检索结果-内蒙古大学图书馆

Clean-label Attack on Face Authentication Systems Through Rolling Shutter Mechanism

IEEE SIGNAL PROCESSING LETTERS 2025年 32卷 36-40页

作者： Wang, Yufei Li, Haoliang Zhang, Liepiao Hu, Yongjian Kot, Alex C. Nanyang Technol Univ Rapid Rich Object Search ROSE Lab Singapore 639798 Singapore City Univ Hong Kong Dept Elect Engn Kowloon Tong Hong Kong Peoples R China GRGTally vis IT Co Ltd Guangzhou 510663 Peoples R China South China Univ Technol Guangzhou 510640 Peoples R China

We introduce a novel clean-label black-box face presentation attack on face authentication systems, i.e., face recognition and verification systems, under mild conditions. Different from other clean-label attacks which require inserting complicated or intensity patterns after the image-capturing phase, our designed pattern can be automatically inserted during the exposure by utilizing the rolling shutter mechanism and modulating environment LEDs in a specialized waveform. This method provides a potential way to conduct backdoor attacks in the physical domain. Additionally, we propose an optimization strategy based on evolutionary computing to optimize the parameters of the stripe patterns, enhancing the attack success rate. The experimental results on several face recognition models and face verification services provided by the leading technology companies demonstrate the effectiveness of our attack method. Our study reveals a new attack applicable in the physical world, highlighting significant security concerns for existing face recognition, verification, and face anti-spoofing techniques.

关键词： Face recognition Light emitting diodes Image color analysis Faces Face detection Data models Computational modeling Closed box Web sites Video on demand Clean-label black-box attack evolutionary computing face recognition/verification rolling shutter

来源：评论

学校读者我要写书评

暂无评论

Robust and Transferable Backdoor Attacks Against Deep Image Compression With Selective Frequency Prior

引用

IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 2025年第3期47卷 1674-1693页

作者： Yu, Yi Wang, Yufei Yang, Wenhan Guo, Lanqing Lu, Shijian Duan, Ling-Yu Tan, Yap-Peng Kot, Alex C. Nanyang Technol Univ Interdisciplinary Grad Programme Rapid Rich Object Search ROSE Lab Singapore 639798 Singapore Nanyang Technol Univ Sch Elect & Elect Engn Singapore 639798 Singapore Pengcheng Lab Shenzhen 518066 Peoples R China Nanyang Technol Univ Sch Comp Sci & Engn Singapore 639798 Singapore Peking Univ Sch Comp Sci Beijing 100871 Peoples R China

Recent advancements in deep learning-based compression techniques have demonstrated remarkable performance surpassing traditional methods. Nevertheless, deep neural networks have been observed to be vulnerable to backdoor attacks, where an added pre-defined trigger pattern can induce the malicious behavior of the models. In this paper, we propose a novel approach to launch a backdoor attack with multiple triggers against learned image compression models. Drawing inspiration from the widely used discrete cosine transform (DCT) in existing compression codecs and standards, we propose a frequency-based trigger injection model that adds triggers in the DCT domain. In particular, we design several attack objectives that are adapted for a series of diverse scenarios, including: 1) attacking compression quality in terms of bit-rate and reconstruction quality;2) attacking task-driven measures, such as face recognition and semantic segmentation in downstream applications. To facilitate more efficient training, we develop a dynamic loss function that dynamically balances the impact of different loss terms with fewer hyper-parameters, which also results in more effective optimization of the attack objectives with improved performance. Furthermore, we consider several advanced scenarios. We evaluate the resistance of the proposed backdoor attack to the defensive pre-processing methods and then propose a two-stage training schedule along with the design of robust frequency selection to further improve resistance. To strengthen both the cross-model and cross-domain transferability on attacking downstream CV tasks, we propose to shift the classification boundary in the attack loss during training. Extensive experiments also demonstrate that by employing our trained trigger injection models and making slight modifications to the encoder parameters of the compression model, our proposed attack can successfully inject multiple backdoors accompanied by their corresponding triggers int

关键词： Image compression backdoor attack frequency trigger deep neural network resistance attack transferability Image compression backdoor attack frequency trigger deep neural network resistance attack transferability

来源：评论

学校读者我要写书评

暂无评论

Toward Model Resistant to Transferable Adversarial Examples via Trigger Activation

引用

IEEE TRANSACTIONS ON INFORMATION FORENSICS AND SECURITY 2025年 20卷 3745-3757页

作者： Yu, Yi Xia, Song Lin, Xun Kong, Chenqi Yang, Wenhan Lu, Shijian Tan, Yap-Peng Kot, Alex C. Nanyang Technol Univ Interdisciplinary Grad Programme Rapid Rich Object Search ROSE Lab Singapore 639798 Singapore Nanyang Technol Univ Sch Elect & Elect Engn Singapore 639798 Singapore Beihang Univ Sch Comp Sci & Engn Beijing 100191 Peoples R China Pengcheng Lab Shenzhen 518000 Peoples R China Nanyang Technol Univ Coll Comp & Data Sci Singapore 639798 Singapore

Adversarial examples, characterized by imperceptible perturbations, pose significant threats to deep neural networks by misleading their predictions. A critical aspect of these examples is their transferability, allowing them to deceive unseen models in closed-box scenarios. Despite the widespread exploration of defense methods, including those on transferability, they show limitations: inefficient deployment, ineffective defense, and degraded performance on clean images. In this work, we introduce a novel training paradigm aimed at enhancing robustness against transferable adversarial examples (TAEs) in a more efficient and effective way. We propose a model that exhibits random guessing behavior when presented with clean data x as input, and generates accurate predictions when with triggered data x +tau. Importantly, the trigger tau remains constant for all data instances. We refer to these models as models with trigger activation. We are surprised to find that these models exhibit certain robustness against TAEs. Through the consideration of first-order gradients, we provide a theoretical analysis of this robustness. Moreover, through the joint optimization of the learnable trigger and the model, we achieve improved robustness to transferable attacks. Extensive experiments conducted across diverse datasets, evaluating a variety of attacking methods, underscore the effectiveness and superiority of our approach.

关键词： Robustness Accuracy Training Data models Predictive models Computational modeling Perturbation methods Closed box Neurons Optimization Adversarial robustness transferable adversarial examples

来源：评论

学校读者我要写书评

暂无评论

Fragrant: frequency-auxiliary guided relational attention network for low-light action recognition

引用

VISUAL COMPUTER 2025年第2期41卷 1379-1394页

作者： Liu, Wenxuan Jia, Xuemei Ju, Yihao Ju, Yakun Jiang, Kui Wu, Shifeng Zhong, Luo Zhong, Xian Wuhan Univ Technol Sch Comp Sci & Artificial Intelligence Hubei Key Lab Transportat Internet Things Wuhan 430070 Peoples R China Wuhan Univ Sch Comp Sci Wuhan 430072 Peoples R China Wuhan Traff Management Bur Wuhan 430024 Peoples R China Nanyang Technol Univ Sch Elect & Elect Engn Rapid Rich Object Search Lab Singapore 639798 Singapore Harbin Inst Technol Sch Comp Sci & Technol Harbin 150001 Peoples R China

Video action recognition aims to classify actions within sequences of video frames, which has important applications in computer vision fields. Existing methods have shown proficiency in well-lit environments but experience a drop in efficiency under low-light conditions. This decline is due to the challenge of extracting relevant information from dark, noisy images. Furthermore, simply introducing enhancement networks as preprocessing will lead to an increase in both parameters and computational burden for the video. To address this dilemma, this paper presents a novel frequency-based method, FRequency-Auxiliary Guided Relational Attention NeTwork (FRAGRANT), designed specifically for low-light action recognition. Its distinctive features can be summarized as: (1) a novel Frequency-Auxiliary Module that focuses on informative object regions, characterizing action and motion while effectively suppressing noise;(2) a sophisticated Relational Attention Module that enhances motion representation by modeling the local s between position neighbors, thereby more efficiently resolving issues, such as fuzzy boundaries. Comprehensive testing demonstrates that FRAGRANT outperforms existing methods, achieving state-of-the-art results on various standard low-light action recognition benchmarks.

关键词： Low-light action recognition Frequency domain analysis Relational attention mechanism Motion representation Action boundary

来源：评论

学校读者我要写书评

暂无评论

Digital Staining with Knowledge Distillation: A Unified Framework for Unpaired and Paired-But-Misaligned Data

引用

IEEE Transactions on Medical Imaging 2025年 PP卷 PP页

作者： Xu, Ziwang Guo, Lanqing Tsutsui, Satoshi Zhang, Shuyan Kot, Alex C. Wen, Bihan Nanyang Technological University Rapid-Rich Object Search Lab China The University of Texas at Austin United States Institute of Materials Research and Engineering China

Staining is essential in cell imaging and medical diagnostics but poses significant challenges, including high cost, time consumption, labor intensity, and irreversible tissue alterations. Recent advances in deep learning have enabled digital staining through supervised model training. However, collecting large-scale, perfectly aligned pairs of stained and unstained images remains difficult. In this work, we propose a novel unsupervised deep learning framework for digital cell staining that reduces the need for extensive paired data using knowledge distillation. We explore two training schemes: (1) unpaired and (2) pairedbut- misaligned settings. For the unpaired case, we introduce a two-stage pipeline, comprising light enhancement followed by colorization, as a teacher model. Subsequently, we obtain a student staining generator through knowledge distillation with hybrid non-reference losses. To leverage the pixel-wise information between adjacent sections, we further extend to the paired-but-misaligned setting, adding the Learning to Align module to utilize pixel-level information. Experiment results on our dataset demonstrate that our proposed unsupervised deep staining method can generate stained images with more accurate positions and shapes of the cell targets in both settings. Compared with competing methods, our method achieves improved results both qualitatively and quantitatively (e.g., NIQE and PSNR). We applied our digital staining method to the White Blood Cell (WBC) dataset, investigating its potential for medical applications. © 1982-2012 IEEE.

关键词： Digital Staining Generative Adversarial Networks Image-to-Image Translation Knowledge Distillation Unsupervised Learning

来源：评论

学校读者我要写书评

暂无评论

Dynamic and static mutual fitting for action recognition

引用

PATTERN RECOGNITION 2025年 157卷

作者： Liu, Wenxuan Jia, Xuemei Zhong, Xian Jiang, Kui Yu, Xiaohan Ye, Mang Wuhan Univ Technol Sch Comp Sci & Artificial Intelligence Hubei Key Lab Transportat Internet Things Wuhan 430070 Peoples R China Wuhan Univ Sch Comp Sci Wuhan 430072 Peoples R China Nanyang Technol Univ Sch Elect & Elect Engn Rapid Rich Object Search Lab Singapore 639798 Singapore Harbin Inst Technol Sch Comp Sci & Technol Harbin 150001 Peoples R China Macquarie Univ Sch Comp Sydney NSW 2109 Australia

Action recognition is intended to classify a video into a certain category by aggregating and summarizing its temporal and spatial information. Existing methods have achieved remarkable performance on the standard datasets. However, how to accurately recognize the action under occluded scenarios remains the major challenge in practical applications and is barely explored. Although previous augmentation methods have made some efforts by introducing additional modalities to deal with occlusion cases, the continuous occlusion related to temporal in the video is rarely focused. To tackle this issue, this paper takes the sample augmentation and feature representation into consideration. Specifically, we design a Dynamic Temporal-aware Erasing (DTE) augmentation strategy to enrich the occlusion samples. The proposed DTE builds upon an explicit analysis conditioned with temporal dimension along with the actors' trajectories, which ensures the capability of carrying the temporal relation to an arbitrary spatial augmentation setting. Specifically, DTE makes the added augmentation samples more temporally consistent with obtaining the motion trajectories of the actors, enhancing the robustness against occlusions. Besides, we revisit the necessity of diverse backgrounds and propose Dynamic and Static Mutual Fitting (DSMF) to optimize the action recognition model. DSMF incorporates background auxiliary mutual fitting actors to distinguish features, which learns a smooth representation with the global temporal consistency. Extensive experiments conducted on standard benchmarks have proved that the proposed DSMF achieves competitive performance over powerful competitors.

关键词： Dynamic temporal-aware erasing Mutual fitting Static augmentation Pseudo-occlusion Action recognition

来源：评论

学校读者我要写书评

暂无评论

Deep Clustering With Variational Autoencoder

引用

IEEE SIGNAL PROCESSING LETTERS 2020年第0期27卷 231-235页

作者： Lim, Kart-Leong Jiang, Xudong Yi, Chenyu Nanyang Technol Univ Sch Elect & Elect Engn Rapid Rich Object Search Lab Singapore 639798 Singapore

An autoencoder that learns a latent space in an unsupervised manner has many applications in signal processing. However, the latent space of an autoencoder does not pursue the same clustering goal as Kmeans or GMM. A recent work proposes to artificially re-align each point in the latent space of an autoencoder to its nearest class neighbors during training (Song et al. 2013). The resulting new latent space is found to be much more suitable for clustering, since clustering information is used. Inspired by previous works (Song et al. 2013), in this letter we propose several extensions to this technique. First, we propose a probabilistic approach to generalize Song's approach, such that Euclidean distance in the latent space is now represented by KL divergence. Second, as a consequence of this generalization we can now use probability distributions as inputs rather than points in the latent space. Third, we propose using Bayesian Gaussian mixture model for clustering in the latent space. We demonstrated our proposed method on digit recognition datasets, MNIST, USPS and SHVN as well as scene datasets, Scene15 and MIT67 with interesting findings.

关键词： Bayes methods Gaussian distribution Optimization Probabilistic logic Probability distribution Random variables Standards

来源：评论

学校读者我要写书评

暂无评论

Image Inpainting Detection via Enriched Attentive Pattern with Near Original Image Augmentation 22

Image Inpainting Detection via Enriched Attentive Pattern wi...

引用

30th ACM International Conference on Multimedia (MM)

作者： Yang, Wenhan Cai, Rizhao Kot, Alex Nanyang Technol Univ Rapid Rich Object Search Lab Singapore Singapore

ISBN: (纸本)9781450392037

As deep learning-based inpainting methods have achieved increasingly better results, its malicious use, e.g. removing objects to report fake news or to provide fake evidence, is becoming threatening. Previous works have provided rich discussions on network architectures, e.g. even performing Neural Architecture search to obtain the optimal model architecture. However, there are rooms in other aspects. In our work, we provide comprehensive efforts from data and feature aspects. From the data aspect, as harder samples in the training data usually lead to stronger detection models, we propose near original image augmentation that pushes the inpainted images closer to the original ones (without distortion and inpainting) as the input images, which is proved to improve the detection accuracy. From the feature aspect, we propose to extract the attentive pattern. With the designed attentive pattern, the knowledge of different inpainting methods can be better exploited during the training phase. Finally, extensive experiments are conducted. In our evaluation, we consider the scenarios where the inpainting masks, which are used to generate the testing set, have a distribution gap from those masks used to produce the training set. Thus, the comparisons are conducted on a newly proposed dataset, where testing masks are inconsistent with the training ones. The experimental results show the superiority of the proposed method and the effectiveness of each component. All our codes and data will be online available.

关键词： Image Inpainting Detection Attentive Pattern Data Augmentation Near Original Image Inpainting Mask

来源：评论

学校读者我要写书评

暂无评论

Diverse Deep Feature Ensemble Learning for Omni-Domain Generalized Person Re-identification 24

Diverse Deep Feature Ensemble Learning for Omni-Domain Gener...

引用

9th International Conference on Multimedia and Image Processing (ICMIP)

作者： Ang, Eugene P. W. Lin, Shan Kot, Alex C. Nanyang Technol Univ Rapid Rich Object Search ROSE Lab Singapore Singapore

ISBN: (纸本)9798400716164

Person Re-identification (Person ReID) has progressed to a level where single-domain supervised Person ReID performance has saturated. However, such methods experience a significant drop in performance when trained and tested across different datasets, motivating the development of domain generalization techniques. However, our research reveals that domain generalization methods significantly underperform single-domain supervised methods on single dataset benchmarks. An ideal Person ReID method should be effective regardless of the number of domains involved, and when test domain data is available for training it should perform as well as state-of-the-art (SOTA) fully supervised methods. This is a paradigm that we call Omni-Domain Generalization Person ReID (ODG-ReID). We propose a way to achieve ODG-ReID by creating deep feature diversity with self-ensembles. Our method, Diverse Deep Feature Ensemble Learning (D2FEL), deploys unique instance normalization patterns that generate multiple diverse views and recombines these views into a compact encoding. To the best of our knowledge, our work is one of few to consider omni-domain generalization in Person ReID, and we advance the study of applying feature ensembles in Person ReID. D2FEL significantly improves and matches the SOTA performance for major domain generalization and single-domain supervised benchmarks.

关键词： person re-identification domain generalization ensemble learning

来源：评论

学校读者我要写书评

暂无评论

Digital Staining with Knowledge Distillation: A Unified Framework for Unpaired and Paired-But-Misaligned Data

arXiv

引用

arXiv 2025年

作者： Xu, Ziwang Guo, Lanqing Tsutsui, Satoshi Zhang, Shuyan Kot, Alex C. Wen, Bihan Rapid-Rich Object Search Lab Nanyang Technological University Singapore The University of Texas Austin United States Institute of Materials Research and Engineering Agency for Science Technology and Research Singapore

Staining is essential in cell imaging and medical diagnostics but poses significant challenges, including high cost, time consumption, labor intensity, and irreversible tissue alterations. Recent advances in deep learning have enabled digital staining through supervised model training. However, collecting large-scale, perfectly aligned pairs of stained and unstained images remains difficult. In this work, we propose a novel unsupervised deep learning framework for digital cell staining that reduces the need for extensive paired data using knowledge distillation. We explore two training schemes: (1) unpaired and (2) paired-but-misaligned settings. For the unpaired case, we introduce a two-stage pipeline, comprising light enhancement followed by colorization, as a teacher model. Subsequently, we obtain a student staining generator through knowledge distillation with hybrid non-reference losses. To leverage the pixel-wise information between adjacent sections, we further extend to the paired-but-misaligned setting, adding the Learning to Align module to utilize pixel-level information. Experiment results on our dataset demonstrate that our proposed unsupervised deep staining method can generate stained images with more accurate positions and shapes of the cell targets in both settings. Compared with competing methods, our method achieves improved results both qualitatively and quantitatively (e.g., NIQE and PSNR). We applied our digital staining method to the White Blood Cell (WBC) dataset, investigating its potential for medical applications. The dataset and code will be released at link. Copyright © 2025, The Authors. All rights reserved.

关键词： Personnel training

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：