6 papers
Cut Less, Fold More: Model Compression through the Lens of Projection Geometry
Olga Saukh, Dong Wang, Haris Šikić +2
Compressing neural networks without retraining is vital for deployment at scale. We study calibration-free compression through the lens of projection geometry: structured pruning i…
AgentTypo: Adaptive Typographic Prompt Injection Attacks against Black-box Multimodal Agents
Yanjie Li, Yiming Cao, Dong Wang +1
Multimodal agents built on large vision-language models (LVLMs) are increasingly deployed in open-world settings but remain highly vulnerable to prompt injection, especially throug…
Algorithmic Complexity Attacks on All Learned Cardinality Estimators: A Data-centric Approach
Yingze Li, Xianglong Liu, Dong Wang +4
Learned cardinality estimators show promise in query cardinality prediction, yet they universally exhibit fragility to training data drifts, posing risks for real-world deployment.…
Gradient-Free Adversarial Purification with Diffusion Models
Xuelong Dai, Dong Wang, Xiuzhen Cheng +1
Adversarial training and adversarial purification are two widely used defense strategies for enhancing model robustness against adversarial attacks. However, adversarial training r…
How Breakable Is Privacy: Probing and Resisting Model Inversion Attacks in Collaborative Inference
Rongke Liu, Youwen Zhu, Dong Wang +3
Collaborative inference (CI) improves computational efficiency for edge devices by transmitting intermediate features to cloud models. However, this process inevitably exposes feat…
Improving Transferable Targeted Attacks with Feature Tuning Mixup
Kaisheng Liang, Xuelong Dai, Yanjie Li +2
Deep neural networks (DNNs) exhibit vulnerability to adversarial examples that can transfer across different DNN models. A particularly challenging problem is developing transferab…