7 papers
VLMShield: Efficient and Robust Defense of Vision-Language Models against Malicious Prompts
Peigui Qi, Kunsheng Tang, Yanpu Yu +7
Vision-Language Models (VLMs) face significant safety vulnerabilities from malicious prompt attacks due to weakened alignment during visual integration. Existing defenses suffer fr…
Hidden Ads: Behavior Triggered Semantic Backdoors for Advertisement Injection in Vision Language Models
Duanyi Yao, Changyue Li, Zhicong Huang +2
Vision-Language Models (VLMs) are increasingly deployed in consumer applications where users seek recommendations about products, dining, and services. We introduce Hidden Ads, a n…
EnchTable: Unified Safety Alignment Transfer in Fine-tuned Large Language Models
Jialin Wu, Kecen Li, Zhicong Huang +3
Many machine learning models are fine-tuned from large language models (LLMs) to achieve high performance in specialized domains like code generation, biomedical analysis, and math…
Understanding the Dilemma of Unlearning for Large Language Models
Qingjie Zhang, Haoting Qian, Zhicong Huang +5
Unlearning seeks to remove specific knowledge from large language models (LLMs), but its effectiveness remains contested. On one side, "forgotten" knowledge can often be recovered…
MARS: A Malignity-Aware Backdoor Defense in Federated Learning
Wei Wan, Yuxuan Ning, Zhicong Huang +7
Federated Learning (FL) is a distributed paradigm aimed at protecting participant data privacy by exchanging model parameters to achieve high-quality model training. However, this…
GaussMarker: Robust Dual-Domain Watermark for Diffusion Models
Kecen Li, Zhicong Huang, Xinwen Hou +1
As Diffusion Models (DM) generate increasingly realistic images, related issues such as copyright and misuse have become a growing concern. Watermarking is one of the promising sol…
VideoMarkBench: Benchmarking Robustness of Video Watermarking
Zhengyuan Jiang, Moyang Guo, Kecen Li +5
The rapid development of video generative models has led to a surge in highly realistic synthetic videos, raising ethical concerns related to disinformation and copyright infringem…