Castagnaro, cybersäkerhetsingenjör på SpritzMatter, en spinoff från Universitetet i Padova.
This is the second ARC@ORU seminar on June 24. Welcome to join the talk at 09.00, Misleading LLMs used in scientific peer reviewing.
About the seminar
Artificial Intelligence is rapidly transforming a wide range of fields, but its rise also raises critical security questions. Its dual-use nature opens a new and often misunderstood threat landscape. Can AI be misused for malicious purposes? Can AI systems themselves be hacked or manipulated? This talk cuts through the terminological confusion in AI security, drawing a clear line between Adversarial AI , Offensive AI and the security of AI Systems. After establishing this conceptual map, the talk grounds these ideas in two pieces of original offensive-security research: one exposing how it is possible to stealthily poison the document-ingestion stage of Retrieval-Augmented Generation (RAG) pipelines, and another reframing directory brute-forcing as a language-modeling problem. Together, the case studies show how AI is simultaneously a new attack surface and a powerful new attack vector for modern cybersecurity.
Speaker
Alberto Castagnaro is a cybersecurity engineer at SpritzMatter, a spin-off of the University of Padova. He earned his MSc in Computer Science, specializing in cybersecurity, from TU Delft in 2024, and has been a research fellow with the SPRITZ group at the University of Padova. His research focuses on AI red teaming and automating vulnerability assessment and penetration testing through Artificial Intelligence.