检索结果-内蒙古大学图书馆

International Conference on computer Vision (ICCV)

作者： Ziqi Zhou Shengshan Hu Ruizhi Zhao Qian Wang Leo Yu Zhang Junhui Hou Hai Jin School of Cyber Science and Engineering Huazhong University of Science and Technology National Engineering Research Center for Big Data Technology and System Services Computing Technology and System Lab Hubei Key Laboratory of Distributed System Security Hubei Engineering Research Center on Big Data Security School of Cyber Science and Engineering Wuhan University School of Information and Communication Technology Griffith University Department of Computer Science City University of Hong Kong School of Computer Science and Technology Huazhong University of Science and Technology Cluster and Grid Computing Lab

Self-supervised learning usually uses a large amount of unlabeled data to pre-train an encoder which can be used as a general-purpose feature extractor, such that downstream users only need to perform fine-tuning operations to enjoy the benefit of "large model". Despite this promising prospect, the security of pre-trained encoder has not been thoroughly investigated yet, especially when the pre-trained encoder is publicly available for commercial *** this paper, we propose AdvEncoder, the first framework for generating downstream-agnostic universal adversarial examples based on the pre-trained encoder. AdvEncoder aims to construct a universal adversarial perturbation or patch for a set of natural images that can fool all the downstream tasks inheriting the victim pre-trained encoder. Unlike traditional adversarial example works, the pre-trained encoder only outputs feature vectors rather than classification labels. Therefore, we first exploit the high frequency component information of the image to guide the generation of adversarial examples. Then we design a generative attack framework to construct adversarial perturbations/patches by learning the distribution of the attack surrogate dataset to improve their attack success rates and transferability. Our results show that an attacker can successfully attack downstream tasks without knowing either the pre-training dataset or the downstream dataset. We also tailor four defenses for pre-trained encoders, the results of which further prove the attack ability of AdvEncoder. Our codes are available at: https://***/CGCL-codes/AdvEncoder.

关键词：

来源：评论

学校读者我要写书评

暂无评论

High-integrated photonic tensor core utilizing high-dimensional lightwave and microwave multidomain multiplexing

引用

Light(science & Applications) 2025年第1期14卷 227-237页

作者： Xiangyan Meng Nuannuan Shi Guojie Zhang Junshen Li Ye Jin Shiyou Sun Yichen Shen Wei Li Ninghua Zhu Ming Li Key Laboratory of Optoelectronic Materials and Devices Institute of SemiconductorsChinese Academy of SciencesBeijing100083China College of Materials Science and Opto-Electronic Technology University of Chinese Academy of SciencesBeijing100049China School of Electronic Electrical and Communication EngineeringUniversity of Chinese Academy of SciencesBeijing100049China China Academy of Space Technology(Xi’an) Xi’anShaanxi710100China WeChat Pay Lab 33 Shenzhen Tencent Computer System Co.Ltd.Shenzhen518054China Lightelligence Group Hangzhou311121China

The burgeoning volume of parameters in artificial neural network models has posed substantial challenges to conventional tensor computing *** from the available optical multidimensional information entropy,optical intelligent computing is used as an alternative solution to address the emerging challenges of electrical *** limitations,in terms of device size and photonic integration scale,have hindered the performance of optical ***,an ultrahigh computing density optical tensor processing unit(OTPU),which is grounded in an individual microring resonator(MRR),is introduced to respond to these *** the independent tuning of multiwavelength lasers,the operational capabilities of an MRR are orchestrated,culminating in the formation of an optical tensor *** design facilitates the execution of tensor convolution operations via the lightwave and microwave multidomain hybrid multiplexing in terms of the time,wavelength,and frequency of *** experimental results for the MRR-based OTPU show an extraordinary computing density of 34.04 TOPS/***,the achieved accuracy rate in recognizing MNIST handwritten digits was 96.41%.These outcomes signify a significant advancement toward the realization of high-performance optical tensor processing chips.

关键词： tuning resonator tensor

来源：评论

学校读者我要写书评

暂无评论

DarkSAM: Fooling Segment Anything Model to Segment Nothing

arXiv

引用

arXiv 2024年

作者： Zhou, Ziqi Song, Yufei Li, Minghui Hu, Shengshan Wang, Xianlong Zhang, Leo Yu Yao, Dezhong Jin, Hai National Engineering Research Center for Big Data Technology and System China Services Computing Technology and System Lab China Cluster and Grid Computing Lab China Hubei Engineering Research Center on Big Data Security China Hubei Key Laboratory of Distributed System Security China School of Cyber Science and Engineering Huazhong University of Science and Technology China School of Software Engineering Huazhong University of Science and Technology China School of Information and Communication Technology Griffith University Australia

Segment Anything Model (SAM) has recently gained much attention for its outstanding generalization to unseen data and tasks. Despite its promising prospect, the vulnerabilities of SAM, especially to universal adversarial perturbation (UAP) have not been thoroughly investigated yet. In this paper, we propose DarkSAM, the first prompt-free universal attack framework against SAM, including a semantic decoupling-based spatial attack and a texture distortion-based frequency attack. We first divide the output of SAM into foreground and background. Then, we design a shadow target strategy to obtain the semantic blueprint of the image as the attack target. DarkSAM is dedicated to fooling SAM by extracting and destroying crucial object features from images in both spatial and frequency domains. In the spatial domain, we disrupt the semantics of both the foreground and background in the image to confuse SAM. In the frequency domain, we further enhance the attack effectiveness by distorting the high-frequency components (i.e., texture information) of the image. Consequently, with a single UAP, DarkSAM renders SAM incapable of segmenting objects across diverse images with varying prompts. Experimental results on four datasets for SAM and its two variant models demonstrate the powerful attack capability and transferability of DarkSAM. Our codes are available at: https://***/CGCL-codes/DarkSAM. Copyright © 2024, The Authors. All rights reserved.

关键词： Frequency domain analysis

来源：评论

学校读者我要写书评

暂无评论

Crucial Interaction Elements Detection for 5G Core Network based on the Complex Network Model

Crucial Interaction Elements Detection for 5G Core Network b...

引用

International Conference on Networking and Network Applications (NaNA)

作者： Ruinan Peng Xinhong Hei Yichuan Wang Xiaoxue Liu Yeqiu Xiao School of Computer Science and Engineering Xi’an University of Technology Shaanxi Key Laboratory for Network Computing and Security Technology Xi’an China

ISBN: (数字)9798350376777

ISBN: (纸本)9798350376784

This paper focuses on the issue of crucial interaction elements detection in the 5G core network. Given the current absence of an effective method for quantification evaluating the importance of interaction elements, we define the crucial interaction elements as crucial nodes in this paper, and also undertake a comprehensive exploration of various data interaction scenarios within the 5G core network. This aims to overcome the limitations of traditional network analysis methods, which often operate independently in multi-layer data processing. By extracting multilayer data interaction elements across diverse scenarios, these elements are characterized within a complex network framework. Subsequently, a quantification scoring system for crucial nodes is constructed, incorporating multiple centrality indicators and node weights. Notably, the fitness function within the genetic algorithm is improved to facilitate a dynamic weight optimization strategy, responding to real-time changes in network topology. The optimized weight combination is then allocated to various evaluation indicators within the system, culminating in the design of a quantification scoring model for crucial node. Based on this model, a suite of 5G core network crucial point detection methods, tailored to different data scenarios is proposed. This method enable the quantification evaluation of each node’s importance, realizes crucial interaction elements detection. Such an approach serves to bolster the stable operation of the 5G core network, thereby more effectively meeting modern society’s security imperatives concerning information transmission.

关键词： 5G mobile communication Transfer learning Complex networks Reinforcement learning Nonhomogeneous media Data models Robustness

来源：评论

学校读者我要写书评

暂无评论

Probability analysis of system vulnerability distribution based on the matroid theory

Probability analysis of system vulnerability distribution ba...

引用

International Conference on Networking and Network Applications (NaNA)

作者： Yichuan Wang Peipei Li Xin Song Yeqiu Xiao Xiaoxue Liu School of computer science and engineering Xi’an University of technology Xi’an China Shanxi Key Laboratory for Network Computing and Security Technology

ISBN: (数字)9798350376777

ISBN: (纸本)9798350376784

In recent years, system vulnerability mining techniques have emerged in an endless stream. At present, the existing vulnerability mining techniques are mainly to analyze the source code or monitor and analyze the behavior in order to find security vulnerabilities. Most of these techniques require the configuration of the operating environment. Based on the mapping function defined in the network in this paper, a matroid based on system services and vulnerability points is established by utilizing the marking of the number of function set elements in the mapping function. The existence of vulnerable points in the system is verified by analyzing the properties of this matroid. The basic characteristics of the circuit, base and rank of the established matroid are given, and the properties such as the closure of the matroid are analyzed. The results show that, under certain circumstances, a change in the number of elements in the set of services does not lead to a change in the number of vulnerable points distributed in these services. Using this method, probability analysis of system vulnerability distribution can be performed, which is important for the discovery of network security threats.

关键词： Source coding Network security Monitoring

来源：评论

学校读者我要写书评

暂无评论

Research on Code Virtualization Methods for Cloud Applications

Research on Code Virtualization Methods for Cloud Applicatio...

引用

International Conference on Networking and Network Applications (NaNA)

作者： Zhoukai Wang Zuoyan Xu Yaling Zhang Xin Song Yichuan Wang School of computer science and engineering Xi’an University of technology Xi’an China Shaanxi Key Laboratory for Network Computing and Security Technology

ISBN: (数字)9798350376777

ISBN: (纸本)9798350376784

With the vigorous development of technologies such as the Internet, mobile devices, and the Internet of Things, cloud services have gained a crucial position in our life and have become an essential part of daily operations for modern enterprises and organizations. However, existing cloud application security solutions have failed to address the problem of runtime binary code security, focusing only on vulnerabilities inherent to the applications themselves. Code virtualization, also known as virtualization obfuscation, is a technique that protects software from malicious analysis by obfuscating the code. It hides the code’s control flow and data flow, thereby preventing the code from being decompiled. In this paper, we introduce a code virtualization method, VMENP, which splits the code into multiple modules and uses polymorphic and modular encryption mechanisms to secure binary code at runtime. Specifically, VMENP divides the protected program into several modules, designs a virtual instruction set along with its corresponding interpreter, encrypts each instruction individually after converting it into intermediate representations, and then encrypts it again at the basic block level. After compiling the segmented code into polymorphic bytecode sequences, it embeds the virtual code’s interpreter into the IR. During runtime, the protected code is dynamically decrypted and executed by the basic block level, with each basic block’s storage and execution in memory in a short time window. VMENP identifies functions annotated in the source code to locate the functions that need protection. This paper uses a large amount of $\mathrm{C} / \mathrm{C}++$ code to perform tests in a real cloud environment. Experimental results show that VMENP achieves a good balance between the granularity of protection and the performance overhead of the protected programs, while successfully protecting $\mathrm{C} / \mathrm{C}++$ code and programs running on X86/64 and ARM32/64 architectures

关键词： Cloud computing Codes Runtime Instruction sets Reverse engineering computer architecture Binary codes

来源：评论

学校读者我要写书评

暂无评论

A Novel Multipath Data Transmission Method with Traffic Obfuscation

A Novel Multipath Data Transmission Method with Traffic Obfu...

引用

International Conference on Networking and Network Applications (NaNA)

作者： Xinhong Hei Jianfeng Chen Yichuan Wang Xin Song Lei Zhu School of Computer Science and Engineering Xi’an University of Technology Shaanxi Key Laboratory for Network Computing and Security Technology Xi’an China

ISBN: (数字)9798350376777

ISBN: (纸本)9798350376784

With the rapid progress of network communication technology, the issues of user identity and privacy are critical. However, there are many disadvantages in traditional single-path data transmission, for example, the link is susceptible to monitoring and the traffic feature can be easily analyzed. To solve this issue, this paper introduces a multipath data transmission strategy with traffic obfuscation. First, we design a multipath traffic transmission scheme that divides traffic into multiple sub streams and transmits them through multiple paths. The ability of the communication system to resist surveillance and traffic analysis is enhanced by this approach. Second, we effectively confused multipath traffic by changing the size of the data packets, controlling the delay of multipath transmission, and adding redundant traffic. Through experiments, it has been shown that in this advanced multipath transmission strategy, there is no significant difference in the traffic features generated by each node. The effectiveness of the proposed method in obfuscating communication traffic is demonstrated by this.

关键词： Privacy Surveillance Resists Traffic control Delays Data communication Streams

来源：评论

学校读者我要写书评

暂无评论

MBTDiff: Multi-segmentation Brain Tumor Model with Diffusion Probabilistic Model 9

MBTDiff: Multi-segmentation Brain Tumor Model with Diffusion...

引用

9th International Conference on Image, Vision and computing, ICIVC 2024

作者： Zeng, Shuqi Liu, Jun Xu, Ji Luo, Yuan School of Computer Science and Engineering Wuhan Institute of Technology Wuhan China Key Laboratory of Biomarkers and In Vitro Diagnosis Translation Zhejiang province Hangzhou China Hubei Engineering Research Center for Intelligent Digital Technology in New Power System Hubei Central China Technology Development of Electric Power Co. Ltd Wuhan China

ISBN: (纸本)9798350385991

The diffusion probability model has recently garnered remarkable achievements in image generation, emerging as a prominent research focus within computer vision. It has been observed through numerous studies that the diffusion probability model holds potential for application across various domains within computer vision. Drawing inspiration from these models, we present a novel approach to extend their utility to the realm of brain tumor medical image segmentation. Our method entails training the model utilizing a diffusion framework, followed by iterative refinement of segmentation outcomes by incorporating additional encoding layers and decoders. This method employs an end-to-end training approach without relying on pre-trained backbone networks. Given the probabilistic nature of the results generated by the diffusion model and to simulate the actual decision-making process of doctors, we will generate multiple segmentation results and apply denoising techniques to obtain a stable output. These results will then be integrated into the final segmentation map to achieve more precise segmentation. © 2024 IEEE.

关键词： Image segmentation

来源：评论

学校读者我要写书评

暂无评论

A Comprehensive Overview of Object Detection Based on Deep Learning

A Comprehensive Overview of Object Detection Based on Deep L...

引用

Intelligent Data and Security (IDS), IEEE International Conference on

作者： Gaojing Yuan Linshu Chen Jiahong Cai Chaoyi Yang Jinnian Liu Hunan Key Laboratory for Service Computing and Novel Software Technology School of Computer Science and Engineering Hunan University of Science and Technology Xiangtan China

ISBN: (数字)9798350389487

ISBN: (纸本)9798350389494

Target detection is an important task in the field of computer vision and its application has been extended to the medical field. In the medical field, target detection techniques can be used for diagnosis, treatment and research. For example, by performing target detection on medical images, lesion regions or foci can be automatically identified and localized to help doctors perform accurate diagnosis and treatment. In traditional machine learning methods, commonly used features include morphological features, texture features, and histogram features. And in deep learning methods, the commonly used structures include Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN), etc. In addition, there are many special problems and challenges for medical images, such as image quality, variation between modalities, and lack of data. Therefore, target detection in the medical field requires special techniques and algorithms to address these issues. By integrating expertise in the medical field, combining traditional and deep learning techniques, and developing improved and optimized algorithms for the special problems of medical images, the accuracy and efficiency of medical target detection can be improved, thus contributing to the advancement of medical research and clinical practice.

关键词： Deep learning Image quality Recurrent neural networks Accuracy Object detection Medical services Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Compiler Differentiation Detection Method for Binary Similarity

Compiler Differentiation Detection Method for Binary Similar...

引用

International Conference on Networking and Network Applications (NaNA)

作者： Xiaoxue Liu Xinwei Duan Xinhong Hei Yichuan Wang Xin Song School of computer science and engineering Xi’an University of technology Xi’an China Shaanxi Key Laboratory for Network Computing and Security Technology

ISBN: (数字)9798350376777

ISBN: (纸本)9798350376784

Compiler differential detection can be performed by detecting the similarity between the binary compiled by the compiler and the ideal compiled file. This article proposes a javac compiler differentiation detection system based on binary file similarity: we extract opcode sequences from binary files, use BERT to extract features of the opcodes, and finally obtain the similarity score by calculating the distance between vectors. Under the premise of ensuring the retrieval speed requirements, this algorithm is superior to most existing models in accuracy, with the accuracy rate of $99.7 \%$. It shows that the method in this article can effectively detect whether two binary files are similar. This method will have a good effect on detecting whether the javac compiler has been invaded by malicious code in the future.

关键词： Accuracy Limiting Source coding Semantics Neural networks Feature extraction Malware

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：