Security and Privacy Seminar Series: Dr. Umar Iqbal, "Building Reliable Systems from Unreliable Models."

Oct 23, 2026 1:00 pm

2013 Electrical and Computing Engineering Building

Seminar/Symposium

Sponsor: Siebel School of Computing and Data Science
Speaker: Dr. Umar Iqbal
Contact: Luyi Xing
E-Mail: lxing2@illinois.edu
Originating Calendar: Siebel School Speakers Calendar; Abstract: Generative AI has enabled a new agentic computing paradigm, in which systems rely on machine learning models to interpret and act on user requests expressed in natural language. While this paradigm promises powerful and flexible automation, systems built on it have serious security, privacy, and safety risks that make them difficult to trust in practice.

This talk argues that building reliable agentic systems is fundamentally an integrity problem. Rather than assuming that models will behave reliably, we must constrain their inputs, processing, and outputs to keep system behavior consistent with user intent and security policies. Through a series of case studies, I will show how systems-level guardrails can protect integrity across these stages, in diverse computing environments, enabling systems that provide strong security properties even when the underlying models remain unreliable.

Bio: Umar Iqbal is an Assistant Professor in the Department of Computer Science & Engineering at Washington University in St. Louis. His research focuses on computer security and privacy, with an emphasis on agentic systems. He is a recipient of an NSF CAREER Award (2026), Caspar Bowden Runner-Up Award for Outstanding Research in Privacy Enhancing Technologies (2024), ACM IMC Best Paper Award (2023), CRA/NSF Computing Innovation Fellowship (2021), and the University of Iowa Research Excellence Award (2021). You can find more about him on his research on his webpage: https://umariqbal.com

link for robots only