TECoSA seminar – The Road to Trustworthy ML: From Security and Privacy to Verifiability

Abstract:

In this talk, Buse Atli will present the evolving landscape of security and privacy in machine learning (ML), drawing on insights from my own research and experiences in both academic work and real-world applications. She will describe the ongoing arms race between attacks and defenses in ML systems and discuss what these dynamics reveal about trust, governance, and accountability in ML. Finally, she will argue that verifiability is one of the most critical yet still unresolved aspects of achieving trustworthy ML, particularly in light of legal frameworks such as the EU AI Act.

Bio: Buse Atli is an Assistant Professor in the Cybersecurity Division at Linköping University. Previously, she was a security researcher at Nokia Bell Labs, developing threat modeling strategies for security and privacy in AI-enabled network systems. She received her PhD from Aalto University in Finland. Her research focuses on trustworthy machine learning, including robustness, data privacy, model confidentiality, verifiability, and AI governance.