检索结果-内蒙古大学图书馆

Proceedings of the 38th IEEE/ACM International Conference on Automated software Engineering

作者： Ze Tang Jidong Ge Shangqing Liu Tingwei Zhu Tongtong Xu Liguo Huang Bin Luo National Key Laboratory for Novel Software Technology Nanjing University China Nanyang Technological University Singapore Huawei Software Engineering Application Technology China Department of Computer Science Southern Methodist University USA

ISBN: (纸本)9798350329964

Large Language Models (LLMs) have demonstrated remarkable performance in code completion. However, due to the lack of domain-specific knowledge, they may not be optimal in completing code that requires intensive domain knowledge for example completing the library names. Although there are several works that have confirmed the effectiveness of fine-tuning techniques to adapt language models for code completion in specific domains. They are limited by the need for constant fine-tuning of the model when the project is in constant *** address this limitation, in this paper, we propose kNM-LM, a retrieval-augmented language model (R-LM), that integrates domain knowledge into language models without fine-tuning. Different from previous techniques, our approach is able to automatically adapt to different language models and domains. Specifically, it utilizes the in-domain code to build the retrieval-based database decoupled from LM, and then combines it with LM through Bayesian inference to complete the code. The extensive experiments on the completion of intra-project and intra-scenario have confirmed that kNM-LM brings about appreciable enhancements when compared to CodeGPT and UnixCoder. A deep analysis of our tool including the responding speed, storage usage, specific type code completion, and API invocation completion has confirmed that kNM-LM provides satisfactory performance, which renders it highly appropriate for domain adaptive code completion. Furthermore, our approach operates without the requirement for direct access to the language model's parameters. As a result, it can seamlessly integrate with black-box code completion models, making it easy to integrate our approach as a plugin to further enhance the performance of these models.

关键词： domain adaptive code completion

来源：评论

学校读者我要写书评

暂无评论

Attack and Improvement of a Hidden Vector Encryption Scheme 12

Attack and Improvement of a Hidden Vector Encryption Scheme

引用

12th International Conference on Electronics, Communications and Networks, CECNet 2022

作者： Wang, Ke Wang, Zhikun Luo, Song Guan, Zhi Key Laboratory of High Confidence Software Technologies Peking University MoE Beijing China Department of Computer Science and Technology Eecs Peking University Beijing China School of Computer Science and Engineering Chongqing University of Technology Chongqing China National Engineering Research Center for Software Engineering Peking University Beijing China

ISBN: (纸本)9781643683683

Hidden Vector Encryption (HVE) is a new kind of attribute-based encryption in which a vector is hidden in the ciphertext or linked with the secret key. In ESORICS 2014, Phuong et al. proposed an HVE scheme with constant-size ciphertext which is constructed in the prime order setting. In this paper, we show that Phuong et al.'s scheme is not vector-hiding due to public parameters in their scheme leak some information about vectors. Furthermore, an improved HVE scheme is proposed in the prime order setting and its security is proven in the security model. Comparison shows our scheme has more efficient in decryption than current other HVE schemes. © 2022 The authors and IOS Press.

关键词： Vectors

来源：评论

学校读者我要写书评

暂无评论

CopGAT: Co-propagation Self-supervised Graph Attention Network

CopGAT: Co-propagation Self-supervised Graph Attention Netwo...

引用

IEEE International Conference on Big Data and Cloud Computing (BdCloud)

作者： Baoming Zhang Ming Xu Mingcai Chen Mingyuan Chen Chongjun Wang Department of Computer Science and Technology State Key Laboratory for Novel Software Technology Nanjing University Nanjing China

Graph Attention Network (GAT) is one of the state-of-the-art architectures for Graph Neural Networks (GNNs). In this paper, we first propose Label Purity to explore the relationship between the graph attention and the node labels. By tracking the label purity of graph attention, we observe that graph attention suppresses message passing between inter-class nodes in the graph with homophily. Upon such basis, we further improve graph attention's capability to capture label information by designing a self-supervised graph attention loss. We propose a novel approach based on self-supervised graph attention loss, Co-propagation Self-supervised Graph Attention Network (CopGAT), which explicitly restrains message passing between inter-class nodes. The Co-propagation means propagating node representations in Euclidean space and pseudo labels in probability simplex via the shared attention matrix, combining Message Passing Neural Networks and Label Propagation Algorithm under the graph attention framework. Constrained with the self-supervised graph attention loss generated by co-propagation, our CopGAT enhances the ability of GAT to attenuate the influence of structural noise. Furthermore, our self-supervised graph attention loss can be extended to other graph attention models in a plug-and-play manner. Experimental results demonstrate that our Cop GAT outperforms previous state-of-the-art methods on five benchmark datasets for the semi-supervised node classification task.

关键词： Message passing Benchmark testing Graph neural networks Task analysis

来源：评论

学校读者我要写书评

暂无评论

Supervised Contrastive Learning with Prototype Distillation for Data Incremental Learning

SSRN

引用

SSRN 2024年

作者： Yang, Suorong Zhang, Tianyue Xu, Baile Shen, Furao Zhao, Jian State Key Laboratory for Novel Software Technology Nanjing University China Department of Computer Science and Technology Nanjing University China School of Artificial Intelligence Nanjing University China School of Electronic Science and Engineering Nanjing University China

The goal of Data Incremental Learning (DIL) is to enable learning from small-scale data batches from non-stationary data streams without clear task divisions. This approach often encounters the issue of catastrophic forgetting in deep neural networks. Meanwhile, practical applications may involve unbalanced data distributions across classes. To address the challenges of DIL, using prototypes as classifiers offers a balance between stability and flexibility. This facilitates the preservation of information from previous classes while evolving to accommodate new ones. Prototypes are particularly effective for classifying separable embeddings in the feature space, as they draw embeddings from the same class closer together while pushing those from different classes further apart. This aligns with the objective of contrastive learning. In this paper, we propose Supervised Contrastive learning with the Prototype Distillation (SCPD) method for the DIL problem. First, we employ supervised contrastive loss (SCL) for model training. To further mitigate the forgetting problem, we introduce the prototype distillation loss, which ensures that features remain close to their corresponding prototypes. Exemplar memory is utilized to store representative samples, enabling the model to maintain performance on previously learned tasks. Moreover, data augmentation techniques are applied to improve the models' robustness and continual learning capability. Experimental results demonstrate that the SCPD model outperforms prior state-of-the-art approaches across several benchmarks, including those with various imbalanced setups. Code will be made publicly available soon. © 2024, The Authors. All rights reserved.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Vulnerable Region-Aware Greybox Fuzzing

引用

Journal of computer science & technology 2021年第5期36卷 1212-1228页

作者： Ling-Yun Situ Zhi-Qiang Zuo Le Guan Lin-Zhang Wang Xuan-Dong Li Jin Shi Peng Liu State Key Laboratory for Novel Software Technology Nanjing UniversityNanjing 210023China School of Information Management Nanjing UniversityNanjing 210023China Department of Computer Science University of GeorgiaAthensGA 30602U.S.A. College of Information Sciences and Technology Pennsylvania State UniversityState CollegePA 16802U.S.A

Fuzzing is known to be one of the most effective techniques to uncover security vulnerabilities of large-scale software *** fuzzing,it is crucial to distribute the fuzzing resource appropriately so as to achieve the best fuzzing performance under a limited *** distribution strategies of American Fuzzy Lop(AFL)based greybox fuzzing focus on increasing coverage blindly without considering the metrics of code regions,thus lacking the insight regarding which region is more likely to be vulnerable and deserves more fuzzing *** tackle the above drawback by proposing a vulnerable region-aware greybox fuzzing ***,we distribute more fuzzing resources towards regions that are more likely to be vulnerable based on four kinds of code *** implemented the approach as an extension to AFL named ***-scale experimental evaluations validate the effectiveness and efficiency of RegionFuzz-11 new bugs including three new CVEs are successfully uncovered by RegionFuzz.

关键词： vulnerability detection greybox fuzzing code metrics resource distribution

来源：评论

学校读者我要写书评

暂无评论

AdaAugment: A Tuning-Free and Adaptive Approach to Enhance Data Augmentation

arXiv

引用

arXiv 2024年

作者： Yang, Suorong Li, Peijia Xiong, Xi Shen, Furao Zhao, Jian State Key Laboratory for Novel Software Technology Nanjing University China Department of Computer Science and Technology Nanjing University China School of Artificial Intelligence Nanjing University China School of Electronic Science and Engineering Nanjing University China

Data augmentation (DA) is widely employed to improve the generalization performance of deep models. However, most existing DA methods use augmentation operations with random magnitudes throughout training. While this fosters diversity, it can also inevitably introduce uncontrolled variability in augmented data, which may cause misalignment with the evolving training status of the target models. Both theoretical and empirical findings suggest that this misalignment increases the risks of underfitting and overfitting. To address these limitations, we propose AdaAugment, an innovative and tuning-free Adaptive Augmentation method that utilizes reinforcement learning to dynamically adjust augmentation magnitudes for individual training samples based on real-time feedback from the target network. Specifically, AdaAugment features a dual-model architecture consisting of a policy network and a target network, which are jointly optimized to effectively adapt augmentation magnitudes. The policy network optimizes the variability within the augmented data, while the target network utilizes the adaptively augmented samples for training. Extensive experiments across benchmark datasets and deep architectures demonstrate that AdaAugment consistently outperforms other state-of-the-art DA methods in effectiveness while maintaining remarkable efficiency. Copyright © 2024, The Authors. All rights reserved.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Nerftap: Enhancing Transferability of Adversarial Patches on Face Recognition Using Neural Radiance Fields

SSRN

引用

SSRN 2024年

作者： Liu, Xiaoliang Shen, Fu-Rao Han, Feng Zhao, Jian Nie, Changhai State Key Laboratory for Novel Software Technology Nanjing University China Department of Computer Science and Technology Nanjing University China School of Artificial Intelligence Nanjing University China School of Electronic Science and Engineering Nanjing University China

Face recognition (FR) technology plays a crucial role in various applications, but its vulnerability to adversarial attacks poses significant security concerns. Existing research primarily focuses on transferability to different FR models, overlooking the direct transferability to victim's face images, which is a practical threat in real-world scenarios. In this study, we propose a novel adversarial attack method that considers both the transferability to the FR model and the victim's face image, called NerfTAP. Leveraging Neural Radiance Fields (NeRF)-based 3D Generative Adversarial Network (GAN), we generate new view face images for the source and target subjects to enhance transferability of adversarial patches. We introduce a style consistency loss to ensure the visual similarity between the adversarial UV map and the target UV map under a 0-1 mask, enhancing the effectiveness and naturalness of the generated adversarial face images. Extensive experiments and evaluations on various FR models demonstrate the superiority of our approach over existing attack techniques. Our work provides valuable insights for enhancing the robustness of FR systems in practical adversarial settings. © 2024, The Authors. All rights reserved.

关键词： Face recognition

来源：评论

学校读者我要写书评

暂无评论

EntAugment: Entropy-Driven Adaptive Data Augmentation Framework for Image Classification

arXiv

引用

arXiv 2024年

作者： Yang, Suorong Shen, Furao Zhao, Jian State Key Laboratory for Novel Software Technology Nanjing University China Department of Computer Science and Technology Nanjing University China School of Artificial Intelligence Nanjing University China School of Electronic Science and Engineering Nanjing University China

Data augmentation (DA) has been widely used to improve the generalization of deep neural networks. While existing DA methods have proven effective, they often rely on augmentation operations with random magnitudes to each sample. However, this approach can inadvertently introduce noise, induce distribution shifts, and increase the risk of overfitting. In this paper, we propose EntAugment, a tuning-free and adaptive DA framework. Unlike previous work, EntAugment dynamically assesses and adjusts the augmentation magnitudes for each sample during training, leveraging insights into both the inherent complexities of training samples and the evolving status of deep models. Specifically, in EntAugment, the magnitudes are determined by the information entropy derived from the probability distribution obtained by applying the softmax function to the model’s output. In addition, to further enhance the efficacy of EntAugment, we introduce a novel entropy regularization term, EntLoss, which complements the EntAugment approach. Theoretical analysis further demonstrates that EntLoss, compared to traditional cross-entropy loss, achieves closer alignment between the model distributions and underlying dataset distributions. Moreover, EntAugment and EntLoss can be utilized separately or jointly. We conduct extensive experiments across multiple image classification tasks and network architectures with thorough comparisons of existing DA methods. Importantly, the proposed methods outperform others without introducing any auxiliary models or noticeable extra computational costs, highlighting both effectiveness and efficiency. Code is available at https://***/Jackbrocp/EntAugment. Copyright © 2024, The Authors. All rights reserved.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

CADEC: A Combinatorial Auction for Dynamic Distributed DNN Inference Scheduling in Edge-Cloud Networks

引用

IEEE Transactions on Mobile Computing 2025年

作者： Xu, Xiaolong Hu, Yuhao Cui, Guangming Qi, Lianyong Dou, Wanchun Nanjing University of Information Science and Technology School of Software Nanjing210044 China Nanjing University of Information Science and Technology Jiangsu Province Engineering Research Center of Advanced Computing and Intelligent Services School of Software China Nanjing University State Key Laboratory for Novel Software Technology China College of Computer science and technology Qingdao China Georgia State University Department of Computer Science United States

Deep Neural Network (DNN) Inference, as a key enabler of intelligent applications, is often computation-intensive and latency-sensitive. Combining the advantages of cloud computing (abundant computing resources) and edge computing (fast transmission), edge-cloud collaborative DNN inference is a powerful solution to these problems. However, in edge-cloud networks with heterogeneous resources, how to obtain reasonable decisions on server selection, model partition and resource allocation for efficient distributed DNN inference is a hard challenge. Furthermore, it is non-trivial to design suitable resource prices to maximize the social welfare. These challenges even escalate in dynamic edge-cloud networks where decisions should be generated as soon as each user arrives without future information. Therefore, we design a combinatorial auction for dynamic distributed DNN inference scheduling, named CADEC. CADEC first constructs a bid set for each user based on convex optimization theory for optimal solution searching. Next, prices of resources in the edge-cloud network are adjusted according to changes in supply-demand relationship, and whether to admit the request of each user is decided. Finally, the dynamic distributed inference scheduling decisions are generated through the primal-dual algorithm to maximize the social welfare. Theoretical analysis shows the good competitive ratio and polynomial time complexity of CADEC. Results of simulation experiments present that CADEC improves social welfare by up to 46% compared with state-of-the-art distributed DNN inference schemes. © 2002-2012 IEEE.

关键词： Convex optimization

来源：评论

学校读者我要写书评

暂无评论

AoI-aware Incentive Mechanism for Mobile Crowdsensing using Stackelberg Game

AoI-aware Incentive Mechanism for Mobile Crowdsensing using ...

引用

IEEE Annual Joint Conference: INFOCOM, IEEE computer and Communications Societies

作者： Mingjun Xiao Yin Xu Jinrui Zhou Jie Wu Sheng Zhang Jun Zheng School of Computer Science and Technology University of Science and Technology of China China Suzhou Institute for Advanced Research University of Science and Technology of China China Department of Computer and Information Sciences Temple Universityy State Key Laboratory for Novel Software Technology Nanjing University China

Mobile CrowdSensing (MCS) is a mobile computing paradigm, through which a platform can coordinate a crowd of workers to accomplish large-scale data collection tasks using their mobile devices. Information freshness has attracted much focus on MCS research worldwide. In this paper, we investigate the incentive mechanism design in MCS systems that take the freshness of collected data and social benefits into concerns. First, we introduce the Age of Information (AoI) metric to measure the freshness of data. Then, we model the incentive mechanism design with AoI guarantees as a novel incomplete information two-stage Stackelberg game with multiple constraints. Next, we derive the optimal strategies of this game so as to determine the optimal reward paid by the platform and the optimal data update frequency for each worker. Moreover, we prove that these optimal strategies form a unique Stackelberg equilibrium. Based on the optimal strategies, we propose an AoI-Aware Incentive (AIAI) mechanism for the MCS system, whereby the platform and all workers can maximize their utilities simultaneously. Meanwhile, the system can ensure that the AoI values of all data uploaded to the platform are not larger than a given threshold to achieve high data freshness. Extensive simulations on real-world traces are conducted to demonstrate the significant performance of AIAI.

关键词：

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：