News

Dongyan Xu Presents Hybrid Twin Model for Freight Rail Defense at DCS-CI 2026

September 2026 — At the Design of Cyber-Secure Critical Infrastructure workshop (DCS-CI 2026) in Buford, GA, Professor Dongyan Xu presented pioneering research from his CERIAS team on safeguarding critical transport infrastructure against next-generation, AI-driven threats. His presentation, titled "A Hybrid Twin Model for Evaluating Protections in Freight Rail against LLM-Based Attacks," addressed the growing intersection between Large Language Models (LLMs) and physical infrastructure security.

ACTION Summer Research Experience: University of Washington 2026

Seattle, Washington -- the University of Washington (UW) served as one of the host sites for the ACTION Institute's Summer Research Experience, advancing critical research into the economic and structural dimensions of cybersecurity.  Under the supervision of Professor Radha Poovendran, student intern Brian Lee from the University of Chicago conducted research investigating how cyber insurance design impacts corporate security investments.  

Rutgers Researchers Advance Multi-Agent System Theory in Social Network Learning

New Brunswick, NJ -- During Summer 2026, undergraduate researchers Eli Yablon (Princeton), John Wu (Princeton), and Roger Chen (Duke) completed an 8-week research experience at Rutgers University under the supervision of Professor Jie Gao. Conducted in conjunction with the Center for Discrete Mathematics and Theoretical Computer Science (DIMACS), the project was mentored by Professor Gao and PhD student Daniel Baumgartner, with funding support from ACTION.  

2026 Summer Research in ACTION at Georgia Tech

Atlanta, Georgia -- Machine learning models used for cyber-threat detection frequently fail when moved to new environments, such as different network configurations or newer timeframes. To overcome this instability, Bell Chen (University of Washington) joined Prof. Xiaoming Huo's lab at the Georgia Institute of Technology to design a framework that verifies whether security models can adapt reliably across shifting domain conditions. Prof. João P. Hespanha (UC Santa Barbara) also collaborated in this Summer Research Experience.

NeurIPS workshop on AI Evaluation

ACTION Faculty member Bin Yu (UC Berkeley) is an invited panelist at a new workshop at NeurIPS. We include information from the workshop webpage below. Find more information or submit a paper here. 

U.S. Department of Homeland Security - Higher Education Opportunities

Homeland Security Professional Opportunities for Student Workforce to Experience (HS-POWER) - Applications Open October 2026

Every June, DHS S&T welcomes students from across the U.S. and its territories to the ten-week HS-POWER internship program that is geared toward undergraduate and graduate students in the science, technology, engineering and mathematics fields and political and social sciences. The application period will run from October 2026 through January 2027.

ACTION Benchmarks Used in Industry

Yuetai Li, a PhD student at the University of Washington, who is working with Prof. Radha Poovendran and ACTION undergraduates Brian Lee from the University of Chicago and Kayla Xu from Northwestern, helped build two AI evaluation benchmarks: JobBench and Agents' Last Exam. Today, both of those benchmarks were name-dropped in the same-day flagship model launches from Meta and OpenAI.

Inaugural Workshop on Agentic AI in Offensive and Defensive Cyber Operations (AIDC'26)

ACTION researchers have organized the inaugural Workshop on Agentic AI in Offensive and Defensive Cyber Operations (AIDC'26), to be held in conjunction with the 2026 Annual Computer Security Applications Conference (ACSAC 2026) in December in Los Angeles, CA.