Hui Liu
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Hui Liu has a total of 33 co-authors. Here are the top co-authors by number of shared publications:
- Bo Zhao
School of Cyber Science and Engineering, Wuhan University, Wuhan, China
4 shared publications - Hongqin Du
School of Computer Science and Hubei Provincial Key Laboratory of Artificial Intelligence and Smart Learning, Central China Normal University, Wuhan, China
3 shared publications - Jiabao Guo
School of Computer Science, Hefei University of Technology, Hefei, China
3 shared publications - Fujv Wen
School of Computer Science and Hubei Provincial Key Laboratory of Artificial Intelligence and Smart Learning, Central China Normal University, Wuhan, China
2 shared publications Jiabao Guo
2 shared publications- Jiageng Chen
Central China Normal University
2 shared publications
Latest Publications
- Open Access
- 20 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 867-878
WARP: A Word-Level Backdoor Attack Targeting RAG Systems via Retrieval Corpus Poisoning
- 20 April 2026
- Published by Association for Computing Machinery (ACM)
- p. 867-878
Open AccessWARP: A Word-Level Backdoor Attack Targeting RAG Systems via Retrieval Corpus Poisoning
- 27 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Social Systems
- Vol. 13 (3) , 3190-3200
Take off Your Disguise: Detecting Disguised Prompt-Based Jailbreak Attacks Against LLMs
Large language models (LLMs) are typically equipped with alignment mechanisms designed to prevent the generation of harmful content. However, recent advances have led to the emergence of highly evasive disguised prompt jailbreak attacks (DPJAs), where attackers conceal real malicious intent within prompts that appear benign on the surface, thereby inducing the model to produce unsafe outputs. In this work, we propose prompt reconstruction-based detection (RePrompt), a training-free and plug-in detection framework designed to identify such jailbreak attacks. The key insight of RePrompt is that, when guided by carefully designed templates, LLMs possess the capability to uncover disguised adversarial prompts and effectively detect jailbreak attacks. RePrompt leverages the reasoning capabilities of the LLM itself to analyze and uncover the true intent behind a user’s query. When a disguised jailbreak prompt is encountered, RePrompt reconstructs the underlying malicious intent hidden beneath the surface-level disguise. Experiments across multiple LLMs, datasets, and three representative DPJAs demonstrate that RePrompt consistently outperforms state-of-the-art defense methods, reducing the average attack success rate from 47.0% to below 1.6%, achieving a near-zero false positive rate, and limiting the average query cost to only 1.13.
- 27 February 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Social Systems
- Vol. 13 (3) , 3190-3200
Take off Your Disguise: Detecting Disguised Prompt-Based Jailbreak Attacks Against LLMs
Large language models (LLMs) are typically equipped with alignment mechanisms designed to prevent the generation of harmful content. However, recent advances have led to the emergence of highly evasive disguised prompt jailbreak attacks (DPJAs), where attackers conceal real malicious intent within prompts that appear benign on the surface, thereby inducing the model to produce unsafe outputs. In this work, we propose prompt reconstruction-based detection (RePrompt), a training-free and plug-in detection framework designed to identify such jailbreak attacks. The key insight of RePrompt is that, when guided by carefully designed templates, LLMs possess the capability to uncover disguised adversarial prompts and effectively detect jailbreak attacks. RePrompt leverages the reasoning capabilities of the LLM itself to analyze and uncover the true intent behind a user’s query. When a disguised jailbreak prompt is encountered, RePrompt reconstructs the underlying malicious intent hidden beneath the surface-level disguise. Experiments across multiple LLMs, datasets, and three representative DPJAs demonstrate that RePrompt consistently outperforms state-of-the-art defense methods, reducing the average attack success rate from 47.0% to below 1.6%, achieving a near-zero false positive rate, and limiting the average query cost to only 1.13.
- Open Access
- 17 January 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Intelligent Systems and Technology
- Vol. 17 (1) , 1-19
Adversarial Face Database against Deep Learning-Enabled Reconstruction Attacks
Face recognition systems offer a range of applications that enhance security, efficiency, and personalization, e.g., access control, identity verification, and personalized services. Mainstream facial recognition systems employ the Edge-Cloud architecture to protect user privacy by storing facial feature data instead of original facial images. However, recently emerging reconstruction attacks based on deep learning can recover the visual information of original facial images from facial features, resulting in face privacy disclosure. Existing anti-reconstruction approaches either compromise facial recognition accuracy or fail to meet real-time requirements. In this article, we propose a practical privacy-preserving approach based on adversarial perturbations against reconstruction attacks. By incorporating subtle adversarial interference into facial features, the mapping relationship from facial features to original facial images is disrupted, and the baseline reconstruction networks cannot recover the original face image. We conducted experiments on two facial recognition models, FaceNet and ArcFace, both widely deployed in practical scenarios. The results show that the face recognition accuracy sacrifice of less than 1% can significantly reduce the quality of the reconstructed image. In terms of efficiency, the average time to generate an adversarial facial feature is less than 10 ms, meeting the real-time requirements of facial recognition.
- 17 January 2026
- Published by Association for Computing Machinery (ACM) in ACM Transactions on Intelligent Systems and Technology
- Vol. 17 (1) , 1-19
Open AccessAdversarial Face Database against Deep Learning-Enabled Reconstruction Attacks
Face recognition systems offer a range of applications that enhance security, efficiency, and personalization, e.g., access control, identity verification, and personalized services. Mainstream facial recognition systems employ the Edge-Cloud architecture to protect user privacy by storing facial feature data instead of original facial images. However, recently emerging reconstruction attacks based on deep learning can recover the visual information of original facial images from facial features, resulting in face privacy disclosure. Existing anti-reconstruction approaches either compromise facial recognition accuracy or fail to meet real-time requirements. In this article, we propose a practical privacy-preserving approach based on adversarial perturbations against reconstruction attacks. By incorporating subtle adversarial interference into facial features, the mapping relationship from facial features to original facial images is disrupted, and the baseline reconstruction networks cannot recover the original face image. We conducted experiments on two facial recognition models, FaceNet and ArcFace, both widely deployed in practical scenarios. The results show that the face recognition accuracy sacrifice of less than 1% can significantly reduce the quality of the reconstructed image. In terms of efficiency, the average time to generate an adversarial facial feature is less than 10 ms, meeting the real-time requirements of facial recognition.
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automation Science and Engineering
- Vol. 23, 16385-16399
Appareo: Runtime Reliability Inspection for Deep Learning-Based Image Recognition Systems
- 1 January 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Automation Science and Engineering
- Vol. 23, 16385-16399
Appareo: Runtime Reliability Inspection for Deep Learning-Based Image Recognition Systems
- Open Access
- 9 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Information Forensics and Security
- Vol. 20 (15566013) , 12903-12917
Patronus : Plug-and-Play and Near-Lossless Facial Privacy Enhancement Against Reconstruction Attacks
Reconstruction attackers can exploit facial features to recover the original user’s face, resulting in user privacy leakage. One new strategy to enhance the “Edge-Cloud” face recognition system’s privacy is to add adversarial perturbations to facial features, preventing the attackers from high-quality user image recovery. However, the existing works following this strategy suffer from unacceptable damage to face recognition accuracy. Achieving robust privacy enhancement and face recognition accuracy simultaneously is still challenging. To tackle this challenge, we propose an adversarial perturbation-based plug-and-play privacy-enhancing method (Patronus) with robustness against face image reconstruction attacks and near-lossless face recognition performance. The key insight is derived from our observation that the feature distance between two face images of the same person is significantly lower than the threshold set in the face recognition system. This leaves room for adding adversarial perturbations to the facial features without compromising face recognition accuracy. Our strategy limits the amount of adversarial perturbations in a fine-grained manner to ensure that they are within the range of not damaging face recognition accuracy. Our evaluation shows the superior performance of Patronus in robustness against reconstruction attacks and near-lossless face recognition accuracy compared to state-of-the-art (SOTA) methods. Patronus can be easily integrated into deployed face recognition systems as a plug-in privacy-enhancing module with low overhead.
- 9 October 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Information Forensics and Security
- Vol. 20 (15566013) , 12903-12917
Open AccessPatronus : Plug-and-Play and Near-Lossless Facial Privacy Enhancement Against Reconstruction Attacks
Reconstruction attackers can exploit facial features to recover the original user’s face, resulting in user privacy leakage. One new strategy to enhance the “Edge-Cloud” face recognition system’s privacy is to add adversarial perturbations to facial features, preventing the attackers from high-quality user image recovery. However, the existing works following this strategy suffer from unacceptable damage to face recognition accuracy. Achieving robust privacy enhancement and face recognition accuracy simultaneously is still challenging. To tackle this challenge, we propose an adversarial perturbation-based plug-and-play privacy-enhancing method (Patronus) with robustness against face image reconstruction attacks and near-lossless face recognition performance. The key insight is derived from our observation that the feature distance between two face images of the same person is significantly lower than the threshold set in the face recognition system. This leaves room for adding adversarial perturbations to the facial features without compromising face recognition accuracy. Our strategy limits the amount of adversarial perturbations in a fine-grained manner to ensure that they are within the range of not damaging face recognition accuracy. Our evaluation shows the superior performance of Patronus in robustness against reconstruction attacks and near-lossless face recognition accuracy compared to state-of-the-art (SOTA) methods. Patronus can be easily integrated into deployed face recognition systems as a plug-in privacy-enhancing module with low overhead.
- Open Access
- 28 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 994-1003
Style-conditional Prompt Token Learning for Generalizable Face Anti-spoofing
Face anti-spoofing (FAS) based on domain generalization (DG) has attracted increasing attention from researchers. The reason for the poor generalization is that the model is overfitted to salient liveness-irrelevant signals. However, the previous methods alleviate the overfitting by mapping the images from multiple domains into a common feature space or promoting the separation of image features from domain-specific features and task-related features. If the text features of vision-language pre-trained (VLP) models (e.g., CLIP) are used to dynamically adjust the image features to gain a better generalization, we can not only explore a wider feature space but also avoid the potential degradation of semantic information. Specifically, we propose a FAS method of Style-Conditional Prompt Token Learning (S-CPTL), which aims to generate generalized text features by training the introduced prompt tokens to carry visual styles and use them as weights for classifiers to improve the model's generalization. Compared to the inherently static prompt token, we propose the dynamic prompt token, which can adaptively capture live-irrelevant signals from the instance-specific styles and increase their diversity through mixed feature statistics to further reduce the overfitting of the model. Thorough experimental analysis demonstrates that S-CPTL exceeds current top-performing methods in four distinct cross-dataset benchmarks.
- 28 October 2024
- Published by Association for Computing Machinery (ACM)
- p. 994-1003
Open AccessStyle-conditional Prompt Token Learning for Generalizable Face Anti-spoofing
Face anti-spoofing (FAS) based on domain generalization (DG) has attracted increasing attention from researchers. The reason for the poor generalization is that the model is overfitted to salient liveness-irrelevant signals. However, the previous methods alleviate the overfitting by mapping the images from multiple domains into a common feature space or promoting the separation of image features from domain-specific features and task-related features. If the text features of vision-language pre-trained (VLP) models (e.g., CLIP) are used to dynamically adjust the image features to gain a better generalization, we can not only explore a wider feature space but also avoid the potential degradation of semantic information. Specifically, we propose a FAS method of Style-Conditional Prompt Token Learning (S-CPTL), which aims to generate generalized text features by training the introduced prompt tokens to carry visual styles and use them as weights for classifiers to improve the model's generalization. Compared to the inherently static prompt token, we propose the dynamic prompt token, which can adaptively capture live-irrelevant signals from the instance-specific styles and increase their diversity through mixed feature statistics to further reduce the overfitting of the model. Thorough experimental analysis demonstrates that S-CPTL exceeds current top-performing methods in four distinct cross-dataset benchmarks.
- 19 November 2023
- Published by Elsevier in Pattern Recognition
- Vol. 147
A lightweight unsupervised adversarial detector based on autoencoder and isolation forest
- 19 November 2023
- Published by Elsevier in Pattern Recognition
- Vol. 147
A lightweight unsupervised adversarial detector based on autoencoder and isolation forest
- 19 May 2022
- Published by Elsevier in Applied Soft Computing
- Vol. 124
Feature-filter: Detecting adversarial examples by filtering out recessive features
- 19 May 2022
- Published by Elsevier in Applied Soft Computing
- Vol. 124
Feature-filter: Detecting adversarial examples by filtering out recessive features
- 18 April 2022
- Published by Elsevier in Computers & Security
- Vol. 118, 102730
The triggers that open the NLP model backdoors are hidden in the adversarial samples
- 18 April 2022
- Published by Elsevier in Computers & Security
- Vol. 118, 102730
The triggers that open the NLP model backdoors are hidden in the adversarial samples
- Open Access
- 17 May 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 7, 65450-65459
A Remote-Sensing Image Encryption Scheme Using DNA Bases Probability and Two-Dimensional Logistic Map
Remote-sensing images contain lots of visual information about transportation systems, geomorphic conditions and communal facilities. The secure storage and transmission of remote-sensing images is of great significance. The paper proposes a novel image encryption scheme based DNA bases probability and applies it to remote-sensing images for data protection. First, the plain-image is randomly encoded in DNA rules and its result participates in DNA addition with DNA mask generated by twodimensional logistic map. Second, the cryptosystem executes two-dimensional logistic map again with DNA bases probability against differential attacks. Third, the pixel-level rearrangement and the DNA baselevel rearrangement are respectively implemented in order of chaos sequences for permutation and diffusion. The cipher-image has a uniform distribution, low coefficients, and ideal entropy. The cryptosystem has acceptable encryption speed ( speed = 0.651308Mb it/s ) and high sensitivity to the plainimage and the security key. The experimental results show the proposed algorithm can resist various existing attack schemes against remote-sensing images.
- 17 May 2019
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 7, 65450-65459
Open AccessA Remote-Sensing Image Encryption Scheme Using DNA Bases Probability and Two-Dimensional Logistic Map
Remote-sensing images contain lots of visual information about transportation systems, geomorphic conditions and communal facilities. The secure storage and transmission of remote-sensing images is of great significance. The paper proposes a novel image encryption scheme based DNA bases probability and applies it to remote-sensing images for data protection. First, the plain-image is randomly encoded in DNA rules and its result participates in DNA addition with DNA mask generated by twodimensional logistic map. Second, the cryptosystem executes two-dimensional logistic map again with DNA bases probability against differential attacks. Third, the pixel-level rearrangement and the DNA baselevel rearrangement are respectively implemented in order of chaos sequences for permutation and diffusion. The cipher-image has a uniform distribution, low coefficients, and ideal entropy. The cryptosystem has acceptable encryption speed ( speed = 0.651308Mb it/s ) and high sensitivity to the plainimage and the security key. The experimental results show the proposed algorithm can resist various existing attack schemes against remote-sensing images.