Yunsong Huang

The LLM-RIMSA dataset, designed to advance 6G networks through ultra-massive connectivity and intelligent radio environments. The dataset is built around a novel framework that integrates large language models (LLMs) with a reconfigurable intelligent metasurface antenna (RIMSA) architecture. This integration addresses limitations in hardware efficiency, dynamic control, and scalability seen in existing RIS technologies.

The integration of Large Language Models (LLMs) into wireless communications for channel state information (CSI) prediction introduces transformative capabilities but also exposes critical security vulnerabilities, particularly backdoor attacks. This paper investigates how adversaries exploit the openness of wireless propagation where signals are inherently susceptible to eavesdropping and adversarial interference, and the black-box nature of neural networks to inject stealthy triggers (e.g., Gaussian white, narrowband, or impulse interference) into online training samples.