Cheng Hong

VLMShield: Efficient and Robust Defense of Vision-Language Models against Malicious Prompts

Fingerprinting LLMs via Prompt Injection

Understanding the Dilemma of Unlearning for Large Language Models

MARS: A Malignity-Aware Backdoor Defense in Federated Learning

GaussMarker: Robust Dual-Domain Watermark for Diffusion Models

FoundationForensics: Traceback Backdoor Attacks for Vision Foundation Models

MARS: A Malignity-Aware Backdoor Defense in Federated Learning