Bin Liu

The paper proposes H-Elo, a Fully Homomorphic Encryption (FHE)-based system that enables private and secure matchmaking by keeping user rating values encrypted during the traditional rating update process.

The paper introduces ImageProtector, a user-side method that embeds an imperceptible perturbation into images to prevent Multi-modal Large Language Models (MLLMs) from analyzing and extracting sensitive information from them.

This paper systematically studies the robustness of vision foundation models to common image perturbations, finding that most models are generally non-robust and proposing a fine-tuning method to improve this resilience.

The paper introduces Checkerboard, a novel, learning-free clean-label backdoor attack that efficiently poisons training data to compromise model integrity with minimal poisoning budget.

This paper presents the benefits of visual pretraining for foundation model intelligence, outperforming text-only pretraining on multiple backbones and benchmarks.