检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Wu, Junyan Xu, Wenbo Lu, Wei Luo, Xiangyang Yang, Rui Guo, Shize MoE Key Laboratory of Information Technology Sun Yat-sen University Guangzhou China State Key Laboratory of Mathematical Engineering and Advanced Computing Zhengzhou China Alibaba Group Hangzhou China

Audio temporal forgery localization (ATFL) aims to find the precise forgery regions of the partial spoof audio that is purposefully modified. Existing ATFL methods rely on training efficient networks using fine-grained annotations, which are obtained costly and challenging in real-world scenarios. To meet this challenge, in this paper, we propose a progressive audio-language co-learning network (LOCO) that adopts co-learning and self-supervision manners to prompt localization performance under weak supervision scenarios. Specifically, an audio-language co-learning module is first designed to capture forgery consensus features by aligning semantics from temporal and global perspectives. In this module, forgery-aware prompts are constructed by using utterance-level annotations together with learnable prompts, which can incorporate semantic priors into temporal content features dynamically. In addition, a forgery localization module is applied to produce forgery proposals based on fused forgery-class activation sequences. Finally, a progressive refinement strategy is introduced to generate pseudo frame-level labels and leverage supervised semantic contrastive learning to amplify the semantic distinction between real and fake content, thereby continuously optimizing forgery-aware features. Extensive experiments show that the proposed LOCO 1 achieves SOTA performance on three public benchmarks. Copyright © 2025, The Authors. All rights reserved.

关键词： Self-supervised learning

来源：评论

学校读者我要写书评

暂无评论

PS-HOMO: A Protection Scheme for iOS SMS 4

PS-HOMO: A Protection Scheme for iOS SMS

引用

2019 4th International Seminar on Computer Technology, Mechanical and Electrical engineering, ISCME 2019

作者： Zhou, Guomiao Wu, Qianqiong Duan, Ming Xi, Qi State Key Laboratory of Mathematical Engineering and Advanced Computing Zhengzhou Henan450000 China Institute of Electrical Engineering and Automation Luoyang Institute of Science and Technology Luoyang Henan471000 China

iOS Short Message Service (SMS) transmits messages in plaintext, which would lead to potential security threats of data. It is meaningful to design a safe and effective SMS. In this paper, all iOS Security Guides are referred to and all the cryptographic algorithms in iOS data protection are analyzed. Because there is no homomorphic encryption, a library of homomorphic encryption algorithms for iOS based on RSA and Paillier is developed. Using this library, a protection scheme PS-HOMO for protecting iOS SMS data is proposed and implemented. Under the premise of ensuring the data confidentiality, PS-HOMO also has the function of ciphertext operation due to the multiplicative homomorphism of RSA and additive homomorphism of Paillier. A comprehensive performance test was performed on PS-HOMO. The results show that the performance is in line with our expectations and will not have a major impact on the original system. The security of PS-HOMO is theoretically analyzed from man-in-the-middle attack, replay attack, and traffic analysis attack. Two application scenarios of the PS-HOMO, remote anonymous voting and privacy telemedicine service are envisaged. We hope PS-HOMO would play a significant role in privacy protection. © 2020 IOP Publishing Ltd. All rights reserved.

关键词： Cryptography

来源：评论

学校读者我要写书评

暂无评论

Improved Impossible Polytopic Attacks on Round-reduced des 4

Improved Impossible Polytopic Attacks on Round-reduced des

引用

2019 4th International Seminar on Computer Technology, Mechanical and Electrical engineering, ISCME 2019

作者： Wang, Jing Wu, Qianqiong Fu, Chaohui Zhou, Guomiao Duan, Ming State Key Laboratory of Mathematical Engineering and Advanced Computing Zhengzhou Henan450000 China College of Electrical Engineering and Automation Luoyang Institute of Science and Technology Luoyang Henan471000 China

In Eurocrypt 2016, Tyge Tiessen introduced the d-difference which considering the differences between multiple plaintexts, and proposed the impossible polytopic attack on block cipher which effectively reducing the data complexity. In this paper, we improve the impossible polytopic attacks of round-reduced DES by some ideas like truncated differentials. Given the input 3-difference of each S-box in the third round, the number of the output 3-difference is actually smaller than the theoretical upper bound, which helps us reduce the memory complexity of the attack on 5-round DES from 212 bytes to 26.9 bytes and increase the success rate of the attack. Using the idea of truncated differentials, the time complexity of the attack on 6-round DES is reduced from 232.2 encryptions to 225.8 encryptions by selecting the output 3-differences of 6 S-boxes for key recovery. We also improve the attack on 7-round DES by using more plaintexts based on our improved attack on 6-round DES. © 2020 IOP Publishing Ltd. All rights reserved.

关键词： Gears

来源：评论

学校读者我要写书评

暂无评论

Distributed Deep Learning System for Efficient Face Recognition in Surveillance System

Distributed Deep Learning System for Efficient Face Recognit...

引用

作者： Jinjin Liu Zhifeng Chen Xiaonan Li Tongxin Wei School of Computer Science Zhongyuan University of Technology State Key Laboratory of Mathematical Engineering and Advanced Computing School of Cyberspace Security Zhengzhou University

In view of the bandwidth consumption caused by data stream transmission in video analysis system and the demand for accurate online real-time analysis of massive data, this paper proposes a deep learning model framework for face recognition employed in the embedded system. Through data collaboration, the cloud could build a more complex data set with a small amount of uploaded data gathered by the end devices. And the framework collaboration makes sure that the fully-trained cloud model directly download or distillate knowledge to the end devices. Experiments show that the deep model not only realizes the real-time response and the accurate response of the cloud system, but also greatly reduces the bandwidth consumption caused by sample data transmission in the model training process.

关键词： Deep learning distributed model data transfer model distillation

来源：评论

学校读者我要写书评

暂无评论

NiCoP nanoleaves array for electrocatalytic alkaline H2 evolution and overall water splitting

引用

Journal of Energy Chemistry 2020年第11期29卷 395-401页

作者： Lei Chen Yaohao Song Yi Liu Liang Xu Jiaqian Qin Yongpeng Lei Yougen Tang State Key Laboratory of Powder Metallurgy Hunan Provincial Key Laboratory of Chemical Power SourcesCollege of Chemistry and Chemical EngineeringCentral South UniversityChangsha 410083HunanChina Energy Materials Computing Center Jiangxi University of Science and TechnologyNanchang 330013JiangxiChina Research Unit of Advanced Materials for Energy Storage Metallurgy and Materials Science Research InstituteChulalongkorn UniversityBangkok 10330Thailand

The development of non-precious, high-efficient and durable electrocatalysts for H2 evolution in alkaline media is highly desirable. Herein we report NiCoP nanoleaves array vertically grown on Ni foam for H2 evolution and overall water splitting via simple hydrothermal treatment and phosphorization. The selfsupported NiCoP nanoleaves architecture contributes to more exposed active sites, the smaller contact resistance between catalyst and substrate, faster ion diffusion and electron transfer. As a result, the optimized electrode requires only overpotentials of 98 and 173 mV to achieve current densities of 10 and100 m A cm-2 in 1.0 M KOH,respectively. Besides, used as both anode and cathode simultaneously, the electrode delivers current densities of 100 and 200 m A cm-2 at cell voltages of only 1.8 and 1.87 V, respectively. Moreover, the relatively high efficiency of about 11.4% for solar-driven water splitting further illustrates the application of our catalyst to sustainable development based on green technologies.

关键词： Electrocatalysis Hydrogen evolution reaction Alkaline media Water splitting Solar-driven electrolysis

来源：评论

学校读者我要写书评

暂无评论

Stealthy Malware Detection Based on Deep Neural Network 2

Stealthy Malware Detection Based on Deep Neural Network

引用

2019 2nd International Symposium on Big Data and Applied Statistics, ISBDAS 2019

作者： Lu, Shoupu Li, Qingbao Zhu, Xinbing State Key Laboratory of Mathematical Engineering and Advanced Computing Henan Province Zhengzhou450001 China Henan University of Economics and Law Henan Province Zhengzhou450000 China

Network attacks using advanced local hiding technology have not only increased, but also become a serious threat. However, attacks using these technologies can not be detected through traffic detection, and some attacks imitate benign traffic to avoid detection. To solve these problems, a malware process detection method based on process behavior in possibly infected terminals is proposed. In this method, a deep neural network is introduced to classify malware processes. Firstly, the recurrent neural network is trained to extract the characteristics of process behavior. Secondly, training convolutional neural network is used to classify feature images generated by trained RNN features. The experiments results show that this method can effectively extract the features of malicious processes, and the AUC of ROC curve is 0.97 in the best case. © Published under licence by IOP Publishing Ltd.

关键词： Malware

来源：评论

学校读者我要写书评

暂无评论

Parallelizing Neural Network Models Effectively on GPU by Implementing Reductions Atomically 22

Parallelizing Neural Network Models Effectively on GPU by Im...

引用

Proceedings of the International Conference on Parallel Architectures and Compilation Techniques

作者： Jie Zhao Cédric Bastoul Yanzhi Yi Jiahui Hu Wang Nie Renwei Zhang Zhen Geng Chong Li Thibaut Tachon Zhiliang Gan State Key Laboratory of Mathematical Engineering and Advanced Computing Zhengzhou China Huawei Technologies France SASU Paris France Huawei Technologies Co. Ltd. Beijing China Huawei Technologies Co. Ltd. Hangzhou China Huawei Technologies Co. Ltd. Shenzhen China

ISBN: (纸本)9781450398688

Due to the missing of a good orchestration of loop transformations, existing optimizing compilers for deploying neural networks on GPU either parallelize reductions ineffectively or miss the fusion opportunities with other operators. Neural network models thus exhibit sub-optimal performance on GPU. We present a practical approach called Panamera for the effective parallelization of reductions in neural networks on GPU. Panamera first leverages loop coalescing to flatten the loop dimensions of reductions, converting all reduction operators into canonical forms eligible for the polyhedral model. Next, Panamera uses polyhedral transformations to reduce the data movements caused by unfused reductions and perform multi-block hardware binding not considered by many compilers. Finally, Panamera embeds a highly optimized routine implemented using GPU atomic instructions, further improving the performance of neural network models while guaranteeing the correctness of parallel reductions. The experimental results demonstrate the effectiveness of our approach: for single operators our code obtains a mean speedup of 33.7×, 3.5×, 5.4× and 9.6× over cuDNN, CUB, TVM and Ansor, for sub-graphs our approach outperforms cuDNN, TVM and Ansor by 9.5×, 2.6× and 2.7×, and for end-to-end workloads, a tensor compiler integrated with our approach outperforms them by 122.5%, 19.3% and 15.2%.

关键词： GPU polyhedral compilation deep learning reduction

来源：评论

学校读者我要写书评

暂无评论

Accurate and Efficient Human Activity Recognition Through Semi-Supervised Deep Learning

引用

IEEE Sensors Journal 2025年第12期25卷 23105-23116页

作者： Long, Xianlei Wang, Ping Gu, Fuqiang Yan, Jinjin Yu, Fangwen Li, Yan Chen, Chao Guo, Songtao Chongqing University College of Computer Science Chongqing400044 China Harbin Engineering University Qingdao Innovation and Development Center Qingdao266400 China Tsinghua University Department of Precision Instrument Beijing China Macquarie University School of Computing Sydney Australia Chongqing University State Key Laboratory of Mechanical Transmission for Advanced Equipment Chongqing400044 China

In recent years, deep learning has been widely used for human activity recognition (HAR) based on wearable sensor data due to its excellent performance. However, deep learning approaches often need a large amount of labeled data to train the model. The collection of huge amounts of labeled activity data is usually labor-intensive and even cost-prohibitive, limiting the practical use of these methods. An effective way to address this issue is to use semi-supervised learning methods. Yet, most existing semi-supervised approaches fail to fully make use of unlabeled data to improve deep learning models' accuracy and generalizability. In this work, we propose an innovative Adaptive Confidence Semi-supervised HAR (ACS-HAR) method that can dynamically adjust the selective confidence threshold based on the model's learning state, thereby making more efficient use of unlabeled data. Experimental results on six public datasets demonstrate that our method outperforms traditional supervised methods by up to 5.20% accuracy and state-of-the-art semi-supervised methods by up to 3.30% *** release our code at https://***/objwww/ACS-HAR. © 2001-2012 IEEE.

关键词： Semi-supervised learning

来源：评论

学校读者我要写书评

暂无评论

引用

IEEE International Conference on Software engineering and Service Sciences (ICSESS)

作者： Duanyi Wang Hui Shu Fei Kang Wenjuan Bu State Key Laboratory of Mathematical Engineering and Advanced Computing Henan Zhengzhou China State Key Laboratory of Mathematical Engineering and Advanced Computing Zhengzhou Henan China

ISBN: (数字)9781728165790

ISBN: (纸本)9781728165806

Recently, graph-based malware similarity analysis has been widely used in the field of malware detection. However, the wide application of code obfuscation, polymorphism, and deformation changes the structure of malicious code, which brings great challenges to the malware similarity analysis. To solve these problems, in this paper, we present a new approach to malware similarity analysis based on the network control structure graph (NCSG). This method analyzed the behavior of malware by application program interface (API) association and constructed NCSG. The graph could reflect the command-and-control(C&C) logic of malware. Therefore, it can resist the interference of code obfuscation technology. The structural features extracted from NCSG will be used as the basis of similarity analysis for training the detection model. Finally, we tested the dataset constructed from five known malware family samples, and the experimental results showed that the accuracy of this method for malware variation analysis reached 92.75%. In conclusion, the malware similarity analysis based on NCSG has a strong application value for identifying the same family of malware.

关键词： Training Software algorithms Resists Feature extraction Malware Strain Software engineering

来源：评论

学校读者我要写书评

暂无评论

A Multi-Variant Voting Algorithm Based on Dynamic Feedback

A Multi-Variant Voting Algorithm Based on Dynamic Feedback

引用

Computer Communication and Network Security (CCNS), International Conference on

作者： Shao Yuwen Zhang Zheng Li Bingzheng Ji Xinsheng State Key Laboratory of Mathematical Engineering and Advanced Computing Zhengzhou China R&D Center National Digital Switching System Engineering & Technological Zhengzhou China

Multi-variant execution framework can protect software security. For the problem of performance loss caused by too many voting times in the existing multi-variant execution framework, this paper proposes a multi-variant voting algorithm based on dynamic feedback, which can make the multi-variant continuously improve its corresponding voting strategy at runtime. Using this algorithm, the multi-variant can get a stable voting range, reduce the number of voting, and effectively improve the performance of multi-variant.

关键词： Runtime Heuristic algorithms Software algorithms Network security Software Computer security

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：