Graduate mentor's supervisor: Prof. Sihang Liu
Modern AI and machine learning systems are increasingly trained and deployed on distributed infrastructures consisting of multiple servers working together. While distributed computing enables larger models and faster processing, it also introduces new security challenges. Communication between nodes, shared resources, and distributed coordination mechanisms can create vulnerabilities that may not exist in single-machine systems. The goal of this project is to understand and evaluate security risks that arise when training or running AI/ML models in distributed environments. By identifying and studying these vulnerabilities, we can help build more secure and trustworthy AI systems.
In the short term, students will set up a distributed computing environment, select AI/ML models, and reproduce vulnerabilities reported in prior research papers. Team members will independently study different papers and perform experiments to determine whether the same vulnerabilities appear in our setup. In the medium term, students will investigate previously unexplored weaknesses in the system and develop proof-of-concept attacks. Long-term goals include implementing multiple attacks, evaluating their impact, and writing a paper. The project will involve substantial hands-on experimentation, system setup, testing, and analysis.
This project is suitable for students with familiarity in network, OS, and AI/ML. Students with an interest in security, distributed systems, or AI are encouraged to apply.