Colloquium - Shuo Chen, "The Web-Verbs Abstraction -- A Crucial Semantic Layer for A Reliable and Secure Agentic Web"

Abstract: 
The agentic web envisions natural language as the new interface for interacting with the web, sparking significant excitement across the industry. One dimension of the effort is to build powerful agents that can see and parse webpages and act on GUI elements. Microsoft’s NLWeb initiative is calling for a complementary effort – to build a semantic layer around the “raw” web so that it is more amenable to agents.

 In this talk, I will present our team’s work within the NLWeb initiative. We advocate for a semantic layer called “Web Verbs” and highlight its importance for reliability and security in the agentic-web future. Web Verbs are a web-scale set of strongly typed functions that enable client-side and server-side interactions. They are created by website developers and retrievable through NLWeb’s RAG system. With the Web-Verbs abstraction, existing coding agents can reliably perform complex tasks. Moreover, this abstraction provides a solid programming-language foundation for rigorous security reasoning. I will describe our ongoing efforts to enable automatic and precise authorization for agents.

Bio:
Shuo Chen is a Senior Principal Researcher at Microsoft Research Redmond. Shuo’s work spans software-as-a-service, browsers, web privacy/security, and blockchain/smart contracts. His research has driven cross-company security initiatives in browsers, e-commerce, and web authentication, and has been featured by CNN, CNET, MIT Tech Review, and Ars Technica. Several of his technologies have been transferred to Microsoft Security Intelligence and Azure Blockchain groups. Shuo has served on program committees for IEEE S&P, USENIX Security, ACM CCS, and WWW. He earned his Ph.D. in Computer Science from the University of Illinois at Urbana-Champaign. His research received two Microsoft Gold Star awards and two paper awards in the IEEE S&P conference and the PETS Symposium.