Dissertation Talk: Privacy and Security in the Era of Data: from Data Processing to Intelligent Agents

Event details Date: Wednesday, July 15, 2026 Time: 1:00 PM to 1:00 PM Location: Zoom Type: Lecture / Workshop Audience: Faculty,Students About this event Advances in data-driven computing and artificial intelligence have created growing demands for secure, privacy-preserving, and trustworthy collaboration. As modern computing evolves from collaborative data processing toward increasingly autonomous intelligent agents, new challenges emerge in system design, evaluation, and deployment. Collaboration across distributed data sources requires privacy-preserving mechanisms, including homomorphic encryption (FHE), secure multi-party computation (MPC), federated learning (FL), and differential privacy (DP), that are both efficient and practical to deploy. Efficient use of FHE enables practical private aggregation over large public databases, while MPC allows data alignment and collaborative computation without compromising individual privacy, supporting applications such as contact tracing. The diversity of data partitioning schemes, data modalities, and application requirements further calls for appropriate choices of FL algorithms and system frameworks. Bridging the gap between cryptographic protocols and real-world deployment, however, requires not only efficient building blocks but also programming abstractions and system designs that enable decentralized collaborative applications without relying on trusted third parties. As large language models continue to advance and intelligent agents become increasingly prevalent, the use of data is shifting from conventional query processing toward semantic understanding and interactive reasoning, extending data collaboration into a new paradigm. This transition introduces new challenges in evaluating LLMs and agent systems, driving the rapid development of benchmarks, datasets, and evaluation methodologies. Open, reproducible, and standardized evaluation has become increasingly important for measuring technological progress, yet becomes substantially more challenging when multiple data providers, execution environments, and interacting agents are involved. Addressing these challenges similarly calls for simple yet general abstractions and system... Official event details: https://events.berkeley.edu/eecs/event/324319-dissertation-talk-privacy-and-security-in-the-era