Data Science Talk Series 

Smarter Doesn’t Mean Safer: Building Secure, Trustworthy Smart Cyber Physical Systems (Full online)

Abstract: Smart cyber-physical systems from homes and grids to farms and factories are becoming deeply integrated into our daily lives. But as these systems grow, so do concerns about data privacy, security, and compliance with regulations like GDPR and HIPAA, especially in distributed environments involving multiple smart agents. In this talk, I’ll present a policy-driven approach to

making CPS environments more secure, privacy-aware, and trustworthy. Our framework combines dynamic access control with contextual policies that govern data sharing, transformation, and compliance enforcement, all without needing to change application code. We also address the

challenge of misinformation in multi-agent systems by using lightweight truth maintenance techniques to detect and resolve false or conflicting data. This approach enables resilient and compliant CPS operations while maintaining the utility of data for cloud-based AI applications.

Bio of the speaker: Dr. Sai Sree Laya Chukkapalli is a Research Scientist at IBM Research, who works on security for large language model (LLM)-based agentic systems. Her research focuses on designing dynamic, context-aware access control mechanisms to secure interaction flows and safeguard

sensitive data within multi-agent AI frameworks. She earned her Ph.D. from the University of Maryland, Baltimore County (UMBC), where she developed privacy-preserving and resilient security models for autonomous AI agents operating in dynamic, real-world environments.