Dr. Xiaoyong (Brian) Yuan Research Focus
Research Focus: Dr. Xiaoyong (Brian) Yuan, Assistant Professor,
Department of Electrical and Computer Engineering, Clemson University
Dr. Xiaoyong (Brian) Yuan is an Assistant Professor in the Department of Electrical and Computer Engineering at Clemson University. His research expertise spans the fields of artificial intelligence, cybersecurity, and edge computing, with a particular focus on the development of trustworthy and efficient AI-driven systems. Dr. Yuan’s research has been supported by multiple grants from the National Science Foundation (NSF) and industry award. His contributions have been recognized through several awards, including the ORAU Ralph E. Powe Junior Faculty Enhancement Award (2022), the Michigan Tech ICC Achievement Award (2022), and the IEEE CIS TETCI Outstanding Paper Award (2025). Dr. Yuan has been serving as an associate editor for IEEE Transactions on Neural Networks and Learning Systems (TNNLS) since 2022.
Dr. Yuan is a PI of a SC EPSCoR Grants for Applications in Industry and Networking (GAIN) project and is very active in Goal 3 of the AI-Enabling Core of the ADAPT in SC Project, focusing on implementing security and testing robustness in AI-enabled biomedical devices in collaboration with Dr. Qi Wang. Dr. Yuan’s research centers on advancing trustworthy and efficient artificial intelligence. Dr. Yuan investigates vulnerabilities in Retrieval-Augmented Generation (RAG), a Large Language Model (LLM)-based technique increasingly used in medical decision support systems to retrieve relevant clinical knowledge.
His team recently uncovered a novel Adversarial Instructional Prompt (AIP) attack, showing that instructional prompts, often used to guide AI models and improve retrieval quality, can be subtly manipulated to promote biased or harmful outputs. For instance, an adversarial prompt could covertly influence the system to suggest a specific medication regardless of clinical appropriateness. The AIP attack leverages diverse query generation and genetic algorithm-based optimization to maximize attack robustness on diverse user queries while maintaining naturalness and clean-task utility. This work underscores the need for robust prompt-level auditing to ensure the safety and reliability of AI-assisted medical applications.
Dr. Yuan’s team has developed an innovative framework that enables the efficient fine-tuning of large-scale foundation models for domain-specific applications, such as healthcare, across resource-constrained devices. By intelligently orchestrating operations across attention modules based on device-specific computing and memory profiles, the framework significantly reduces both computational and communication overhead without compromising predictive accuracy. The design incorporates selective execution of forward and backward passes in fine-tuning, guided by Fisher information and weight magnitude metrics, and optimizes workload distribution via a hierarchical multi-knapsack scheduling algorithm. This approach addresses challenges in deploying AI models on edge and embedded systems.