检索结果-内蒙古大学图书馆

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Huang, Wenke Ye, Mang Shi, Zekun Wan, Guancheng Li, He Du, Bo National Engineering Research Center for Multimedia Software Institute of Artificial Intelligence Hubei Key Laboratory of Multimedia and Network Communication Engineering School of Computer Science Wuhan University Wuhan China Taikang Center for Life and Medical Sciences Wuhan University Wuhan China

Backdoor attacks pose a serious threat to federated systems, where malicious clients optimize on the triggered distribution to mislead the global model towards a predefined target. Existing backdoor defense methods typically require either homogeneous assumption, validation datasets, or client optimization conflicts. In our work, we observe that benign heterogeneous distributions and malicious triggered distributions exhibit distinct parameter importance degrees. We introduce the Fisher Discrepancy Cluster and Rescale (FDCR) method, which utilizes Fisher Information to calculate the degree of parameter importance for local distributions. This allows us to reweight client parameter updates and identify those with large discrepancies as backdoor attackers. Furthermore, we prioritize rescaling important parameters to expedite adaptation to the target distribution, encouraging significant elements to contribute more while diminishing the influence of trivial ones. This approach enables FDCR to handle backdoor attacks in heterogeneous federated learning environments. Empirical results on various heterogeneous federated scenarios under backdoor attacks demonstrate the effectiveness of our method. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Zenith: Real-time Identification of DASH Encrypted Video Traffic with Distortion 24

Zenith: Real-time Identification of DASH Encrypted Video Tra...

引用

32nd ACM International Conference on Multimedia, MM 2024

作者： Tang, Weitao Li, Jianqiang Du, Meijie Hu, Die Liu, Qingyun Institute of Information Engineering Chinese Academy of Sciences School of Cyber Security University of Chinese Academy of Sciences Beijing China National Computer Network Emergency Response Technical Team/Coordination Center of China Beijing China

ISBN: (纸本)9798400706868

Some video traffic carries harmful content, such as hate speech and child abuse, primarily encrypted and transmitted through Dynamic Adaptive Streaming over HTTP (DASH). Promptly identifying and intercepting traffic of harmful videos is crucial in network regulation. However, QUIC is becoming another DASH transport protocol in addition to TCP. On the other hand, complex network environments and diverse playback modes lead to significant distortions in traffic. The issues above have not been effectively addressed. This paper proposes a real-time identification method for DASH encrypted video traffic with distortion, named Zenith. We extract stable video segment sequences under various itags as video fingerprints to tackle resolution changes and propose a method of traffic fingerprint extraction under QUIC and VPN. Subsequently, simulating the sequence matching problem as a natural language problem, we propose Traffic Language Model (TLM), which can effectively address video data loss and retransmission. Finally, we propose a frequency dictionary to accelerate Zenith's speed further. Zenith significantly improves accuracy and speed compared to other SOTA methods in various complex scenarios, especially in QUIC, VPN, automatic resolution, and low bandwidth. Zenith requires traffic for just half a minute of video content to achieve precise identification, demonstrating its real-time effectiveness. © 2024 Owner/Author.

关键词： Video streaming

来源：评论

学校读者我要写书评

暂无评论

FedSSP: Federated Graph Learning with Spectral Knowledge and Personalized Preference 38

FedSSP: Federated Graph Learning with Spectral Knowledge and...

引用

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Tan, Zihan Wan, Guancheng Huang, Wenke Ye, Mang National Engineering Research Center for Multimedia Software Institute of Artificial Intelligence Hubei Key Laboratory of Multimedia and Network Communication Engineering School of Computer Science Wuhan University Wuhan China Taikang Center for Life and Medical Sciences Wuhan University Wuhan China

Personalized Federated Graph Learning (pFGL) facilitates the decentralized training of Graph Neural networks (GNNs) without compromising privacy while accommodating personalized requirements for non-IID participants. In cross-domain scenarios, structural heterogeneity poses significant challenges for pFGL. Nevertheless, previous pFGL methods incorrectly share non-generic knowledge globally and fail to tailor personalized solutions locally under domain structural shift. We innovatively reveal that the spectral nature of graphs can well reflect inherent domain structural shifts. Correspondingly, our method overcomes it by sharing generic spectral knowledge. Moreover, we indicate the biased message-passing schemes for graph structures and propose the personalized preference module. Combining both strategies, we propose our pFGL framework FedSSP which Shares generic Spectral knowledge while satisfying graph Preferences. Furthermore, We perform extensive experiments on cross-dataset and cross-domain settings to demonstrate the superiority of our framework. The code is available at https://***/OakleyTan/FedSSP. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Brain Tumor Detection Using Image Processing 5

Brain Tumor Detection Using Image Processing

引用

5th International Conference on Advancements in Computing, ICAC 2023

作者： Wijerathna, Udari Dissanayake, Nimavi Nimasha, Dinithi Senarathne, Udara Weerasinghe, Lokeesha Kahandawaarachchi, Chathurangika Sri Lanka Institute of Information Technology Faculty of Computing Dept of Computer System Engineering Malabe Sri Lanka Sri Lanka Institute of Information Technology Faculty of Computing Dept of Information Technology Malabe Sri Lanka Sri Lanka Institute of Information Technology Faculty of Computing Dept of Computer System Network Engineering Malabe Sri Lanka

ISBN: (纸本)9798350358131

Abnormal formation of cells in a layer. The brain is known as the brain cancer complex, and a brain cancer detection system is needed to detect cancer in its early stages Phase. As a result, the primary goal of this initiative is to assist existing doctors in their training for the early identification o f brain cancer by utilizing magnetic resonance imaging (MRI), which is the most effective approach to identifying brain cancer. This system is done through four components, and it has made it easier to detect brain tumors. These techniques with different image processing techniques include noise removal, segmentation, classification, a nd u se o fa nalysis segments f or detection. Pre-processing techniques prepare the image for further processing by removing image noise. During segmentation, the brain tumor system determines the size of the brain tumor and estimates its size. Specialists should examine them and classify brain tumors as meningiomas, gliomas, and pituitary tumors. In analysis, this component is expected to be designed so that users or medical experts can know the future through a dashboard. We explored the methodology of using generative adversarial networks (GANs) implemented in TensorFlow to denominate brain tumor images. Analyses the outcomes of various types of established deep learning models on the brain tumor MR image datasets, such as classification methods ResNet-50, VGG 16, CNN, and Inception V3. The validation precision, AUC, recall, and validation loss for CNN are 95.3%, 99.43%, 92.1 %, and 0.360 adjustments respectively. © 2023 IEEE.

关键词： Magnetic resonance imaging

来源：评论

学校读者我要写书评

暂无评论

Malware Variant Detection Based on Knowledge Transfer and Ensemble Learning 6th

Malware Variant Detection Based on Knowledge Transfer and ...

引用

6th International Conference on Science of Cyber Security, SciSec 2024

作者： Ding, Yu Sun, Haoliang Li, Binbin Qi, Zisen Jia, Siyu Wang, Haiping Tan, XingBang Institute of Information Engineering Chinese Academy of Sciences Beijing China National Computer Network Emergency Response Technical Team/Coordination Center of China Beijing China School of Cyber Security University of Chinese Academy of Sciences Beijing China

ISBN: (纸本)9789819624164

In the realm of cybersecurity, malware adopts various evasion tactics, such as obfuscation and code rewriting, to evade detection by network security protection systems, perpetually evolving in sophistication. This presents a formidable challenge for traditional signature-based and machine learning-based malware detection methods, rendering them ineffective against emerging malware variants. To mitigate this problem, security analysts strive to detect malware variant samples at the earliest possible stage. This typically involves labor-intensive manual analysis, where analysts meticulously scrutinize network activities, system calls, and other suspicious behaviors. However, given the vast volume of potentially malicious samples identified daily by network security systems, discerning the ‘high-value’ samples worthy of further investigation becomes a daunting task. To address this challenge, we propose a novel malware variant detection method rooted in knowledge transfer and ensemble learning. This method aims to detect and identify malware variant samples, streamlining the process of screening out highly suspected files for security analysts. By doing so, it alleviates the burden of manual analysis and judgment. This study leveraged the Microsoft Kaggle dataset to simulate real-world scenarios of malware variant detection. The results demonstrate a recall rate of 61.9% for malware variant samples, with an accuracy rate of 56.52%. Notably, security analysts can pinpoint a malware variant sample by analyzing an average of 1.77 samples, significantly reducing the manual analysis workload. These experimental findings underscore the effectiveness and robustness of our proposed method. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

Automatic Alert Categories Standardization for Heterogeneous Devices with Incomplete Semantic Knowledge Based on LSTM 6th

Automatic Alert Categories Standardization for Heterogeneou...

引用

6th International Conference on Science of Cyber Security, SciSec 2024

作者： Wang, Haiping Li, Jianqiang Li, Binbin Zang, Tianning Yang, Yifei Jia, Siyu Qi, Zisen Ding, Yu Institute of Information Engineering Chinese Academy of Sciences Beijing China School of Cyber Security University of Chinese Academy of Sciences Beijing China National Computer Network Emergency Response Technical Team/Coordination Center of China Beijing China

ISBN: (纸本)9789819624164

In the realm of cybersecurity situational awareness, systems compile alert logs from diverse network monitors. These logs exhibit large-scale dimensions, numerous alert types, and autonomous alert categories. Ensuring precise and efficacious analysis necessitates the standardization and incorporation of these logs into a unified alert type categorization. Conventionally, security experts can manually categorize and standardize alert types by scrutinizing their semantic descriptions. However, this approach is inherently subjective and limited to alerts with comprehensive descriptions already known to the experts. To confront these challenges, we pioneer a groundbreaking method that automates alert type standardization without dependence on semantic information, harnessing solely the inherent attack behavior attributes and attacker-victim ratios extracted from the logs themselves. Specifically, we commence by encapsulating each alert type within a 27-dimensional feature vector, derived from its corresponding alert logs over a defined time frame. Subsequently, we utilize Long Short-Term Memory (LSTM) neural networks to homogenize these alert types. We undertake experiments on online alert logs, achieving an outstanding 92.308% accuracy in automatically categorizing 223 distinct alert types into 12 standardized classifications. These outcomes attest to the significant enhancement in efficiency and precision engendered by our method in alert categorization. Given that the alert logs encompass data from multiple NDR (network Detection and Response Device) devices, this experimentation further substantiates the robustness of the proposed methodology. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Classification (of information)

来源：评论

学校读者我要写书评

暂无评论

Sorted Consecutive Occurrence Queries in Substrings

arXiv

引用

arXiv 2024年

作者： Akram, Waseem Mieno, Takuya Department of Computer Science and Engineering Indian Institute of Technology Kanpur India Department of Computer and Network Engineering University of Electro-Communications Chofu Japan

The string indexing problem is a fundamental computational problem with numerous applications, including information retrieval and bioinformatics. It aims to efficiently solve the pattern matching problem: given a text T of length n for preprocessing and a pattern P of length m as a query, the goal is to report all occurrences of P as substrings of T. Navarro and Thankachan [CPM 2015, Theor. Comput. Sci. 2016] introduced a variant of this problem called the gap-bounded consecutive occurrence query, which reports pairs of consecutive occurrences of P in T such that their gaps (i.e., the distances between them) lie within a query-specified range [g1, g2]. Recently, Bille et al. [FSTTCS 2020, Theor. Comput. Sci. 2022] proposed the top-k close consecutive occurrence query, which reports the k closest consecutive occurrences of P in T, sorted in non-descending order of distance. Both problems are optimally solved in query time with O(n log n)-space data structures. In this paper, we generalize these problems to the range query model, which focuses only on occurrences of P in a specified substring T[a..b] of T. Our contributions are as follows: We propose an O(n log2 n)-space data structure that answers the range top-k consecutive occurrence query in O(|P| + log log n + k) time. We propose an O(n log2+ϵ n)-space data structure that answers the range gap-bounded consecutive occurrence query in O(|P| + log log n + output) time, where ϵ is a positive constant and output denotes the number of outputs. Additionally, as by-products, we present algorithms for geometric problems involving weighted horizontal segments in a 2D plane, which are of independent interest. Copyright © 2024, The Authors. All rights reserved.

关键词： Trees (mathematics)

来源：评论

学校读者我要写书评

暂无评论

Using GPT and authentic contextual recognition to generate math word problems with difficulty levels

引用

Education and Information Technologies 2024年第13期29卷 1-29页

作者： Hwang, Wu-Yuin Utami, Ika Qutsiati Graduate Institute of Network Learning Technology National Central University Taoyuan 320317 Taiwan Department of Computer Science and Information Engineering National Dong Hwa University Hualien 974301 Taiwan Data Science Technology Engineering Department Faculty of Advanced Technology and Multidiscipline Universitas Airlangga Surabaya 60115 Indonesia

Automatic generation of math word problems (MWPs) is a challenging task in Natural Language Processing (NLP), particularly connecting it to real-life problems because it can benefit students in developing a higher level of mathematical thinking. However, most of the MWPs are presented within a scholastic setting in a decontextualized way. This paper describes a prototype system that generates authentic math word problems (i.e., real-life problems) using authentic contextual recognition with controlled difficulty levels. Our innovative approach includes the acquisition of authentic contextual information through recognition technology, an instructional-based prompt generator for three different difficulty levels, and question generation through the Generative Pre-trained Transformer (GPT) model. We evaluated the performance of the system in terms of the quality of generated questions using automatic evaluation and human evaluation. Further, we assessed the usability of the system using heuristic evaluation. The automatic evaluation showed our generated MWPs were relevant for geometry topics and varied in sentence generation. The human evaluation found our system generated more realistic problems with satisfactory language quality and mathematical validity. Our system also produced more questions with higher difficulty levels based on human evaluation. Heuristic evaluation captured the usability of the system and highlighted the potential to be applied in a pedagogical context for mathematics learning. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.

关键词： Authentic context Difficulty level Math word problems Natural language processing Primary school Question generation Technology education

来源：评论

学校读者我要写书评

暂无评论

An Efficient IOC-Driven BigData Tracing and Backtracking Model for Emergency Response 6th

An Efficient IOC-Driven BigData Tracing and Backtracking Mo...

引用

6th International Conference on Science of Cyber Security, SciSec 2024

作者： Wang, Haiping Li, Jianqiang Li, Binbin Zang, Tianning Qi, Zisen Jia, Siyu Ding, Yu Yang, Yifei Institute of Information Engineering Chinese Academy of Sciences Beijing China School of Cyber Security University of Chinese Academy of Sciences Beijing China National Computer Network Emergency Response Technical Team/Coordination Center of China Beijing China

ISBN: (纸本)9789819624164

In the realm of intelligence-driven emergency response, utilizing Indicators of Compromise (IOCs) like IP addresses and domain names is crucial for the swift analysis of and response to cyber-attacks within vast network security logs. However, with daily log increases of hundreds of TBs and total volumes reaching into the hundreds of PBs, the tasks of tracking and backtracking critical data face significant challenges, including substantial computational resource consumption and extended processing times. In this paper, we introduce a data tracking and backtracking model based on the principles of data tiered organization and pre-tracking backtracking to enhance emergency response efficiency and reduce resource consumption. The model initially scores IOC clues based on dimensions such as activity and confidence levels, periodically filtering and caching those with higher scores. Based on these clues, it pre-tracks and backtracks associated data, thereby avoiding the computational overhead caused by multiple repetitive offline filtering operations. Furthermore, to improve the retrieval efficiency of threat intelligence clues in scenarios involving massive data volumes, we have designed a three-tier data pre-tracking and backtracking state index to manage the data involved in pre-tracking and backtracking. We also propose a data query task decomposition algorithm to optimize the data retrieval process and enhance emergency response efficiency. Experimental results demonstrate that our method reduces processing time by more than 85.6% compared to traditional direct raw data access, highlighting our model’s real-world efficiency and practicality. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： network security

来源：评论

学校读者我要写书评

暂无评论

A Theory of Transfer-Based Black-Box Attacks: Explanation and Implications 37

A Theory of Transfer-Based Black-Box Attacks: Explanation an...

引用

37th Conference on Neural Information Processing Systems, NeurIPS 2023

作者： Chen, Yanbo Liu, Weiwei School of Computer Science Wuhan University National Engineering Research Center for Multimedia Software Wuhan University Institute of Artificial Intelligence Wuhan University Hubei Key Laboratory of Multimedia and Network Communication Engineering Wuhan University

ISBN: (纸本)9781713899921

Transfer-based attacks [1] are a practical method of black-box adversarial attacks in which the attacker aims to craft adversarial examples from a source model that is transferable to the target model. Many empirical works [2-6] have tried to explain the transferability of adversarial examples from different angles. However, these works only provide ad hoc explanations without quantitative analyses. The theory behind transfer-based attacks remains a mystery. This paper studies transfer-based attacks under a unified theoretical framework. We propose an explanatory model, called the manifold attack model, that formalizes popular beliefs and explains the existing empirical results. Our model explains why adversarial examples are transferable even when the source model is inaccurate as observed in Papernot et al. [7]. Moreover, our model implies that the existence of transferable adversarial examples depends on the "curvature" of the data manifold, which further explains why the success rates of transfer-based attacks are hard to improve. We also discuss our model's expressive power and applicability. © 2023 Neural information processing systems foundation. All rights reserved.

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：