检索结果-内蒙古大学图书馆

DBSSL: A Scheme to Detect Backdoor Attacks in Self-Supervised Learning Models

IEEE Transactions on Dependable and Secure Computing 2025年

作者： Huang, Yuxian Yang, Geng Yuan, Dong Yu, Shui Nanjing University of Posts and Telecommunication College of Computer Science and Software China Jiangsu Key Laboratory of Big Data Security and Intelligent Processing China University of Sydney School of Electrical and Information Engineering Australia University of Technology Sydney School of Computer Science Australia

Recently, self-supervised learning has garnered significant attention for its ability to extract high-quality features from unlabeled data. However, existing research indicates that backdoor attacks can pose significant threats to self-supervised learning. Additionally, due to the substantial training overhead, traditional backdoor detection methods in supervised learning, such as meta-learning, are not well-suited for self-supervised learning. Current backdoor detection methods for self-supervised learning can only defend against backdoor attacks triggered by patch. To address this challenge, we proposes DBSSL, a scheme that can efficiently detect backdoor attacks in self-supervised learning models. In DBSSL, we connect backdoor attacks with adversarial attacks and use adversarial perturbations for detection. Specifically, we first conduct targeted adversarial attacks on the samples, classifying them into different classes. Then, we employ the Z-Score method to analyze these adversarial perturbations. If outliers are present, we can infer that the self-supervised learning model has been backdoored. Unlike previous works, our scheme is better suited for downstream users with constrained computation resources and exhibits excellent detection capabilities for backdoor attacks caused by patch -type or global triggers. Abundant theoretical analysis has demonstrated the feasibility of our scheme. Moreover, extensive experiments show that our method performs well in detecting prevalent backdoor attacks in self-supervised learning, achieving detection accuracy exceeding 95%. © 2004-2012 IEEE.

关键词： Self-supervised learning

来源：评论

学校读者我要写书评

暂无评论

Galliot: Path Merging Based Betweenness Centrality Algorithm on GPU 42

Galliot: Path Merging Based Betweenness Centrality Algorithm...

引用

42nd IEEE International Conference on computer Communications, INFOCOM 2023

作者： Zheng, Zhigao Zhao, Chen Xie, Peichen Dum, Bo School of Computer Science Wuhan University Wuhan 430072 China National Engineering Research Center for Multimedia Software Wuhan University Wuhan 430072 Chin. Institute of Artificial Intelligence Wuhan University Wuhan 430072 Chin. Hubei Key Laboratory of Multimedia and Network Communication Engineering Wuhan University Wuhan 430072 China Hubei Luojia Laboratory Wuhan 430072 China

ISBN: (纸本)9798350334142

Betweenness centrality (BC) is widely used to measure a vertex's significance by using the frequency of a vertex appearing in the shortest path between other vertices. However, most recent algorithms in BC computation suffer from the problem of high auxiliary memory consumption. To reduce BC computing's memory consumption, we propose a path-mergingbased algorithm called Galliot to calculate the BC values on GPU, which aims to minimize the on-board memory consumption and enable the BC computation of large-scale graphs. The proposed algorithm requires O(n) space and runs in O(mn) time on unweighted graphs. We present the theoretical principle for the proposed path merging method. Moreover, we propose a locality-oriented policy to maintain and update the worklist to improve GPU data locality. In addition, we conducted extensive experiments on NVIDIA GPUs to show the performance of Galliot. The results show that Galliot can process the larger graphs, which have 11.32× more vertices and 5.67× more edges than the graphs that recent works. Moreover, Galliot can achieve up to 38.77× speedup over the existing methods. © 2023 IEEE.

关键词： Graphics processing unit

来源：评论

学校读者我要写书评

暂无评论

Thinking in Granularity: Dynamic Quantization for Image Super-Resolution by Intriguing Multi-Granularity Clues

arXiv

引用

arXiv 2024年

作者： Wang, Mingshen Zhang, Zhao Li, Feng Xu, Ke Miao, Kang Wang, Meng School of Computer Science and Information Engineering Hefei University of Technology Hefei China Yunnan Key Laboratory of Software Engineering Yunan China School of Artificial Intelligence Anhui University Hefei China

Dynamic quantization has attracted rising attention in image super-resolution (SR) as it expands the potential of heavy SR models onto mobile devices while preserving competitive performance. Existing methods explore layer-to-bit configuration upon varying local regions, adaptively allocating the bit to each layer and patch. Despite the benefits, they still fall short in the trade-off of SR accuracy and quantization efficiency. Apart from this, adapting the quantization level for each layer individually can disturb the original inter-layer relationships, thus diminishing the representation capability of quantized models. In this work, we propose Granular-DQ, which capitalizes on the intrinsic characteristics of images while dispensing with the previous consideration for layer sensitivity in quantization. Granular-DQ conducts a multi-granularity analysis of local patches with further exploration of their information densities, achieving a distinctive patch-wise and layer-invariant dynamic quantization paradigm. Specifically, Granular-DQ initiates by developing a granularity-bit controller (GBC) to apprehend the coarse-to-fine granular representations of different patches, matching their proportional contribution to the entire image to determine the proper bit-width allocation. On this premise, we investigate the relation between bit-width and information density, devising an entropy-to-bit (E2B) mechanism that enables further fine-grained dynamic bit adaption of high-bit patches. Extensive experiments validate the superiority and generalization ability of Granular-DQ over recent state-of-the-art methods on various SR models. Code and supplementary statement can be found at https://***/MmmingS/***. Copyright © 2024, The Authors. All rights reserved.

关键词： Granular computing

来源：评论

学校读者我要写书评

暂无评论

When Neural Code Completion Models Size up the Situation: Attaining Cheaper and Faster Completion through Dynamic Model Inference

When Neural Code Completion Models Size up the Situation: At...

引用

International Conference on software engineering (ICSE)

作者： Zhensu Sun Xiaoning Du Fu Song Shangwen Wang Li Li Beihang University Beijing China Monash University Melbourne Victoria Australia State Key Laboratory of Computer Science Institute of Software Chinese Academy of Sciences Beijing China National University of Defense Technology Changsha China Beijing Yunnan Key Laboratory of Software Engineering Beihang University China

ISBN: (数字)9798400702174

ISBN: (纸本)9798350382143

Leveraging recent advancements in large language models, modern neural code completion models have demonstrated the capability to generate highly accurate code suggestions. However, their massive size poses challenges in terms of computational costs and environmental impact, hindering their widespread adoption in practical scenarios. Dynamic inference emerges as a promising solution, as it allocates minimal computation during inference while maintaining the model's performance. In this research, we explore dynamic inference within the context of code completion. Initially, we conducted an empirical investigation on GPT-2, focusing on the inference capabilities of intermediate layers for code completion. We found that 54.4% of tokens can be accurately generated using just the first layer, signifying significant computational savings potential. Moreover, despite using all layers, the model still fails to predict 14.5% of tokens correctly, and the subsequent completions continued from them are rarely considered helpful, with only a 4.2% Acceptance Rate. These findings motivate our exploration of dynamic inference in code completion and inspire us to enhance it with a decision-making mechanism that stops the generation of incorrect code. We thus propose a novel dynamic inference method specifically tailored for code completion models. This method aims not only to produce correct predictions with largely reduced computation but also to prevent incorrect predictions proactively. Our extensive evaluation shows that it can averagely skip 1.7 layers out of 16 layers in the models, leading to an 11.2% speedup with only a marginal 1.1 % reduction in ROUGE- L.

关键词： Codes Accuracy Computational modeling Decision making Focusing Predictive models Dynamic scheduling

来源：评论

学校读者我要写书评

暂无评论

Vision-Language Models are Strong Noisy Label Detectors

arXiv

引用

arXiv 2024年

作者： Wei, Tong Li, Hao-Tian Li, Chun-Shu Shi, Jiang-Xin Li, Yu-Feng Zhang, Min-Ling School of Computer Science and Engineering Southeast University Nanjing China Key Laboratory of Computer Network and Information Integration Southeast University Ministry of Education China National Key Laboratory for Novel Software Technology Nanjing University China School of Artificial Intelligence Nanjing University China

Recent research on fine-tuning vision-language models has demonstrated impressive performance in various downstream tasks. However, the challenge of obtaining accurately labeled data in real-world applications poses a significant obstacle during the fine-tuning process. To address this challenge, this paper presents a Denoising Fine-Tuning framework, called DEFT, for adapting vision-language models. DEFT utilizes the robust alignment of textual and visual features pre-trained on millions of auxiliary image-text pairs to sieve out noisy labels. The proposed framework establishes a noisy label detector by learning positive and negative textual prompts for each class. The positive prompt seeks to reveal distinctive features of the class, while the negative prompt serves as a learnable threshold for separating clean and noisy samples. We employ parameter-efficient fine-tuning for the adaptation of a pre-trained visual encoder to promote its alignment with the learned textual prompts. As a general framework, DEFT can seamlessly fine-tune many pre-trained models to downstream tasks by utilizing carefully selected clean samples. Experimental results on seven synthetic and real-world noisy datasets validate the effectiveness of DEFT in both noisy label detection and image classification tasks. Our source code is available at https://***/HotanLee/DeFT. © 2024, CC BY.

关键词： Visual languages

来源：评论

学校读者我要写书评

暂无评论

Watch Your Head: Assembling Projection Heads to Save the Reliability of Federated Models

arXiv

引用

arXiv 2024年

作者： Chen, Jinqian Zhu, Jihua Zheng, Qinghai Li, Zhongyu Tian, Zhiqiang School of Software Engineering Xi’an Jiaotong University China College of Computer and Data Science Fuzhou University China Shaanxi Joint Key Laboratory for Artificial Intelligence China

Federated learning encounters substantial challenges with heterogeneous data, leading to performance degradation and convergence issues. While considerable progress has been achieved in mitigating such an impact, the reliability aspect of federated models has been largely disregarded. In this study, we conduct extensive experiments to investigate the reliability of both generic and personalized federated models. Our exploration uncovers a significant finding: federated models exhibit unreliability when faced with heterogeneous data, demonstrating poor calibration on in-distribution test data and low uncertainty levels on out-of-distribution data. This unreliability is primarily attributed to the presence of biased projection heads, which introduce miscalibration into the federated models. Inspired by this observation, we propose the"Assembled Projection Heads" (APH) method for enhancing the reliability of federated models. By treating the existing projection head parameters as priors, APH randomly samples multiple initialized parameters of projection heads from the prior and further performs targeted fine-tuning on locally available data under varying learning rates. Such a head ensemble introduces parameter diversity into the deterministic model, eliminating the bias and producing reliable predictions via head averaging. We evaluate the effectiveness of the proposed APH method across three prominent federated benchmarks. Experimental results validate the efficacy of APH in model calibration and uncertainty estimation. Notably, APH can be seamlessly integrated into various federated approaches but only requires less than 30% additional computation cost for 100× inferences within large models. Copyright © 2024, The Authors. All rights reserved.

关键词： Reliability

来源：评论

学校读者我要写书评

暂无评论

Anomaly Detection for Multivariate Time Series Based on Contrastive Learning and Autoformer

Anomaly Detection for Multivariate Time Series Based on Cont...

引用

International Conference on computer Supported Cooperative Work in Design

作者： Xuwen Shang Jue Zhang Xingguo Jiang Hong Luo Beijing Key Laboratory of Intelligent Telecommunication Software and Multimedia Beijing University of Posts and Telecommunications Beijing China Department of Computer Science and Engineering Beihang University Beijing China

ISBN: (数字)9798350349184

ISBN: (纸本)9798350349191

Nowadays, due to the increasingly complex and huge data flow, anomaly detection has become an important technology to ensure the high security and reliability of the systems. It helps identify unusual observations in the data, thereby protecting safety and reducing risk. However, in the field of multivariate temporal anomaly detection, there are still some difficulties in spatio-temporal correlation mining of variables, generalization ability in dealing with complex temporal sequences, and feature learning under different granularity. Therefore, we propose an unsupervised anomaly detection method for multivariate time series based on contrastive learning and Autoformer autoencoder. In order to fully explore the spatio-temporal correlation of time series data, this paper uses Graph Attention Network and Auto-Correlation to jointly learn spatio-temporal correlation, and obtain positive and negative pairs through time-domain and frequency-domain data augmentation. Finally, the Autoformer autoencoder is used for anomaly detection in a predictive manner. In the loss function section, a joint loss function is formed by introducing contrastive learning losses at the time step and time window granularity, combined with prediction loss, to improve the learning effect. This paper investigates the performance of the proposed method through experiments on the SMAP and MSL public datasets, and achieves better results than baseline methods.

关键词： Correlation Computational modeling Time series analysis Contrastive learning computer architecture Stability analysis Safety

来源：评论

学校读者我要写书评

暂无评论

PIP: Perturbation-based Iterative Pruning for Large Language Models

arXiv

引用

arXiv 2025年

作者： Cao, Yi Xu, Wei-Jie Shen, Yucheng Shi, Weijie Chan, Chi-Min Xu, Jiajie School of Computer Science and Technology Soochow University China State Key Laboratory for Novel Software Technology Nanjing University China Department of Computer Science and Engineering The Hong Kong University of Science and Technology Hong Kong

The rapid increase in the parameter counts of Large Language Models (LLMs), reaching billions or even trillions, presents significant challenges for their practical deployment, particularly in resource-constrained environments. To ease this issue, we propose PIP (Perturbation-based Iterative Pruning), a novel double-view structured pruning method to optimize LLMs, which combines information from two different views: the unperturbed view and the perturbed view. With the calculation of gradient differences, PIP iteratively prunes those that struggle to distinguish between these two views. Our experiments show that PIP reduces the parameter count by approximately 20% while retaining over 85% of the original model’s accuracy across varied benchmarks. In some cases, the performance of the pruned model is within 5% of the unpruned version, demonstrating PIP’s ability to preserve key aspects of model effectiveness. Moreover, PIP consistently outperforms existing state-of-the-art (SOTA) structured pruning methods, establishing it as a leading technique for optimizing LLMs in environments with constrained resources. Our code is available at: https://***/caoyiiiiii/PIP. Copyright © 2025, The Authors. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Interpretability Research of Variational Autoencoder Generation Process Based on Feature Disentanglement

Interpretability Research of Variational Autoencoder Generat...

引用

Artificial Intelligence, Networking and Information Technology (AINIT), International Seminar on

作者： Sicheng Xi Yanhui Peng Chongqing Key Laboratory of Computational Intelligence College of Computer Science and Technology Chongqing University of Posts and Telecommunications Chongqing China Chongqing Key Laboratory of Computational Intelligence School of Software Engineering Chongqing University of Posts and Telecommunications Chongqing China

ISBN: (数字)9798350385557

ISBN: (纸本)9798350385564

Variational Autoencoder (VAE), as one of the main generative models, has a powerful representation learning capability. However, the hidden space representation learned by VAE is a high-dimensional and complex vector space, which makes it difficult to explain how the model gradually learns and composes the final generated results on different semantic features. To address this problem, firstly, this paper increases the degree of decoupling between different semantic features by increasing the independence between the hidden variables of the modal hermitian space, and explains the learning process of the model on different hermitian spaces by visualization based on the feature decoupling model. In addition, this paper also proposes a hidden variable contribution index to measure the influence of different dimensional hidden variables on the generation results, so as to explain the learning process of the model.

关键词： Training Seminars Representation learning Visualization Correlation Semantics Vectors

来源：评论

学校读者我要写书评

暂无评论

Contextual Insight: Detecting Abnormal Device Behaviors in IoT Systems

Contextual Insight: Detecting Abnormal Device Behaviors in I...

引用

International Conference on computer Supported Cooperative Work in Design

作者： Jian Ge Jianwu Rui Hengtai Ma Bin Li Yeping He National Engineering Research Center of Fundamental Software Institute of Software Chinese Academy of Sciences Beijing China University of Chinese Academy of Sciences Beijing China State Key Laboratory of Computer Science Institute of Software Chinese Academy of Sciences Beijing China

ISBN: (数字)9798350349184

ISBN: (纸本)9798350349191

The proliferation of collaborative technologies and the Internet of Things (IoT) has led to a greater diversity of devices. However, it also increased their vulnerability to potential attacks. The inherent heterogeneity and complexity of IoT devices make it difficult to accurately detect abnormal behavior, leaving them open to manipulation by attackers. To address this issue, we propose Talos, an innovative solution to detect abnormal device behavior in IoT systems. Talos employs advanced deep neural networks to analyze and comprehend historical event sequences, enabling it to accurately model the normal behaviors of IoT devices and detect any anomalies through reconstruction error. We evaluate our prototype on a real-world testbed and test it for three different anomaly cases. The experimental results demonstrate that Talos achieved high precision and recall, achieving an average F1-Score of 99.73% while raising only a few false alarms.

关键词： Analytical models Federated learning Prototypes Collaboration Cyberspace Artificial neural networks Complexity theory

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：