معرفی
Xiaojun Xu is a Professor at the School of Computer Science, Beijing Institute of Technology, with a prolific research career spanning machine learning security, medical AI, and robotics. Their work demonstrates strong interdisciplinary collaboration across computer science, healthcare, and engineering domains.
- Institution: Beijing Institute of Technology, School of Computer Science
- Research Focus: AI security, medical imaging, robotics, and remote sensing applications
- Collaborations: Extensive work with Bo Li (29 papers), Dawn Song (11 papers), and medical researchers
Xu's research interests center on adversarial machine learning, with significant contributions to model security, backdoor detection, and LLM unlearning. They've pioneered techniques like Meta Neural Analysis for Trojan detection and developed frameworks for certified robustness. Their medical imaging work focuses on quantitative susceptibility mapping for neurodegenerative diseases, particularly Parkinson's and Alzheimer's. In robotics, they've advanced control systems for quadruped and amphibious vehicles.
Recent publications reveal a strong trend toward large language model security, with multiple 2024-2025 papers on machine unlearning and watermarking techniques. Their work bridges theoretical security with practical healthcare applications, particularly in medical image analysis where they've developed tools for subcortical nucleus segmentation and brain age prediction.
- Key venues: NeurIPS, CCS, IEEE S&P, NeuroImage, IEEE Transactions
- Research impact: High citation count with consistent top-tier publication record
Xu has secured significant research funding, evidenced by the volume and diversity of publications across multiple domains. Their work on blockchain-enabled IoT systems and RAFT-based private blockchain demonstrates expertise in distributed systems. The medical imaging research shows strong hospital collaborations, particularly in developing tools for Parkinson's diagnosis. Current projects appear focused on LLM security challenges and multimodal medical AI systems with potential clinical applications.
