Hide in thicket: Generating imperceptible and rational adversarial perturbations on 3d point clouds

T Lou, X Jia, J Gu, L Liu, S Liang… - Proceedings of the …, 2024‏ - openaccess.thecvf.com
Adversarial attack methods based on point manipulation for 3D point cloud classification
have revealed the fragility of 3D models yet the adversarial examples they produce are …

Vl-trojan: Multimodal instruction backdoor attacks against autoregressive visual language models

J Liang, S Liang, A Liu, X Cao - International Journal of Computer Vision, 2025‏ - Springer
Abstract Autoregressive Visual Language Models (VLMs) demonstrate remarkable few-shot
learning capabilities within a multimodal context. Recently, multimodal instruction tuning has …

Sibling-attack: Rethinking transferable adversarial attacks against face recognition

Z Li, B Yin, T Yao, J Guo, S Ding… - Proceedings of the …, 2023‏ - openaccess.thecvf.com
A hard challenge in develo** practical face recognition (FR) attacks is due to the black-
box nature of the target FR model, ie, inaccessible gradient and parameter information to …

Inducing high energy-latency of large vision-language models with verbose images

K Gao, Y Bai, J Gu, ST **a, P Torr, Z Li… - arxiv preprint arxiv …, 2024‏ - arxiv.org
Large vision-language models (VLMs) such as GPT-4 have achieved exceptional
performance across various multi-modal tasks. However, the deployment of VLMs …

Boosting transferability in vision-language attacks via diversification along the intersection region of adversarial trajectory

S Gao, X Jia, X Ren, I Tsang, Q Guo - European Conference on Computer …, 2024‏ - Springer
Vision-language pre-training (VLP) models exhibit remarkable capabilities in
comprehending both images and text, yet they remain susceptible to multimodal adversarial …

Improving fast adversarial training with prior-guided knowledge

X Jia, Y Zhang, X Wei, B Wu, K Ma… - IEEE Transactions on …, 2024‏ - ieeexplore.ieee.org
Fast adversarial training (FAT) is an efficient method to improve robustness in white-box
attack scenarios. However, the original FAT suffers from catastrophic overfitting, which …

Object detectors in the open environment: Challenges, solutions, and outlook

S Liang, W Wang, R Chen, A Liu, B Wu… - arxiv preprint arxiv …, 2024‏ - arxiv.org
With the emergence of foundation models, deep learning-based object detectors have
shown practical usability in closed set scenarios. However, for real-world tasks, object …

Jailbreak vision language models via bi-modal adversarial prompt

Z Ying, A Liu, T Zhang, Z Yu, S Liang, X Liu… - arxiv preprint arxiv …, 2024‏ - arxiv.org
In the realm of large vision language models (LVLMs), jailbreak attacks serve as a red-
teaming approach to bypass guardrails and uncover safety implications. Existing jailbreaks …

Revisiting backdoor attacks against large vision-language models

S Liang, J Liang, T Pang, C Du, A Liu… - arxiv preprint arxiv …, 2024‏ - arxiv.org
Instruction tuning enhances large vision-language models (LVLMs) but raises security risks
through potential backdoor attacks due to their openness. Previous backdoor studies focus …

Does few-shot learning suffer from backdoor attacks?

X Liu, X Jia, J Gu, Y Xun, S Liang, X Cao - Proceedings of the AAAI …, 2024‏ - ojs.aaai.org
The field of few-shot learning (FSL) has shown promising results in scenarios where training
data is limited, but its vulnerability to backdoor attacks remains largely unexplored. We first …