检索结果-内蒙古大学图书馆

VulnSlicer: A Java Vulnerability Analysis Tool Based on Vulnerability Slice 10

VulnSlicer: A Java Vulnerability Analysis Tool Based on Vuln...

10th IEEE International Conference on Dependability in Sensor, Cloud and Big Data Systems and Applications, DependSys 2024

作者： Liao, Chaohao Zhang, Shihe Liu, Renhua University of Electronic Science and Technology of China School of Computer Science and Engineering Chengdu China

ISBN: (纸本)9798331540500

Due to its exceptional performance and memory security, Java is extensively employed in enterprise applications, thereby rendering any vulnerability in Java applications capable of causing significant repercussions. In recent years, due to the frequent occurrence of network attacks, new attack methods continue to emerge while software systems become increasingly intricate and unstable under the weight of historical burdens. Consequently, static analysis techniques are imperative for identifying security risks in code during the coding phase. The prevailing approach in static analysis involves transforming code into a standardized intermediate representation and conducting vulnerability detection, without code execution. However, this methodology often yields detection results accompanied by a considerable false positive rate. We propose VulnSlicer, a tool based on graph similarity matching and vulnerability subgraph slice technology. This tool not only enables efficient vulnerability retrieval but also effectively reduces the false positive rate by leveraging the slice mechanism to save historical vulnerabilities. Additionally, it facilitates rapid matching of accumulated vulnerabilities in the database and further mitigates false positives through the introduction of a keyframe mechanism. In the experiment, we conducted a comparative analysis of several existing static detection tools to assess their effectiveness in detecting vulnerabilities within the same codebase. Experiments show that VulnSlicer has decent accuracy and lower false positive rate. © 2024 IEEE.

关键词： Java programming language

来源：评论

学校读者我要写书评

暂无评论

MFN: A Brain-Like Auditory Model Based on Spiking Neural Networks 21

MFN: A Brain-Like Auditory Model Based on Spiking Neural Net...

引用

21st International computer Conference on Wavelet Active Media technology and Information Processing, ICCWAMTIP 2024

作者： Hongyu, Qing Guangyuan, Gao Yang, Zhang Xuntong, Lin School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu China

ISBN: (纸本)9798331519254

In the human sensing system, information is transmitted, received, and processed in the form of action voltages or electrical pulses. The birth of Spiking Neural Networks (SNNs) has brought neural networks closer to the human brain. However, there is still a lack of a brain like model that can effectively simulate the biological auditory pathway in terms of its operational mechanism. This paper proposes a brain like auditory model that is more in line with biological speech processing mechanisms. Specifically, in the data input stage, mask encoding was introduced to simulate the masking effect in acoustics. In the feature extraction stage, a multi-scale feedback neural network was designed based on the top-down feedback mechanism in the biological auditory system. Meanwhile, the experimental results on the speech recognition task in the Google Speech Command dataset demonstrate that our auditory model has superior performance. © 2024 IEEE.

关键词： Encoding (symbols)

来源：评论

学校读者我要写书评

暂无评论

Optimizing Spiking Neural Networks with Transformer-Based Spatial Attention 4

Optimizing Spiking Neural Networks with Transformer-Based Sp...

引用

4th International Symposium on computer technology and Information science, ISCTIS 2024

作者： Zheng, Peixiang Shen, Zhao University of Electronic Science and Technology of China School of Computer Science and Engineering Chengdu China

ISBN: (纸本)9798350354560

This study introduces SpikeNetTransformer, a novel approach to developing Spiking Neural Networks (SNNs) by integrating the Transformer architecture. The proposed model optimizes the interplay between self-attention mechanism and SNN features to enhance performance across various datasets. A key contribution of this research is the introduction of the Temporal Spike-Scaled Attention (TSSA) module, which is an optimization of the existing Sparse Spatial Attention (SSA) structure. The TSSA module efficiently utilizes spike-shaped Queries, Keys, and Values to effectively blend temporal features. Moreover, a scaling factor is incorporated into the TSSA module to deal with vanishing and exploding gradients during training. This modification not only stabilizes the learning process, but also accelerates both the training and inference phases while reducing the memory consumption by 20 percent. SpikeNetTransformer demonstrates superior performance compared to existing SNN models, achieving competitive Top-1 accuracy rates on static datasets such as CIFAR-10 and CIFAR-100. On CIFAR-10, our model achieves a Top-1 accuracy of 93.34%, outperforming most state-of-the-art SNN models. Similarly, on CIFAR-100, our model achieves Top-1 accuracy of 76.22%, outperforming many existing SNN architectures. This study highlights the potential of refined SNNs in combination with Transformer techniques for efficient and accurate neural computation. © 2024 IEEE.

关键词： Spatio-temporal data

来源：评论

学校读者我要写书评

暂无评论

TGFFD: A Two-Stream Graph Neural Network for Financial Fraud Detection Based on Graph Convolution and Wavelet Analysis 21

TGFFD: A Two-Stream Graph Neural Network for Financial Fraud...

引用

21st International computer Conference on Wavelet Active Media technology and Information Processing, ICCWAMTIP 2024

作者： Zhen, Gong Jianpin, Li School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu China

ISBN: (纸本)9798331519254

Financial fraud has caused lots of losses for both consumers and banks. One of the modern methods to detect financial fraud is graph neural network, which checks relationships in financial transactions, but graph convolution-based methods cannot capture the global information of the graph well, and graph Fourier transform-based methods cannot capture the local relationships of the graph well. Therefore, we propose a two-stream graph neural network (TGFFD) based on attentional graph convolution and wavelet analysis for financial fraud detection. Specifically, we use the history of user behavior as input data to construct a graph, which consists of users (nodes) and relationships between them (edges). Then we use a two-stream network to extract features from the graph structure, where one branch uses a similarity attention-based neighborhood sampling, which can extract the local feature of the graph and identify the camouflage behaviors, and the other stream uses a wavelet-analysis-based graph spectral transform to extract the global features of the graph. Finally, fashioning the features from both streams to get the final prediction of the model. We conducted some experiments on two public datasets of financial fraud detection. The results show that our proposed model outperforms other baselines on the financial fraud detection task. © 2024 IEEE.

关键词： Graph neural networks

来源：评论

学校读者我要写书评

暂无评论

Transaction transmission model for blockchain channels based on non-cooperative games

引用

science China(Information sciences) 2023年第1期66卷 97-114页

作者： Peiyun ZHANG Chenxi LI Mengchu ZHOU Wenjun HUANG Abdullah ABUSORRAH Omaimah O.BAMASAG Engineering Research Center of Digital Forensics of Ministry of Education School of Computer and SoftwareNanjing University of Information Science and Technology School of Computer and Information Anhui Normal University Department of Electrical and Computer Engineering New Jersey Institute of Technology Center of Research Excellence in Renewable Energy and Power Systems Department of Electrical and Computer Engineering Faculty of Engineering and K.A.CARE Energy Research and Innovation CenterKing Abdulaziz University Department of Computer Science Faculty of Computing and Information Technology Center of Excellence in Smart Environment Research King Abdulaziz University

In blockchain networks, transactions can be transmitted through channels. The existing transmission methods depend on their routing information. If a node randomly chooses a channel to transmit a transaction, the transmission may be aborted due to insufficient funds(also called balance) or a low transmission rate. To increase the success rate and reduce transmission delay across all transactions, this work proposes a transaction transmission model for blockchain channels based on non-cooperative game *** balance, channel states, and transmission probability are fully considered. This work then presents an optimized channel transaction transmission algorithm. First, channel balances are analyzed and suitable channels are selected if their balance is sufficient. Second, a Nash equilibrium point is found by using an iterative sub-gradient method and its related channels are then used to transmit transactions. The proposed method is compared with two state-of-the-art approaches: Silent Whispers and Speedy Murmurs. Experimental results show that the proposed method improves transmission success rate, reduces transmission delay,and effectively decreases transmission overhead in comparison with its two competitive peers.

关键词： Blockchain channel transaction transmission channel balance channel state channel transmission probability non-cooperative game

来源：评论

学校读者我要写书评

暂无评论

A Feature Adversarial Domain Adaptation Method in Single-Stage Object Detection Models 21

A Feature Adversarial Domain Adaptation Method in Single-Sta...

引用

21st International computer Conference on Wavelet Active Media technology and Information Processing, ICCWAMTIP 2024

作者： Hanying, Yu Yi, Su Shijiao, Han School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu China

ISBN: (纸本)9798331519254

Thanks to the rapid progress of automatic driving technology, the problem of object detection in automatic driving scenarios has become one of the hottest research topics at present. Automatic driving scenarios face different day and night scenes, complex and variable weather conditions, which further increase the demand for the generalization ability of detection models. At the same time, the automatic driving scenario requires a high response speed for the model, and the object detection model in this scenario is usually a single-stage model. In this paper, a feature adversarial domain adaptation method was proposed for single-stage models, and experiments are conducted on multiple datasets to verify the effectiveness of the proposed algorithm. © 2024 IEEE.

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

Biologically Inspired Spike-UNet for Enhanced and Efficient Image Generati 21

Biologically Inspired Spike-UNet for Enhanced and Efficient ...

引用

21st International computer Conference on Wavelet Active Media technology and Information Processing, ICCWAMTIP 2024

作者： Zheng, Peixiang Shen, Zhao School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu China

ISBN: (纸本)9798331519254

Achieving high-quality image generation remains a challenging problem, particularly when balancing computational efficiency, representational richness, and energy consumption. Traditional deep learning models often lack the biological plausibility and dynamic processing capabilities found in natural neural systems. To address these limitations, we propose SUM (Spike-UNet Model), a novel Spiking Neural Network (SNN) architecture inspired by the U-Net framework By harnessing the temporal dynamics of spiking neurons, SUM aims to enhance image realism while improving computational efficiency. Experiments on benchmark datasets, including MNIST and CIFAR-10, demonstrate that SUM not only outperforms existing SNN architectures but also surpasses conventional Artificial Neural Networks (ANNs) in certain tasks. The architecture effectively captures intricate patterns and textures, resulting in images with enhanced realism and detail. Moreover, its low energy consumption makes it suitable for neuromorphic hardware and edge devices with power constraints. This work underscores the potential of SNNs in computationally intensive visual tasks, demonstrating that biologically inspired architectures can deliver both efficiency and quality. By advancing the integration of spike-based processing into modern AI pipelines, SUM paves the way for more adaptive, sustainable, and high-performing image generation approaches. © 2024 IEEE.

关键词： Pipeline processing systems

来源：评论

学校读者我要写书评

暂无评论

Resource-saving Design Scheme Based on Look-Up Table for Activation Functions 4

Resource-saving Design Scheme Based on Look-Up Table for Act...

引用

4th International Conference on Neural Networks, Information and Communication engineering, NNICE 2024

作者： Yuan, Shuyu University of Electronic Science and Technology of China School of Computer Science and Engineering Chengdu China

ISBN: (纸本)9798350394375

In the field of deep learning, model quantization is an important technique aimed at reducing the storage requirements and computational complexity of models to enable efficient deployment on resource-constrained devices. However, traditional model quantization methods often face high computational demands when dealing with exponential operations, especially those involving activation functions. To address this issue, this paper proposes a LUT-based method for simplifying exponential operations. Look-Up Table (LUT) is a technique that precomputes and stores function values in a table for fast querying. In this paper, LUT is employed to approximate the exponential function, significantly alleviating the computational demands of exponential operations in models. To validate the effectiveness of this method, a series of experiments were conducted. The results demonstrate that using the proposed method for activation functions improves computational speed by 22% compared to using only pseudo-quantization for activation functions. Additionally, in practical applications, this method substantially reduces the L2 error during the quantization process, further demonstrating its superiority in model quantization. © 2024 IEEE.

关键词： Exponential functions

来源：评论

学校读者我要写书评

暂无评论

Solving Directed Multiway Cut Faster Than 2n 32

Solving Directed Multiway Cut Faster Than 2n

引用

32nd Annual European Symposium on Algorithms, ESA 2024

作者： Xiao, Mingyu School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu China

ISBN: (纸本)9783959773386

In the Directed Multiway Cut problem, we are given a directed graph G = (V, E) and a subset T ⊆ V, called the terminal set. The aim is to find a minimum sized set S ⊆ V \ T, such that after deleting S, no directed path exists from one terminal to another terminal in the remaining graph. It has been an open question whether Directed Multiway Cut can be solved faster than the trivial running-time bound O∗(2|V|). In this paper, we provide a positive answer to this question, presenting an algorithm with a running-time bound O(1.9967|V|) © Mingyu Xiao;licensed under Creative Commons License CC-BY 4.0.

关键词： Directed graphs

来源：评论

学校读者我要写书评

暂无评论

Heterogeneous distributed data mining method based on decision tree classification algorithm 5

Heterogeneous distributed data mining method based on decisi...

引用

5th International Conference on computers, Information Processing and Advanced Education, CIPAE 2024

作者： Zheng, Yingzi School of Computer Science and Engineering Guangzhou Institute of Science and Technology Guangzhou510540 China

ISBN: (纸本)9798331527662

The mining of heterogeneous distributed data is a key technology for insight data traceability and classification feature analysis. A heterogeneous distributed data mining method based on decision tree classification algorithm is proposed, and a structured heterogeneous distributed big data label classification model is constructed. The decision tree shift of heterogeneous distributed data is achieved by combining the label classification statistical method. The decision tree classification algorithm is used to achieve label classification statistics and state transfer of distributed big data. Based on the purity deviation difference and direction deviation, the segmentation values in data label classification are reconstructed and the decision tree structure layer is reorganized. The arithmetic difference of each segmentation block of distributed heterogeneous big data is solved, and more embedding capacity is obtained through histogram shift algorithm. The statistical related parameter feature quantity of data is extracted, and screening is carried out according to feature attributes, using mutual information to implement decision tree classification and big data mining for the feature attributes to be embedded. And simulation tests were conducted on the data mining model under different attack models. The simulation results showed that using this method for heterogeneous distributed data mining can design corresponding shift directions and distances for each heterogeneous data distribution area. There will be no errors in the mining process, which improves the traceability and anti attack ability of the data. Moreover, the distortion of the data is small, and the traceability and classification ability of the data are strong. © 2024 IEEE.

关键词： Data consistency

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：