1 code implementation • 25 Mar 2024 • Siyuan Cheng, Guanhong Tao, Yingqi Liu, Guangyu Shen, Shengwei An, Shiwei Feng, Xiangzhe Xu, Kaiyuan Zhang, Shiqing Ma, Xiangyu Zhang
Backdoor attack poses a significant security threat to Deep Learning applications.
1 code implementation • 8 Feb 2024 • Guangyu Shen, Siyuan Cheng, Kaiyuan Zhang, Guanhong Tao, Shengwei An, Lu Yan, Zhuo Zhang, Shiqing Ma, Xiangyu Zhang
Large Language Models (LLMs) have become prevalent across diverse sectors, transforming human life with their extraordinary reasoning and comprehension abilities.
no code implementations • 8 Dec 2023 • Zhuo Zhang, Guangyu Shen, Guanhong Tao, Siyuan Cheng, Xiangyu Zhang
Instead, it exploits the fact that even when an LLM rejects a toxic request, a harmful response often hides deep in the output logits.
1 code implementation • 27 Nov 2023 • Shengwei An, Sheng-Yen Chou, Kaiyuan Zhang, QiuLing Xu, Guanhong Tao, Guangyu Shen, Siyuan Cheng, Shiqing Ma, Pin-Yu Chen, Tsung-Yi Ho, Xiangyu Zhang
Diffusion models (DM) have become state-of-the-art generative models because of their capability to generate high-quality images from noises without adversarial training.
1 code implementation • CVPR 2023 • Shiwei Feng, Guanhong Tao, Siyuan Cheng, Guangyu Shen, Xiangzhe Xu, Yingqi Liu, Kaiyuan Zhang, Shiqing Ma, Xiangyu Zhang
We show the effectiveness of our method on image encoders pre-trained on ImageNet and OpenAI's CLIP 400 million image-text pairs.
1 code implementation • 16 Jan 2023 • Siyuan Cheng, Guanhong Tao, Yingqi Liu, Shengwei An, Xiangzhe Xu, Shiwei Feng, Guangyu Shen, Kaiyuan Zhang, QiuLing Xu, Shiqing Ma, Xiangyu Zhang
Attack forensics, a critical counter-measure for traditional cyber attacks, is hence of importance for defending model backdoor attacks.
no code implementations • CVPR 2023 • QiuLing Xu, Guanhong Tao, Jean Honorio, Yingqi Liu, Shengwei An, Guangyu Shen, Siyuan Cheng, Xiangyu Zhang
It trains the clone model from scratch on a very small subset of samples and aims to minimize a cloning loss that denotes the differences between the activations of important neurons across the two models.
no code implementations • 29 Nov 2022 • Guanhong Tao, Zhenting Wang, Siyuan Cheng, Shiqing Ma, Shengwei An, Yingqi Liu, Guangyu Shen, Zhuo Zhang, Yunshu Mao, Xiangyu Zhang
We leverage 20 different types of injected backdoor attacks in the literature as the guidance and study their correspondences in normally trained models, which we call natural backdoor vulnerabilities.
1 code implementation • 23 Oct 2022 • Kaiyuan Zhang, Guanhong Tao, QiuLing Xu, Siyuan Cheng, Shengwei An, Yingqi Liu, Shiwei Feng, Guangyu Shen, Pin-Yu Chen, Shiqing Ma, Xiangyu Zhang
In this work, we theoretically analyze the connection among cross-entropy loss, attack success rate, and clean accuracy in this setting.
no code implementations • 18 Jun 2022 • Guanhong Tao, Yingqi Liu, Siyuan Cheng, Shengwei An, Zhuo Zhang, QiuLing Xu, Guangyu Shen, Xiangyu Zhang
As such, using the samples derived from our attack in adversarial training can harden a model against these backdoor vulnerabilities.
no code implementations • 11 Feb 2022 • Guangyu Shen, Yingqi Liu, Guanhong Tao, QiuLing Xu, Zhuo Zhang, Shengwei An, Shiqing Ma, Xiangyu Zhang
We develop a novel optimization method for NLPbackdoor inversion.
no code implementations • CVPR 2022 • Guanhong Tao, Guangyu Shen, Yingqi Liu, Shengwei An, QiuLing Xu, Shiqing Ma, Pan Li, Xiangyu Zhang
A popular trigger inversion method is by optimization.
1 code implementation • CVPR 2022 • Yingqi Liu, Guangyu Shen, Guanhong Tao, Zhenting Wang, Shiqing Ma, Xiangyu Zhang
Our results on the TrojAI competition rounds 2-4, which have patch backdoors and filter backdoors, show that existing scanners may produce hundreds of false positives (i. e., clean models recognized as trojaned), while our technique removes 78-100% of them with a small increase of false negatives by 0-30%, leading to 17-41% overall accuracy improvement.
no code implementations • 16 Mar 2021 • Yingqi Liu, Guangyu Shen, Guanhong Tao, Zhenting Wang, Shiqing Ma, Xiangyu Zhang
A prominent challenge is hence to distinguish natural features and injected backdoors.
1 code implementation • 9 Feb 2021 • Guangyu Shen, Yingqi Liu, Guanhong Tao, Shengwei An, QiuLing Xu, Siyuan Cheng, Shiqing Ma, Xiangyu Zhang
By iteratively and stochastically selecting the most promising labels for optimization with the guidance of an objective function, we substantially reduce the complexity, allowing to handle models with many classes.
no code implementations • 22 Jan 2020 • Ziyang Tang, Xiang Liu, Guangyu Shen, Baijian Yang
Aerial imagery has been increasingly adopted in mission-critical tasks, such as traffic surveillance, smart cities, and disaster assistance.
no code implementations • 6 Oct 2019 • Guangyu Shen, Chengzhi Mao, Junfeng Yang, Baishakhi Ray
Due to the inherent robustness of segmentation models, traditional norm-bounded attack methods show limited effect on such type of models.