Article(id=1149768944640241740, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1149768937925165147, articleNumber=null, orderNo=null, doi=10.12404/j.issn.1671-1815.2404568, pmid=null, cstr=null, oa=null, hot=null, price=null, onlineType=0, articleFormat=0, articleType=null, articleTypeStr=null, receivedDate=1718640000000, receivedDateStr=2024-06-18, revisedDate=1763308800000, revisedDateStr=2025-11-17, acceptedDate=null, acceptedDateStr=null, onlineDate=1752055878075, onlineDateStr=2025-07-09, pubDate=1748361600000, pubDateStr=2025-05-28, doiRegisterDate=null, doiRegisterDateStr=null, onlineIssueDate=1752055878075, onlineIssueDateStr=2025-07-09, onlineJustAcceptDate=null, onlineJustAcceptDateStr=null, onlineFirstDate=null, onlineFirstDateStr=null, sourceXml=null, magXml=null, createTime=1752055878075, creator=13701087609, updateTime=1752055878075, updator=13701087609, issue=Issue{id=1149768937925165147, tenantId=1146029695717560320, journalId=1146123166801305609, year='2025', volume='25', issue='15', pageStart='6155', pageEnd='6586', issueExtLink='null', onlineDate='null', pubDate='null', beforeIssueId=null, nextIssueId=null, price=null, status=1, issueComplete=1, articleOrder=1, issueType=-1, specialIssue=0, createTime=1752055876475, creator=13701087609, updateTime=1768456822194, updator=13701087609, preIssue=null, nextIssue=null, ext={EN=IssueExt(id=1218559490207699090, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1149768937925165147, language=EN, specialIssueTitle=, coverIllustrator=, specialIssueEditor=, specialIssueAbout=), CN=IssueExt(id=1218559490211893395, tenantId=1146029695717560320, journalId=1146123166801305609, issueId=1149768937925165147, language=CN, specialIssueTitle=, coverIllustrator=, specialIssueEditor=, specialIssueAbout=)}, issueFiles=null}, startPage=6419, endPage=6430, ext={EN=ArticleExt(id=1149768944791236685, articleId=1149768944640241740, tenantId=1146029695717560320, journalId=1146123166801305609, language=EN, title=Network Traffic Classification Method Improved Based on Data Augmentation and CNN-Optuna-Attention, columnId=1156262729162810294, journalTitle=Science Technology and Engineering, columnName=Papers·Automation and Computational Technology, runingTitle=null, highlight=null, articleAbstract=

In order to improve the accuracy of network traffic classification, a traffic classification method combining an attention mechanism and a convolutional neural network was proposed. An attention mechanism layer was designed and implemented on the basis of the convolutional neural network model, which received the output of the fully connected layer as input, calculated the weight of the input features, and multiplied it by the original features to strengthen the key features. This, in turn, helped to improve the model's ability to capture key information. Secondly, in order to solve the problem that the model was overfitting to the high-proportion category due to the unbalanced sample number of network traffic categories, and it was difficult to identify the small-proportion categories, a method to augment the dataset was proposed. Considering the perspective of hyperparameter combination optimization, a hyperparameter search strategy based on Bayesian optimization and five-fold cross-validation was proposed to optimize the hyperparameter combination of the model. The combination of hyperparameters of the model was determined by the above methods. The public dataset was used for the above experiments and model tests. The results show that compared with other methods, the overall accuracy, precision, and F1 score are significantly improved, which verifies that the proposed method has better classification performance.

, correspAuthors=Xin CUI, authorNote=null, correspAuthorsNote=null, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=null, magXml=null, pdfUrl=null, pdf=null, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=null, mapNumber=null, authorCompany=null, fund=null, authors=null, authorsList=Hao-yao TANG, Xin CUI, Yi-wei ZHANG, Qing-hui ZHAO), CN=ArticleExt(id=1149768980300214425, articleId=1149768944640241740, tenantId=1146029695717560320, journalId=1146123166801305609, language=CN, title=基于数据增强与CNN-Optuna-Attention改进的网络流量分类方法, columnId=1156262729783567290, journalTitle=科学技术与工程, columnName=论文·自动化技术、计算机技术, runingTitle=null, highlight=null, articleAbstract=

针对基于卷积神经网络的流量分类方法难以捕捉序列中不同部分的重要性、特征提取不足的问题,为提升网络流量分类精度,提出一种注意力机制与卷积神经网络相结合的流量分类方法,在卷积神经网络模型的基础上设计实现一种注意力机制层,通过接收全连接层的输出作为输入,计算输入特征的权重并乘以原始特征,实现对关键特征的加强,进而有助于提高模型对关键信息的捕捉能力。其次针对网络流量类别样本数不均衡导致模型过拟合于高比例类别,难以识别分类小比例类别的问题,提出了一种对数据集进行数据增强的方法。并且考虑到超参数组合优化的角度,提出一种基于贝叶斯优化的超参数搜索策略和五折交叉验证的方式对模型的超参数组合进行优化。通过上述方法研究确定模型的超参数组合。使用公开数据集进行上述实验与模型测试,结果表明:与其他方法相比,总体准确率、精确率以及F1分数都有明显的提升,验证了本文所提方法具有更好的分类性能。

, correspAuthors=崔鑫, authorNote=null, correspAuthorsNote=
* 崔鑫(1972—),女,汉族,山东淄博人,博士,副教授。研究方向:下一代互联网技术、网络安全、网络大数据、无线传感网。E-mail:
, copyrightStatement=null, copyrightOwner=null, extLink=null, articleAbsUrl=null, sourceXml=pfEH0cHldLysDdvb3C1srw==, magXml=9JNtIfllDlqWluKikLpHUg==, pdfUrl=null, pdf=Sjy4TRRqw8E5PqSZvrPs0Q==, pdfFileSize=null, pdfExtLink=null, richHtmlUrl=null, mobilePdfUrl=null, reviewReport=null, pdfFirstPage=null, abstractGraph=null, abstractGraphContent=null, abstractVideo=null, citation=null, cebUrl=null, magXmlContent=nTup8tIZWzjWksIPJngJXQ==, mapNumber=null, authorCompany=null, fund=null, authors=

唐浩耀(2000—),男,汉族,山东济南人,硕士研究生。研究方向:网络服务与信息安全。E-mail:

, authorsList=唐浩耀, 崔鑫, 张艺炜, 赵庆慧)}, authors=[Author(id=1172924320054653089, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, orderNo=0, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=1393124747@qq.com, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1172924320172093603, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, authorId=1172924320054653089, language=EN, stringName=Hao-yao TANG, firstName=Hao-yao, middleName=null, lastName=TANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Computer Science and Technology, Shandong University of Technology, Zibo 255049, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1172924320230813860, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, authorId=1172924320054653089, language=CN, stringName=唐浩耀, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=山东理工大学计算机科学与技术学院, 淄博 255049, bio={"content":"

唐浩耀(2000—),男,汉族,山东济南人,硕士研究生。研究方向:网络服务与信息安全。E-mail:

"}, bioImg=null, bioContent=

唐浩耀(2000—),男,汉族,山东济南人,硕士研究生。研究方向:网络服务与信息安全。E-mail:

, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1172924319958184093, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, xref=null, ext=[AuthorCompanyExt(id=1172924319962378398, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, companyId=1172924319958184093, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Computer Science and Technology, Shandong University of Technology, Zibo 255049, China), AuthorCompanyExt(id=1172924319966572703, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, companyId=1172924319958184093, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=山东理工大学计算机科学与技术学院, 淄博 255049)])]), Author(id=1172924320302117030, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, orderNo=1, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=cx@sdut.edu.cn, emailSecond=null, emailThird=null, correspondingAuthor=1, authorType=1, ext={EN=AuthorExt(id=1172924320369225896, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, authorId=1172924320302117030, language=EN, stringName=Xin CUI, firstName=Xin, middleName=null, lastName=CUI, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=School of Computer Science and Technology, Shandong University of Technology, Zibo 255049, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1172924320448917673, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, authorId=1172924320302117030, language=CN, stringName=崔鑫, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=*, address=山东理工大学计算机科学与技术学院, 淄博 255049, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1172924319958184093, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, xref=null, ext=[AuthorCompanyExt(id=1172924319962378398, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, companyId=1172924319958184093, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Computer Science and Technology, Shandong University of Technology, Zibo 255049, China), AuthorCompanyExt(id=1172924319966572703, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, companyId=1172924319958184093, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=山东理工大学计算机科学与技术学院, 淄博 255049)])]), Author(id=1172924320532803755, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, orderNo=2, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1172924320620884141, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, authorId=1172924320532803755, language=EN, stringName=Yi-wei ZHANG, firstName=Yi-wei, middleName=null, lastName=ZHANG, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Computer Science and Technology, Shandong University of Technology, Zibo 255049, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1172924320708964526, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, authorId=1172924320532803755, language=CN, stringName=张艺炜, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=山东理工大学计算机科学与技术学院, 淄博 255049, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1172924319958184093, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, xref=null, ext=[AuthorCompanyExt(id=1172924319962378398, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, companyId=1172924319958184093, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Computer Science and Technology, Shandong University of Technology, Zibo 255049, China), AuthorCompanyExt(id=1172924319966572703, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, companyId=1172924319958184093, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=山东理工大学计算机科学与技术学院, 淄博 255049)])]), Author(id=1172924320776073392, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, orderNo=3, firstName=null, middleName=null, lastName=null, nameCn=null, orcid=null, stid=null, country=null, authorPic=null, dead=0, email=null, emailSecond=null, emailThird=null, correspondingAuthor=0, authorType=1, ext={EN=AuthorExt(id=1172924320847376562, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, authorId=1172924320776073392, language=EN, stringName=Qing-hui ZHAO, firstName=Qing-hui, middleName=null, lastName=ZHAO, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=School of Computer Science and Technology, Shandong University of Technology, Zibo 255049, China, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null), CN=AuthorExt(id=1172924320910291123, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, authorId=1172924320776073392, language=CN, stringName=赵庆慧, firstName=null, middleName=null, lastName=null, prefix=null, suffix=null, authorComment=null, nameInitials=null, affiliation=null, department=null, xref=null, address=山东理工大学计算机科学与技术学院, 淄博 255049, bio=null, bioImg=null, bioContent=null, aboutCorrespAuthor=null)}, companyList=[AuthorCompany(id=1172924319958184093, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, xref=null, ext=[AuthorCompanyExt(id=1172924319962378398, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, companyId=1172924319958184093, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Computer Science and Technology, Shandong University of Technology, Zibo 255049, China), AuthorCompanyExt(id=1172924319966572703, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, companyId=1172924319958184093, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=山东理工大学计算机科学与技术学院, 淄博 255049)])])], keywords=[Keyword(id=1172924321174532276, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, orderNo=1, keyword=deep learning), Keyword(id=1172924321250029749, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, orderNo=2, keyword=traffic classification), Keyword(id=1172924321312944310, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, orderNo=3, keyword=attention mechanisms), Keyword(id=1172924321388441783, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, orderNo=4, keyword=convolutional neural networks), Keyword(id=1172924321531048120, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, orderNo=5, keyword=hyperparameter optimization), Keyword(id=1172924321673654457, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, orderNo=6, keyword=five-fold cross-validation), Keyword(id=1172924321791094970, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, orderNo=1, keyword=深度学习), Keyword(id=1172924321845620923, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, orderNo=2, keyword=流量分类), Keyword(id=1172924321929507004, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, orderNo=3, keyword=注意力机制), Keyword(id=1172924321975644349, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, orderNo=4, keyword=卷积神经网络), Keyword(id=1172924322046947518, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, orderNo=5, keyword=超参数优化), Keyword(id=1172924322105667775, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, orderNo=6, keyword=五折交叉验证)], refs=[Reference(id=1172924326891368713, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=https://www.cnnic.net.cn/n4/2024/0321/c208-10962.html, language=null, rfNumber=[1], rfOrder=0, authorNames=中国互联网信息中心CNNIC, journalName=null, refType=null, unstructuredReference=中国互联网信息中心CNNIC. 第53次《中国互联网络发展状况统计报告》[EB/OL]. (2024-03-22) [2024-06-18]. https://www.cnnic.net.cn/n4/2024/0321/c208-10962.html., articleTitle=第53次《中国互联网络发展状况统计报告》, refAbstract=null), Reference(id=1172924326992032011, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=null, volume=null, issue=null, pageStart=null, pageEnd=null, url=https://www.cnnic.net.cn/n4/2024/0321/c208-10962.html, language=null, rfNumber=[1], rfOrder=1, authorNames=China Internet Information Center CNNIC, journalName=null, refType=null, unstructuredReference=China Internet Information Center CNNIC. The 53rd statistical report on the internet development in China[EB/OL]. (2024-03-22) [2024-06-18] https://www.cnnic.net.cn/n4/2024/0321/c208-10962.html., articleTitle=The 53rd statistical report on the internet development in China, refAbstract=null), Reference(id=1172924327117861133, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2022, volume=28, issue=21, pageStart=77, pageEnd=79, url=null, language=null, rfNumber=[2], rfOrder=2, authorNames=郑春艳, journalName=现代计算机, refType=null, unstructuredReference=郑春艳. 浅谈计算机网络发展现状[J]. 现代计算机, 2022, 28(21): 77-79., articleTitle=浅谈计算机网络发展现状, refAbstract=null), Reference(id=1172924327264661777, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2022, volume=28, issue=21, pageStart=77, pageEnd=79, url=null, language=null, rfNumber=[2], rfOrder=3, authorNames=Zheng Chunyan, journalName=Modern Computer, refType=null, unstructuredReference=Zheng Chunyan. Introduction to the current status of computer network development[J]. Modern Computer, 2022, 28(21): 77-79., articleTitle=Introduction to the current status of computer network development, refAbstract=null), Reference(id=1172924327361130771, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2022, volume=10, issue=3, pageStart=676, pageEnd=692, url=null, language=null, rfNumber=[3], rfOrder=4, authorNames=Azab A, Khasawneh M, Alrabaee S, journalName=Digital Communications and Networks, refType=null, unstructuredReference=Azab A, Khasawneh M, Alrabaee S, et al. Network traffic classification: techniques, datasets, and challenges[J]. Digital Communications and Networks, 2022, 10(3): 676-692., articleTitle=Network traffic classification: techniques, datasets, and challenges, refAbstract=null), Reference(id=1172924327465988372, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2022, volume=2, issue=3, pageStart=136, pageEnd=144, url=null, language=null, rfNumber=[4], rfOrder=5, authorNames=Munarko A, journalName=AKADEMIK: Jurnal Mahasiswa Humanis, refType=null, unstructuredReference=Munarko A. Determinants of service quality: efficiency and workspace[J]. AKADEMIK: Jurnal Mahasiswa Humanis, 2022, 2(3): 136-144., articleTitle=Determinants of service quality: efficiency and workspace, refAbstract=null), Reference(id=1172924327596011800, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2023, volume=null, issue=11, pageStart=148, pageEnd=150, url=null, language=null, rfNumber=[5], rfOrder=6, authorNames=任成刚, journalName=网络安全和信息化, refType=null, unstructuredReference=任成刚. 大数据下的计算机网络安全技术分析[J]. 网络安全和信息化, 2023, (11): 148-150., articleTitle=大数据下的计算机网络安全技术分析, refAbstract=null), Reference(id=1172924327742812442, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2023, volume=null, issue=11, pageStart=148, pageEnd=150, url=null, language=null, rfNumber=[5], rfOrder=7, authorNames=Ren Chenggang, journalName=Network Security and Informatization, refType=null, unstructuredReference=Ren Chenggang. Analysis of computer network security technology under big data[J]. Network Security and Informatization, 2023, (11): 148-150., articleTitle=Analysis of computer network security technology under big data, refAbstract=null), Reference(id=1172924327881224476, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2021, volume=72, issue=null, pageStart=22, pageEnd=47, url=null, language=null, rfNumber=[6], rfOrder=8, authorNames=Zhao J, Jing X, Yan Z, journalName=Information Fusion, refType=null, unstructuredReference=Zhao J, Jing X, Yan Z, et al. Network traffic classification for data fusion: a survey[J]. Information Fusion, 2021, 72: 22-47., articleTitle=Network traffic classification for data fusion: a survey, refAbstract=null), Reference(id=1172924328011247904, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2023, volume=23, issue=18, pageStart=7607, pageEnd=7621, url=null, language=null, rfNumber=[7], rfOrder=9, authorNames=杨宇, 闫钰, 申芳, journalName=科学技术与工程, refType=null, unstructuredReference=杨宇, 闫钰, 申芳, 等. 基于机器和深度学习的入侵检测综述[J]. 科学技术与工程, 2023, 23(18): 7607-7621., articleTitle=基于机器和深度学习的入侵检测综述, refAbstract=null), Reference(id=1172924328124494116, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2023, volume=23, issue=18, pageStart=7607, pageEnd=7621, url=null, language=null, rfNumber=[7], rfOrder=10, authorNames=Yang Yu, Yan Yu, Shen Fang, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Yang Yu, Yan Yu, Shen Fang, et al. A review of intrusion detection based on machine and deep learning[J]. Science Technology and Engineering, 2023, 23(18): 7607-7621., articleTitle=A review of intrusion detection based on machine and deep learning, refAbstract=null), Reference(id=1172924328254517543, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2021, volume=8, issue=1, pageStart=422, pageEnd=439, url=null, language=null, rfNumber=[8], rfOrder=11, authorNames=Shahraki A, Abbasi M, Taherkordi A, journalName=IEEE Transactions on Cognitive Communications and Networking, refType=null, unstructuredReference=Shahraki A, Abbasi M, Taherkordi A, et al. Active learning for network traffic classification: a technical study[J]. IEEE Transactions on Cognitive Communications and Networking, 2021, 8(1): 422-439., articleTitle=Active learning for network traffic classification: a technical study, refAbstract=null), Reference(id=1172924328413901099, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2005, volume=null, issue=null, pageStart=50, pageEnd=60, url=null, language=null, rfNumber=[9], rfOrder=12, authorNames=Moore A W, Zuev D, journalName=Proceedings of the 2005 ACM Sigmetrics International Conference on Measurement and Modeling of Computer Systems, refType=null, unstructuredReference=Moore A W, Zuev D. Internet traffic classification using Bayesian analysis techniques[C]// Proceedings of the 2005 ACM Sigmetrics International Conference on Measurement and Modeling of Computer Systems. New York: ACM, 2005: 50-60., articleTitle=Internet traffic classification using Bayesian analysis techniques, refAbstract=null), Reference(id=1172924328518758704, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2006, volume=null, issue=null, pageStart=179, pageEnd=188, url=null, language=null, rfNumber=[10], rfOrder=13, authorNames=Madhukar A, Williamson C, journalName=14th IEEE International Symposium on Modeling, Analysis, and Simulation, refType=null, unstructuredReference=Madhukar A, Williamson C. A longitudinal study of P2P traffic classification[C]// 14th IEEE International Symposium on Modeling, Analysis, and Simulation. Monterey: IEEE, 2006: 179-188., articleTitle=A longitudinal study of P2P traffic classification, refAbstract=null), Reference(id=1172924328636199218, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2009, volume=39, issue=null, pageStart=12, pageEnd=18, url=null, language=null, rfNumber=[11], rfOrder=14, authorNames=Gringoli F, Salgarelli L, Dusi M, journalName=Computer Communication Review, refType=null, unstructuredReference=Gringoli F, Salgarelli L, Dusi M, et al. GT: picking up the truth from the ground for internet traffic[J]. Computer Communication Review, 2009, 39: 12-18., articleTitle=GT: picking up the truth from the ground for internet traffic, refAbstract=null), Reference(id=1172924328732668213, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2017, volume=17, issue=28, pageStart=103, pageEnd=107, url=null, language=null, rfNumber=[12], rfOrder=15, authorNames=罗冬梅, journalName=科学技术与工程, refType=null, unstructuredReference=罗冬梅. 网络协议流不平衡环境下基于机器学习算法的在线流量分类方法[J]. 科学技术与工程, 2017, 17(28): 103-107., articleTitle=网络协议流不平衡环境下基于机器学习算法的在线流量分类方法, refAbstract=null), Reference(id=1172924328816554295, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2017, volume=17, issue=28, pageStart=103, pageEnd=107, url=null, language=null, rfNumber=[12], rfOrder=16, authorNames=Luo Dongmei, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Luo Dongmei. Online traffic classification method based on machine learning algorithm in network protocol flow imbalance environment[J]. Science Technology and Engineering, 2017, 17(28): 103-107., articleTitle=Online traffic classification method based on machine learning algorithm in network protocol flow imbalance environment, refAbstract=null), Reference(id=1172924328921411897, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2024, volume=null, issue=2, pageStart=104, pageEnd=106, url=null, language=null, rfNumber=[13], rfOrder=17, authorNames=王瑞敏, 汤云凯, 郝高鑫, journalName=山西电子技术, refType=null, unstructuredReference=王瑞敏, 汤云凯, 郝高鑫, 等. 基于机器学习的网络流量分类研究[J]. 山西电子技术, 2024(2): 104-106, 119., articleTitle=基于机器学习的网络流量分类研究, refAbstract=null), Reference(id=1172924329017880889, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2024, volume=null, issue=2, pageStart=104, pageEnd=106, url=null, language=null, rfNumber=[13], rfOrder=18, authorNames=Wang Ruimin, Tang Yunkai, Hao Gaoxin, journalName=Shanxi Electronic Technology, refType=null, unstructuredReference=Wang Ruimin, Tang Yunkai, Hao Gaoxin, et al. Research on network traffic classification based on machine learning[J]. Shanxi Electronic Technology, 2024(2): 104-106, 119., articleTitle=Research on network traffic classification based on machine learning, refAbstract=null), Reference(id=1172924329122738490, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2023, volume=24, issue=4, pageStart=447, pageEnd=453, url=null, language=null, rfNumber=[14], rfOrder=19, authorNames=于治平, 刘彩霞, 刘树新, journalName=信息工程大学学报, refType=null, unstructuredReference=于治平, 刘彩霞, 刘树新, 等. 基于机器学习的网络流量分类综述[J]. 信息工程大学学报, 2023, 24(4): 447-453, 483., articleTitle=基于机器学习的网络流量分类综述, refAbstract=null), Reference(id=1172924329215013180, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2023, volume=24, issue=4, pageStart=447, pageEnd=453, url=null, language=null, rfNumber=[14], rfOrder=20, authorNames=Yu Zhiping, Liu Caixia, Liu Shuxin, journalName=Journal of Information Engineering University, refType=null, unstructuredReference=Yu Zhiping, Liu Caixia, Liu Shuxin, et al. A review of network traffic classification based on machine learning[J]. Journal of Information Engineering University, 2023, 24(4): 447-453, 483., articleTitle=A review of network traffic classification based on machine learning, refAbstract=null), Reference(id=1172924329269539134, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2021, volume=null, issue=null, pageStart=1, pageEnd=6, url=null, language=null, rfNumber=[15], rfOrder=21, authorNames=Perna G, journalName=2021 IEEE 18th Annual Consumer Communications & Networking Conference (CCNC), refType=null, unstructuredReference=Perna G. Online classification of RTC traffic[C]// 2021 IEEE 18th Annual Consumer Communications & Networking Conference (CCNC). Las Vegas: IEEE, 2021: 1-6., articleTitle=Online classification of RTC traffic, refAbstract=null), Reference(id=1172924329340842304, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2022, volume=22, issue=33, pageStart=14786, pageEnd=14794, url=null, language=null, rfNumber=[16], rfOrder=22, authorNames=张超群, 易云恒, 周文娟, journalName=科学技术与工程, refType=null, unstructuredReference=张超群, 易云恒, 周文娟, 等. 基于深度学习与数据增强技术的小样本岩石分类[J]. 科学技术与工程, 2022, 22(33): 14786-14794., articleTitle=基于深度学习与数据增强技术的小样本岩石分类, refAbstract=null), Reference(id=1172924329403756866, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2022, volume=22, issue=33, pageStart=14786, pageEnd=14794, url=null, language=null, rfNumber=[16], rfOrder=23, authorNames=Zhang Chaoqun, Yi Yunheng, Zhou Wenjuan, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Zhang Chaoqun, Yi Yunheng, Zhou Wenjuan, et al. Small-sample rock classification based on deep learning and data enhancement techniques[J]. Science Technology and Engineering, 2022, 22(33): 14786-14794., articleTitle=Small-sample rock classification based on deep learning and data enhancement techniques, refAbstract=null), Reference(id=1172924329588306244, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2022, volume=152, issue=null, pageStart=267, pageEnd=275, url=null, language=null, rfNumber=[17], rfOrder=24, authorNames=Matsuo Y, Lecun Y, Sahani M, journalName=Neural Networks, refType=null, unstructuredReference=Matsuo Y, Lecun Y, Sahani M, et al. Deep learning, reinforcement learning, and world models[J]. Neural Networks, 2022, 152: 267-275., articleTitle=Deep learning, reinforcement learning, and world models, refAbstract=null), Reference(id=1172924329651220806, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2021, volume=37, issue=3, pageStart=105, pageEnd=113, url=null, language=null, rfNumber=[18], rfOrder=25, authorNames=顾玥, 李丹, 高凯辉, journalName=电信科学, refType=null, unstructuredReference=顾玥, 李丹, 高凯辉. 基于机器学习和深度学习的网络流量分类研究[J]. 电信科学, 2021, 37(3): 105-113., articleTitle=基于机器学习和深度学习的网络流量分类研究, refAbstract=null), Reference(id=1172924329730912586, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2021, volume=37, issue=3, pageStart=105, pageEnd=113, url=null, language=null, rfNumber=[18], rfOrder=26, authorNames=Gu Yue, Li Dan, Gao Kaihui, journalName=Telecommunications Science, refType=null, unstructuredReference=Gu Yue, Li Dan, Gao Kaihui. Research on network traffic classification based on machine learning and deep learning[J]. Telecommunications Science, 2021, 37(3): 105-113., articleTitle=Research on network traffic classification based on machine learning and deep learning, refAbstract=null), Reference(id=1172924329798021455, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2021, volume=33, issue=2, pageStart=49, pageEnd=51, url=null, language=null, rfNumber=[19], rfOrder=27, authorNames=宋继红, 王桐桐, journalName=信息与电脑(理论版), refType=null, unstructuredReference=宋继红, 王桐桐. 基于特征融合卷积神经网络的加密流量分类方法[J]. 信息与电脑(理论版), 2021, 33(2): 49-51., articleTitle=基于特征融合卷积神经网络的加密流量分类方法, refAbstract=null), Reference(id=1172924329911267667, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2021, volume=33, issue=2, pageStart=49, pageEnd=51, url=null, language=null, rfNumber=[19], rfOrder=28, authorNames=Song Jihong, Wang Tongtong, journalName=Information and Computer Science (Theoretical Edition), refType=null, unstructuredReference=Song Jihong, Wang Tongtong. An encrypted traffic classification method based on feature fusion convolutional neural network[J]. Information and Computer Science (Theoretical Edition), 2021, 33(2): 49-51., articleTitle=An encrypted traffic classification method based on feature fusion convolutional neural network, refAbstract=null), Reference(id=1172924330003542360, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2020, volume=56, issue=3, pageStart=94, pageEnd=99, url=null, language=null, rfNumber=[20], rfOrder=29, authorNames=李道全, 王雪, 于波, journalName=计算机工程与应用, refType=null, unstructuredReference=李道全, 王雪, 于波, 等. 基于一维卷积神经网络的网络流量分类方法[J]. 计算机工程与应用, 2020, 56(3): 94-99., articleTitle=基于一维卷积神经网络的网络流量分类方法, refAbstract=null), Reference(id=1172924330150343005, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2020, volume=56, issue=3, pageStart=94, pageEnd=99, url=null, language=null, rfNumber=[20], rfOrder=30, authorNames=Li Daoquan, Wang Xue, Yu Bo, journalName=Computer Engineering and Applications, refType=null, unstructuredReference=Li Daoquan, Wang Xue, Yu Bo, et al. A network traffic classification method based on one-dimensional convolutional neural network[J]. Computer Engineering and Applications, 2020, 56(3): 94-99., articleTitle=A network traffic classification method based on one-dimensional convolutional neural network, refAbstract=null), Reference(id=1172924330301337951, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2018, volume=39, issue=1, pageStart=14, pageEnd=23, url=null, language=null, rfNumber=[21], rfOrder=31, authorNames=王勇, 周慧怡, 俸皓, journalName=通信学报, refType=null, unstructuredReference=王勇, 周慧怡, 俸皓, 等. 基于深度卷积神经网络的网络流量分类方法[J]. 通信学报, 2018, 39(1): 14-23., articleTitle=基于深度卷积神经网络的网络流量分类方法, refAbstract=null), Reference(id=1172924330360058210, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2018, volume=39, issue=1, pageStart=14, pageEnd=23, url=null, language=null, rfNumber=[21], rfOrder=32, authorNames=Wang Y, Zhou Huiyi, Feng Hao, journalName=Journal on Communications, refType=null, unstructuredReference=Wang Y, Zhou Huiyi, Feng Hao, et al. Network traffic classification method based on deep convolutional neural network[J]. Journal on Communications, 2018, 39(1): 14-23., articleTitle=Network traffic classification method based on deep convolutional neural network, refAbstract=null), Reference(id=1172924330422972775, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2021, volume=41, issue=3, pageStart=100, pageEnd=108, url=null, language=null, rfNumber=[22], rfOrder=33, authorNames=孔镇, 董育宁, journalName=南京邮电大学学报(自然科学版), refType=null, unstructuredReference=孔镇, 董育宁. 一种基于深度学习的网络流量细粒度分类方法[J]. 南京邮电大学学报(自然科学版), 2021, 41(3): 100-108., articleTitle=一种基于深度学习的网络流量细粒度分类方法, refAbstract=null), Reference(id=1172924330511053161, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2021, volume=41, issue=3, pageStart=100, pageEnd=108, url=null, language=null, rfNumber=[22], rfOrder=34, authorNames=Kong Zhen, Dong Yuning, journalName=Journal of Nanjing University of Posts and Telecommunications (Natural Science Edition), refType=null, unstructuredReference=Kong Zhen, Dong Yuning. A deep learning-based approach for fine-grained classification of network traffic[J]. Journal of Nanjing University of Posts and Telecommunications (Natural Science Edition), 2021, 41(3): 100-108., articleTitle=A deep learning-based approach for fine-grained classification of network traffic, refAbstract=null), Reference(id=1172924330594939244, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2022, volume=42, issue=3, pageStart=82, pageEnd=89, url=null, language=null, rfNumber=[23], rfOrder=35, authorNames=于帅, 董育宁, 邱晓晖, journalName=南京邮电大学学报(自然科学版), refType=null, unstructuredReference=于帅, 董育宁, 邱晓晖. 一种基于深度特征融合的网络流量分类方法[J]. 南京邮电大学学报(自然科学版), 2022, 42(3): 82-89., articleTitle=一种基于深度特征融合的网络流量分类方法, refAbstract=null), Reference(id=1172924330670436718, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2022, volume=42, issue=3, pageStart=82, pageEnd=89, url=null, language=null, rfNumber=[23], rfOrder=36, authorNames=Yu Shuai, Dong Yuning, Qiu Xiaohui, journalName=Journal of Nanjing University of Posts and Telecommunications (Natural Science Edition), refType=null, unstructuredReference=Yu Shuai, Dong Yuning, Qiu Xiaohui. A network traffic classification method based on deep feature fusion[J]. Journal of Nanjing University of Posts and Telecommunications (Natural Science Edition), 2022, 42(3): 82-89., articleTitle=A network traffic classification method based on deep feature fusion, refAbstract=null), Reference(id=1172924330729156978, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2023, volume=50, issue=S2, pageStart=775, pageEnd=781, url=null, language=null, rfNumber=[24], rfOrder=37, authorNames=马继烨, 朱国胜, 卫操, journalName=计算机科学, refType=null, unstructuredReference=马继烨, 朱国胜, 卫操, 等. 一种噪声容忍的网络流量分类方法[J]. 计算机科学, 2023, 50(S2): 775-781., articleTitle=一种噪声容忍的网络流量分类方法, refAbstract=null), Reference(id=1172924330804654453, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2023, volume=50, issue=S2, pageStart=775, pageEnd=781, url=null, language=null, rfNumber=[24], rfOrder=38, authorNames=Ma Jiye, Zhu Guosheng, Wei Cao, journalName=Computer Science, refType=null, unstructuredReference=Ma Jiye, Zhu Guosheng, Wei Cao, et al. A noise-tolerant network traffic classification method[J]. Computer Science, 2023, 50(S2): 775-781., articleTitle=A noise-tolerant network traffic classification method, refAbstract=null), Reference(id=1172924330901123448, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2023, volume=23, issue=30, pageStart=13014, pageEnd=13022, url=null, language=null, rfNumber=[25], rfOrder=39, authorNames=何迎利, 胡光宇, 张浩, journalName=科学技术与工程, refType=null, unstructuredReference=何迎利, 胡光宇, 张浩, 等. 基于局部信息增强注意力机制的网络流量预测[J]. 科学技术与工程, 2023, 23(30): 13014-13022., articleTitle=基于局部信息增强注意力机制的网络流量预测, refAbstract=null), Reference(id=1172924331035341181, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2023, volume=23, issue=30, pageStart=13014, pageEnd=13022, url=null, language=null, rfNumber=[25], rfOrder=40, authorNames=He Yingli, Hu Guangyu, Zhang Hao, journalName=Science Technology and Engineering, refType=null, unstructuredReference=He Yingli, Hu Guangyu, Zhang Hao, et al. Network traffic prediction based on local information-enhanced attention mechanism[J]. Science Technology and Engineering, 2023, 23(30): 13014-13022., articleTitle=Network traffic prediction based on local information-enhanced attention mechanism, refAbstract=null), Reference(id=1172924331232473472, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2023, volume=23, issue=32, pageStart=13896, pageEnd=13902, url=null, language=null, rfNumber=[26], rfOrder=41, authorNames=李艳辉, 王衍萌, journalName=科学技术与工程, refType=null, unstructuredReference=李艳辉, 王衍萌. 基于数据增强技术与CNN-BiLSTM-Attention的油田注水流量预测及效果[J]. 科学技术与工程, 2023, 23(32): 13896-13902., articleTitle=基于数据增强技术与CNN-BiLSTM-Attention的油田注水流量预测及效果, refAbstract=null), Reference(id=1172924331354108291, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2023, volume=23, issue=32, pageStart=13896, pageEnd=13902, url=null, language=null, rfNumber=[26], rfOrder=42, authorNames=Li Yanhui, Wang Yanmeng, journalName=Science Technology and Engineering, refType=null, unstructuredReference=Li Yanhui, Wang Yanmeng. Oilfield water injection flow prediction and effect based on data enhancement technique with CNN-BiLSTM-Attention[J]. Science Technology and Engineering, 2023, 23(32): 13896-13902., articleTitle=Oilfield water injection flow prediction and effect based on data enhancement technique with CNN-BiLSTM-Attention, refAbstract=null), Reference(id=1172924331429605768, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2023, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[27], rfOrder=43, authorNames=金麒旻, journalName=生成对抗网络的质量提升技术研究, refType=null, unstructuredReference=金麒旻. 生成对抗网络的质量提升技术研究[D]. 北京: 北京邮电大学, 2023., articleTitle=null, refAbstract=null), Reference(id=1172924331500908939, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, doi=null, pmid=null, pmcid=null, year=2023, volume=null, issue=null, pageStart=null, pageEnd=null, url=null, language=null, rfNumber=[27], rfOrder=44, authorNames=Jin Qimin, journalName=Research on quality improvement technology of generative adversarial network, refType=null, unstructuredReference=Jin Qimin. Research on quality improvement technology of generative adversarial network[D]. Beijing: Beijing University of Posts and Telecommunications, 2023., articleTitle=null, refAbstract=null)], funds=[Fund(id=1172924326715207940, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, awardId=4041422007, language=CN, fundingSource=山东理工大学科技博士项目(4041422007), fundOrder=null, country=null)], companyList=[AuthorCompany(id=1172924319958184093, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, xref=null, ext=[AuthorCompanyExt(id=1172924319962378398, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, companyId=1172924319958184093, language=EN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=School of Computer Science and Technology, Shandong University of Technology, Zibo 255049, China), AuthorCompanyExt(id=1172924319966572703, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, companyId=1172924319958184093, language=CN, country=null, province=null, city=null, postcode=null, companyName=null, departmentName=null, remark=山东理工大学计算机科学与技术学院, 淄博 255049)])], figs=[ArticleFig(id=1172924322269245632, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Fig.1, caption=Flowchart of the methodology in this paper, figureFileSmall=N4fy2SmnWYaXyA/ULcyoXg==, figureFileBig=5enooCnHHM7Bh5tbfSZSaw==, tableContent=null), ArticleFig(id=1172924322340548801, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=图1, caption=本文方法流程图, figureFileSmall=N4fy2SmnWYaXyA/ULcyoXg==, figureFileBig=5enooCnHHM7Bh5tbfSZSaw==, tableContent=null), ArticleFig(id=1172924322424434882, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Fig.2, caption=Data format with application type WWW, figureFileSmall=W4VUPbMs6gNfwC0FQ5d6/w==, figureFileBig=tYrpaGZixx2PfClMzOdP4w==, tableContent=null), ArticleFig(id=1172924322474766531, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=图2, caption=应用类型为WWW的数据格式, figureFileSmall=W4VUPbMs6gNfwC0FQ5d6/w==, figureFileBig=tYrpaGZixx2PfClMzOdP4w==, tableContent=null), ArticleFig(id=1172924322525098180, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Fig.3, caption=Grayscale graph of Moore traffic data corresponding to the category, figureFileSmall=xZZTk22vohJBLVCT8TtjlQ==, figureFileBig=iLfS03p652uinyWsm14baw==, tableContent=null), ArticleFig(id=1172924322592207045, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=图3, caption=Moore流量数据对应类别的灰度图, figureFileSmall=xZZTk22vohJBLVCT8TtjlQ==, figureFileBig=iLfS03p652uinyWsm14baw==, tableContent=null), ArticleFig(id=1172924322667704518, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Fig.4, caption=Grayscale graph of ISCX non-VPN traffic data corresponding to the category, figureFileSmall=FmRdQ3JgJSd/7kkXWXri9Q==, figureFileBig=9LZ7NhM1TEugz9RhkSELTQ==, tableContent=null), ArticleFig(id=1172924322722230471, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=图4, caption=ISCX non-VPN流量数据对应类别的灰度图, figureFileSmall=FmRdQ3JgJSd/7kkXWXri9Q==, figureFileBig=9LZ7NhM1TEugz9RhkSELTQ==, tableContent=null), ArticleFig(id=1172924322810310856, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Fig.5, caption=Data augmentation generates images, figureFileSmall=SqTj6Y5uoD85BTqTzFGgyQ==, figureFileBig=nZ1vz0AATx4Hb2DStx7Tbw==, tableContent=null), ArticleFig(id=1172924322885808329, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=图5, caption=数据增强生成图像, figureFileSmall=SqTj6Y5uoD85BTqTzFGgyQ==, figureFileBig=nZ1vz0AATx4Hb2DStx7Tbw==, tableContent=null), ArticleFig(id=1172924322948722890, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Fig.6, caption=Five-fold cross-validation graph, figureFileSmall=dz4h1SlqAvIK5G1pdOe71Q==, figureFileBig=CCOfAxGbS2lcdHj4lA3JQQ==, tableContent=null), ArticleFig(id=1172924323066163403, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=图6, caption=五折交叉验证方式图, figureFileSmall=dz4h1SlqAvIK5G1pdOe71Q==, figureFileBig=CCOfAxGbS2lcdHj4lA3JQQ==, tableContent=null), ArticleFig(id=1172924323137466572, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Fig.7, caption=The architecture of the CNN-optuna-attention model, figureFileSmall=+IQTVE+rAIokDY+Zp6bYmQ==, figureFileBig=qODO8hFFMIA+Cd0wnzL24A==, tableContent=null), ArticleFig(id=1172924323225546957, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=图7, caption=CNN-Optuna-Attention模型架构图, figureFileSmall=+IQTVE+rAIokDY+Zp6bYmQ==, figureFileBig=qODO8hFFMIA+Cd0wnzL24A==, tableContent=null), ArticleFig(id=1172924323292655822, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Fig.8, caption=Comparison chart of the categorical precision of different methods in the Moore dataset, figureFileSmall=AwRJMoPkJdl61/wfYSg7kw==, figureFileBig=dUE7+3rgKxEVlWFghOhFaA==, tableContent=null), ArticleFig(id=1172924323401707727, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=图8, caption=Moore数据集下不同方法的类别精确率对比图表, figureFileSmall=AwRJMoPkJdl61/wfYSg7kw==, figureFileBig=dUE7+3rgKxEVlWFghOhFaA==, tableContent=null), ArticleFig(id=1172924323573674192, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Fig.9, caption=Comparison chart of F1 scores of different methods in the Moore dataset, figureFileSmall=7JlMcZoGy/IgPC15GxGoHw==, figureFileBig=kjnNNBR/FpMwjW7kqIraSg==, tableContent=null), ArticleFig(id=1172924323636588753, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=图9, caption=Moore数据集下不同方法的类别F1分数对比图表, figureFileSmall=7JlMcZoGy/IgPC15GxGoHw==, figureFileBig=kjnNNBR/FpMwjW7kqIraSg==, tableContent=null), ArticleFig(id=1172924323770806482, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Fig.10, caption=Comparison chart of the categorical precision of different methods in the ISCX non-VPN dataset, figureFileSmall=8cldtDU/LG4OqaWo+ha75A==, figureFileBig=77X54pcYI0TEYtNKsFh8BQ==, tableContent=null), ArticleFig(id=1172924323993104595, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=图10, caption=ISCX non-VPN数据集下不同方法的类别精确率对比图表, figureFileSmall=8cldtDU/LG4OqaWo+ha75A==, figureFileBig=77X54pcYI0TEYtNKsFh8BQ==, tableContent=null), ArticleFig(id=1172924324097962196, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Fig.11, caption=Comparison chart of F1 scores of different methods in the ISCX non-VPN dataset, figureFileSmall=9hVjhY/6LaFAo478AANsEQ==, figureFileBig=JkmXBgzfIKgkP3jHE7pgRg==, tableContent=null), ArticleFig(id=1172924324248957141, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=图11, caption=ISCX non-VPN数据集下不同方法的类别F1分数对比图表, figureFileSmall=9hVjhY/6LaFAo478AANsEQ==, figureFileBig=JkmXBgzfIKgkP3jHE7pgRg==, tableContent=null), ArticleFig(id=1172924324408340694, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Fig.12, caption=The confusion matrix of the proposed method, figureFileSmall=Fnx95sP70abwsVdHmeEOSQ==, figureFileBig=Nc965iKhFGfvXn8aISVi6g==, tableContent=null), ArticleFig(id=1172924324488032471, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=图12, caption=本文方法的混淆矩阵, figureFileSmall=Fnx95sP70abwsVdHmeEOSQ==, figureFileBig=Nc965iKhFGfvXn8aISVi6g==, tableContent=null), ArticleFig(id=1172924324609667289, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Table 1, caption=

Statistics of Moore traffic data

, figureFileSmall=null, figureFileBig=null, tableContent=
类别 数量 应用来源
WWW 328 092 Web browsers,web applications
MAIL 28 567 IMAP,POP,SMTP
FTP-CONTROL 3 054 FTP
FTP-PASV 2 688 FTP
ATTACK 1 793 Port scans,worms,viruses,sql injections
P2P 2 094 Napster,WKazaa,Gnutella,eDonkey,Bit Torrent
DATABASE 2 648 MySQL,dbase,Oracle
FTP-DATA 5 797 FTP
MULTIMEDIA 576 Windows Media Player,Real,iTunes
SERVICES 2 099 X11,DNS,IDENT,LDAP,NTP
INTERACTIVE 110 SSH,TELNET,VNC,GotoMyPC
GAMES 8 Half-Life
总计 377 526
), ArticleFig(id=1172924324710330588, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=表1, caption=

Moore流量数据统计信息

, figureFileSmall=null, figureFileBig=null, tableContent=
类别 数量 应用来源
WWW 328 092 Web browsers,web applications
MAIL 28 567 IMAP,POP,SMTP
FTP-CONTROL 3 054 FTP
FTP-PASV 2 688 FTP
ATTACK 1 793 Port scans,worms,viruses,sql injections
P2P 2 094 Napster,WKazaa,Gnutella,eDonkey,Bit Torrent
DATABASE 2 648 MySQL,dbase,Oracle
FTP-DATA 5 797 FTP
MULTIMEDIA 576 Windows Media Player,Real,iTunes
SERVICES 2 099 X11,DNS,IDENT,LDAP,NTP
INTERACTIVE 110 SSH,TELNET,VNC,GotoMyPC
GAMES 8 Half-Life
总计 377 526
), ArticleFig(id=1172924324810993886, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Table 2, caption=

Statistics of ISCX non-VPN traffic data

, figureFileSmall=null, figureFileBig=null, tableContent=
类别 数量 应用来源
语音 500 AIM,ICQ,Facebook,Skype,Hangout,Google
聊天 500 Skype,Hangout,Facebook
视频 500 Skype,Hangout,Facebook
IP语音 500 Voipbuster
总计 2 000
), ArticleFig(id=1172924324920045793, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=表2, caption=

ISCX non-VPN 流量数据统计信息

, figureFileSmall=null, figureFileBig=null, tableContent=
类别 数量 应用来源
语音 500 AIM,ICQ,Facebook,Skype,Hangout,Google
聊天 500 Skype,Hangout,Facebook
视频 500 Skype,Hangout,Facebook
IP语音 500 Voipbuster
总计 2 000
), ArticleFig(id=1172924325012320483, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Table 3, caption=

Parameter settings for ImageDataGenerator

, figureFileSmall=null, figureFileBig=null, tableContent=
参数 设置值 具体描述
rotation_range 20 随机旋转的角度范围,以度为单位,设置为-20°~+20°
width_shift_range 0.1 随机水平平移的宽度范围,设置为10%
height_shift_range 0.1 随机垂直平移的高度范围,设置为10%
horizontal_flip True 是否随机水平翻转,设置为True表示会进行水平翻转
vertical_flip True 是否随机垂直翻转,设置为True表示会进行垂直翻转
shear_range 0.2 随机剪切变换的角度范围,设置为-0.2~+0.2 rad
zoom_range 0.2 随机缩放的范围,设置为原始尺寸的80%~120%
fill_mode “nearest” 当进行变换时,超出图像边界的像素填充方式,设置为“nearest”表示最近邻插值
), ArticleFig(id=1172924325108789477, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=表3, caption=

ImageDataGenerator参数设置

, figureFileSmall=null, figureFileBig=null, tableContent=
参数 设置值 具体描述
rotation_range 20 随机旋转的角度范围,以度为单位,设置为-20°~+20°
width_shift_range 0.1 随机水平平移的宽度范围,设置为10%
height_shift_range 0.1 随机垂直平移的高度范围,设置为10%
horizontal_flip True 是否随机水平翻转,设置为True表示会进行水平翻转
vertical_flip True 是否随机垂直翻转,设置为True表示会进行垂直翻转
shear_range 0.2 随机剪切变换的角度范围,设置为-0.2~+0.2 rad
zoom_range 0.2 随机缩放的范围,设置为原始尺寸的80%~120%
fill_mode “nearest” 当进行变换时,超出图像边界的像素填充方式,设置为“nearest”表示最近邻插值
), ArticleFig(id=1172924325205258472, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Table 4, caption=

Changes in the sample size of the Moore dataset

, figureFileSmall=null, figureFileBig=null, tableContent=
类别 原样本数量 数据增强后样本数量
INTERACTIVE 110 3 410
GAMES 8 3 208
总计 377 526 384 026
), ArticleFig(id=1172924325322698987, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=表4, caption=

Moore数据集的样本数量变化

, figureFileSmall=null, figureFileBig=null, tableContent=
类别 原样本数量 数据增强后样本数量
INTERACTIVE 110 3 410
GAMES 8 3 208
总计 377 526 384 026
), ArticleFig(id=1172924325444333805, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Table 5, caption=

Comparison between pptuna and grid search methods

, figureFileSmall=null, figureFileBig=null, tableContent=
参数 Optuna方法 网格搜索方法
灵活性
超参数搜索范围 连续、离散参数等自动定义 手动定义参数组合
搜索方式 TPE、随即搜索等 穷举迭代搜索
收敛速度
使用场景 复杂超参数空间 简单超参数空间
适用性 多样化模型优化 简单模型调优
), ArticleFig(id=1172924325528219886, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=表5, caption=

Optuna与网格搜索方法对比

, figureFileSmall=null, figureFileBig=null, tableContent=
参数 Optuna方法 网格搜索方法
灵活性
超参数搜索范围 连续、离散参数等自动定义 手动定义参数组合
搜索方式 TPE、随即搜索等 穷举迭代搜索
收敛速度
使用场景 复杂超参数空间 简单超参数空间
适用性 多样化模型优化 简单模型调优
), ArticleFig(id=1172924325645660399, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Table 6, caption=

Search space for hyperparameters

, figureFileSmall=null, figureFileBig=null, tableContent=
超参数 搜索空间
学习率 0.000 1~0.1
丢弃率 0.0~0.3
批次大小 64/128/256
迭代次数 200(使用早停策略)
), ArticleFig(id=1172924325712769264, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=表6, caption=

超参数的搜索空间

, figureFileSmall=null, figureFileBig=null, tableContent=
超参数 搜索空间
学习率 0.000 1~0.1
丢弃率 0.0~0.3
批次大小 64/128/256
迭代次数 200(使用早停策略)
), ArticleFig(id=1172924325792461041, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Table 7, caption=

The network structure of the CNN-optuna-attention model

, figureFileSmall=null, figureFileBig=null, tableContent=
层类型 输入尺寸 卷积核/
池化大小
卷积核/
神经元数
步长 填充
方式
输出
尺寸
激活
函数
卷积层1 (batch_size,16, 16, 1) (3, 3) 8 1 ‘same' (batch_size,16, 16, 8) ReLU
池化层1 (batch_size,16, 16, 8) (2, 2) 2 ‘same' (batch_size,8, 8, 8)
卷积层2 (batch_size,8, 8, 8) (3, 3) 16 1 ‘same' (batch_size,8, 8, 16) ReLU
池化层2 (batch_size,8, 8, 16) (2, 2) 2 ‘same' (batch_size,4, 4, 16)
扁平化层 (batch_size,4, 4, 16) (batch_size,256)
失活层1 (batch_size,256) (batch_size,256)
全连接层1 (batch_size,256) 256 (batch_size,256) ReLU
失活层2 (batch_size,256) (batch_size,256)
全连接层2 (batch_size,256) 128 (batch_size,128) ReLU
注意力机制层 (batch_size,128) (batch_size,128)
全连接层3(输出层) (batch_size,128) 12/4 (batch_size,12/4) softmax
), ArticleFig(id=1172924325876347122, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=表7, caption=

CNN-Optuna-Attention模型网络参数结构

, figureFileSmall=null, figureFileBig=null, tableContent=
层类型 输入尺寸 卷积核/
池化大小
卷积核/
神经元数
步长 填充
方式
输出
尺寸
激活
函数
卷积层1 (batch_size,16, 16, 1) (3, 3) 8 1 ‘same' (batch_size,16, 16, 8) ReLU
池化层1 (batch_size,16, 16, 8) (2, 2) 2 ‘same' (batch_size,8, 8, 8)
卷积层2 (batch_size,8, 8, 8) (3, 3) 16 1 ‘same' (batch_size,8, 8, 16) ReLU
池化层2 (batch_size,8, 8, 16) (2, 2) 2 ‘same' (batch_size,4, 4, 16)
扁平化层 (batch_size,4, 4, 16) (batch_size,256)
失活层1 (batch_size,256) (batch_size,256)
全连接层1 (batch_size,256) 256 (batch_size,256) ReLU
失活层2 (batch_size,256) (batch_size,256)
全连接层2 (batch_size,256) 128 (batch_size,128) ReLU
注意力机制层 (batch_size,128) (batch_size,128)
全连接层3(输出层) (batch_size,128) 12/4 (batch_size,12/4) softmax
), ArticleFig(id=1172924325947650291, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Table 8, caption=

Confusion matrix for calculating assessment metrics

, figureFileSmall=null, figureFileBig=null, tableContent=
真实值 预测值
Negative Positive
Positive False Negative(FN) True Positive(TP)
Negative True Negative(TN) False Positive(FP)
), ArticleFig(id=1172924326090256628, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=表8, caption=

计算评估指标的混淆矩阵

, figureFileSmall=null, figureFileBig=null, tableContent=
真实值 预测值
Negative Positive
Positive False Negative(FN) True Positive(TP)
Negative True Negative(TN) False Positive(FP)
), ArticleFig(id=1172924326220280056, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Table 9, caption=

The accuracy of different methods under the Moore dataset

, figureFileSmall=null, figureFileBig=null, tableContent=
方法 准确率
BP神经网络 0.993.1
文献[21]的方法 0.993 0
对比方法1 0.992 7
对比方法2 0.992 5
本文方法 0.995 9
), ArticleFig(id=1172924326325137659, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=表9, caption=

Moore数据集下采用不同方法的准确率

, figureFileSmall=null, figureFileBig=null, tableContent=
方法 准确率
BP神经网络 0.993.1
文献[21]的方法 0.993 0
对比方法1 0.992 7
对比方法2 0.992 5
本文方法 0.995 9
), ArticleFig(id=1172924326455161085, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=EN, label=Table 10, caption=

The accuracy of different methods under the ISCX non-VPN dataset

, figureFileSmall=null, figureFileBig=null, tableContent=
方法 准确率
文献[15]的方法 0.977 2
文献[22]的方法 0.985 8
文献[23]的方法 0.996 5
本文方法 0.997 5
), ArticleFig(id=1172924326547435777, tenantId=1146029695717560320, journalId=1146123166801305609, articleId=1149768944640241740, language=CN, label=表10, caption=

ISCX non-VPN数据集下采用不同方法的准确率

, figureFileSmall=null, figureFileBig=null, tableContent=
方法 准确率
文献[15]的方法 0.977 2
文献[22]的方法 0.985 8
文献[23]的方法 0.996 5
本文方法 0.997 5
)], attaches=null, journal=Journal(id=1146119176004939786, delFlag=0, nameCn=科学技术与工程, nameEn=Science Technology and Engineering, nameHistory1=null, nameHistory2=null, issn=1671-1815, eissn=, cn=11-4688/T, coden=null, periodic=4, language=CN, oaType=是, ccby=null, superviseOffice=null, ownerOffice=null, pubOffice=null, editorOffice=null, officeType=null, aims=null, clcCode=null, officeProv=null, officeCity=null, officeAddr=null, officeZip=null, officeEmail=null, officePhone=null, editDirector=null, officeDirector=null, officeDirectorPhone=null, officeStaffNum=null, officeEmpNum=null, coverPicUrl=UKU/O7GSka5polgCTkbIIw==, journalPrice=null, startedYear=null, abbrevIsoEn=Sci Technol Eng, journalRemark=null, publicationField=null, createdTime=null, updatedTime=1754445529766, createdBy=null, updatedBy=13701087609, firstLetterCn=S, firstLetterEn=S, subjectCode=Natural Sciences, subjectName=自然科学, subjectCodeEn=Natural Sciences, subjectNameEn=null, picCn=UKU/O7GSka5polgCTkbIIw==, picEn=5hwlULoNwcbj3xUmVi9MAQ==, jcr=null, cjcr=null, exts=[JournalExt(id=1159791870395564357, language=CN, name=科学技术与工程, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=null, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=http://www.stae.com.cn/jsygc/home, createdTime=1754445529793, updatedTime=1754445529793, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=http://www.stae.com.cn/jsygc/site/menus/20090429150146001, submissionAuthorUrl=http://www.stae.com.cn/jsygc/author/login, submissionEditorUrl=http://www.stae.com.cn/jsygc/editor/login, submissionReviewUrl=http://www.stae.com.cn/jsygc/reviewer/login, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""}), JournalExt(id=1159791870441701702, language=EN, name=Science Technology and Engineering, nameHistory1=null, nameHistory2=null, managedBy=, sponsoredBy=, publishedBy=, editorOffice=, officeProv=null, officeCity=null, officeAddr=, officeZip=, editDirector=null, officeDirector=null, officePhone=null, coverPicUrl=null, journalRemark=, submitArticleUrl=null, websiteUrl=http://www.stae.com.cn/jsygc/home, createdTime=1754445529804, updatedTime=1754445529804, createdBy=13701087609, updatedBy=13701087609, submissionGuidelinesUrl=, submissionAuthorUrl=http://www.stae.com.cn/jsygc/author/login, submissionEditorUrl=http://www.stae.com.cn/jsygc/editor/login, submissionReviewUrl=http://www.stae.com.cn/jsygc/reviewer/login, submissionCeEditorUrl=, submissionAeEditorUrl=, option={"copyright":""})], databaseList=null, tenantJournalId=1146123166801305609, websiteList=[Website(id=1148243202391400884, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1146123166801305609, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/kxjsygc/CN, language=CN, createTime=1751692112777, createBy=18614031015, updateTime=1753520965431, updateBy=18614031015, name=科学技术与工程-中文站点, tplId=1146099689490845704, title=科学技术与工程, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1148622798802673703, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=articleTextType, value=kx, createTime=1751782615614, updateTime=1751782615614, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798781702180, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=banner, value=null, createTime=1751782615609, updateTime=1751782615609, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798769119267, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=logo, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic?fileId=j86gbwi+p0Idkyl5SzIlmQ==, createTime=1751782615606, updateTime=1751782615606, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798794285094, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic, createTime=1751782615612, updateTime=1751782615612, creator=18614031015, updator=18614031015), WebsiteProps(id=1148622798790090789, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1148243202391400884, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1751782615611, updateTime=1751782615611, creator=18614031015, updator=18614031015)]), Website(id=1155914124811976731, webName=null, webTitle=null, webDomain=null, webCopyrigh=null, webIpcNo=null, seoTitle=null, seoKeywords=null, seoDescription=null, tenantJournalId=null, journalId=1146123166801305609, journalNameCn=null, journalNameEn=null, grayFlag=null, tenantId=1146029695717560320, platformId=null, journalGroupId=null, journalGroupNameCn=null, journalGroupNameEn=null, type=1, domain=https://castjournals.cast.org.cn/joweb/kxjsygc/EN, language=EN, createTime=1753521003206, createBy=18614031015, updateTime=1753521003206, updateBy=18614031015, name=科学技术与工程-英文站点, tplId=1146101810881728533, title=Science Technology and Engineering, delFlag=0, indexPage=/home, props=[WebsiteProps(id=1155914371227308235, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=articleTextType, value=kx, createTime=1753521061952, updateTime=1753521061952, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371210531016, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=banner, value=null, createTime=1753521061947, updateTime=1753521061947, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371202142407, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=logo, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic?fileId=j86gbwi+p0Idkyl5SzIlmQ==, createTime=1753521061945, updateTime=1753521061945, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371223113930, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=picServerUrl, value=https://castjournals.cast.org.cn/joweb/kjdb/CN/file/pic, createTime=1753521061950, updateTime=1753521061950, creator=18614031015, updator=18614031015), WebsiteProps(id=1155914371218919625, tenantId=1146029695717560320, journalId=null, journalGroupId=null, siteId=1155914124811976731, code=staticResourcePath, value=https://castjournals.cast.org.cn/joweb/cast_kjdb_cn_619/, createTime=1753521061949, updateTime=1753521061949, creator=18614031015, updator=18614031015)])], journalTitle=科学技术与工程, weixinUrl=null, journalUrl=null, iacademicId=null, status=0, seqNo=null, journalTitleEn=Science Technology and Engineering, journalPhotoCn=UKU/O7GSka5polgCTkbIIw==, journalPhotoEn=5hwlULoNwcbj3xUmVi9MAQ==, journalFirstLetter=S, journalRecommend=null, journalNew=null, journalCollection=null, jcrJf=null, cjcrJf=null, jcrJfStr=null, cjcrJfStr=null, submissionFirstDecision=null, sciSubjectClassification=null, casSubjectClassification=null, citeScore=null, totalCitationFrequency=null, icpCode=null, psCode=null, advertisingLicenseCode=null, copyrightInformation=null, country=null, option=null, provinceCode=null, provinceName=null, collectFlag=false), detailUrlCn=https://castjournals.cast.org.cn/joweb/kxjsygc/CN/10.12404/j.issn.1671-1815.2404568, detailUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/10.12404/j.issn.1671-1815.2404568, pdfUrlCn=https://castjournals.cast.org.cn/joweb/kxjsygc/CN/PDF/10.12404/j.issn.1671-1815.2404568, pdfUrlEn=https://castjournals.cast.org.cn/joweb/kxjsygc/EN/PDF/10.12404/j.issn.1671-1815.2404568, aliStartDate=null, aliEndDate=null, collectionFlag=false, citedCount=null, citedUrl=null, reference=null)
收藏切换
基于数据增强与CNN-Optuna-Attention改进的网络流量分类方法
收藏切换
PDF下载
唐浩耀 , 崔鑫 * , 张艺炜 , 赵庆慧
科学技术与工程 | 论文·自动化技术、计算机技术 2025,25(15): 6419-6430
收起
收藏切换
科学技术与工程 | 论文·自动化技术、计算机技术 2025, 25(15): 6419-6430
基于数据增强与CNN-Optuna-Attention改进的网络流量分类方法
全屏
唐浩耀 , 崔鑫* , 张艺炜, 赵庆慧
作者信息
  • 山东理工大学计算机科学与技术学院, 淄博 255049
  • 唐浩耀(2000—),男,汉族,山东济南人,硕士研究生。研究方向:网络服务与信息安全。E-mail:

通讯作者:

* 崔鑫(1972—),女,汉族,山东淄博人,博士,副教授。研究方向:下一代互联网技术、网络安全、网络大数据、无线传感网。E-mail:
Network Traffic Classification Method Improved Based on Data Augmentation and CNN-Optuna-Attention
Hao-yao TANG , Xin CUI* , Yi-wei ZHANG, Qing-hui ZHAO
Affiliations
  • School of Computer Science and Technology, Shandong University of Technology, Zibo 255049, China
出版时间: 2025-05-28 doi: 10.12404/j.issn.1671-1815.2404568
文章导航
收藏切换

针对基于卷积神经网络的流量分类方法难以捕捉序列中不同部分的重要性、特征提取不足的问题,为提升网络流量分类精度,提出一种注意力机制与卷积神经网络相结合的流量分类方法,在卷积神经网络模型的基础上设计实现一种注意力机制层,通过接收全连接层的输出作为输入,计算输入特征的权重并乘以原始特征,实现对关键特征的加强,进而有助于提高模型对关键信息的捕捉能力。其次针对网络流量类别样本数不均衡导致模型过拟合于高比例类别,难以识别分类小比例类别的问题,提出了一种对数据集进行数据增强的方法。并且考虑到超参数组合优化的角度,提出一种基于贝叶斯优化的超参数搜索策略和五折交叉验证的方式对模型的超参数组合进行优化。通过上述方法研究确定模型的超参数组合。使用公开数据集进行上述实验与模型测试,结果表明:与其他方法相比,总体准确率、精确率以及F1分数都有明显的提升,验证了本文所提方法具有更好的分类性能。

深度学习  /  流量分类  /  注意力机制  /  卷积神经网络  /  超参数优化  /  五折交叉验证

In order to improve the accuracy of network traffic classification, a traffic classification method combining an attention mechanism and a convolutional neural network was proposed. An attention mechanism layer was designed and implemented on the basis of the convolutional neural network model, which received the output of the fully connected layer as input, calculated the weight of the input features, and multiplied it by the original features to strengthen the key features. This, in turn, helped to improve the model's ability to capture key information. Secondly, in order to solve the problem that the model was overfitting to the high-proportion category due to the unbalanced sample number of network traffic categories, and it was difficult to identify the small-proportion categories, a method to augment the dataset was proposed. Considering the perspective of hyperparameter combination optimization, a hyperparameter search strategy based on Bayesian optimization and five-fold cross-validation was proposed to optimize the hyperparameter combination of the model. The combination of hyperparameters of the model was determined by the above methods. The public dataset was used for the above experiments and model tests. The results show that compared with other methods, the overall accuracy, precision, and F1 score are significantly improved, which verifies that the proposed method has better classification performance.

deep learning  /  traffic classification  /  attention mechanisms  /  convolutional neural networks  /  hyperparameter optimization  /  five-fold cross-validation
唐浩耀, 崔鑫, 张艺炜, 赵庆慧. 基于数据增强与CNN-Optuna-Attention改进的网络流量分类方法. 科学技术与工程, 2025 , 25 (15) : 6419 -6430 . DOI: 10.12404/j.issn.1671-1815.2404568
Hao-yao TANG, Xin CUI, Yi-wei ZHANG, Qing-hui ZHAO. Network Traffic Classification Method Improved Based on Data Augmentation and CNN-Optuna-Attention[J]. Science Technology and Engineering, 2025 , 25 (15) : 6419 -6430 . DOI: 10.12404/j.issn.1671-1815.2404568
据中国互联网络信息中心(China Internet Network Information Center,CNNIC)在北京发布的第53 次报告[1]可知,截至2023 年12 月,中国网民总数突破10.92 亿人。移动互联网流量达3 015 亿GB。这一数据既体现了网络应用的日益丰富与普及,也揭示了网络资源使用紧张、分配不均衡和利用率低等问题[2]
网络流量分类(network traffic classification, NTC)[3]作为现代网络管理与安全不可或缺的一环,对维持网络稳定和保障网络安全具有不可忽视的作用。NTC的核心任务是将互联网流量按照预设的类别进行区分,如协议类型(UDP、TCP等)、应用类型(Youtube、Facebook等)以及异常检测(正常、异常)。它不仅可以被互联网服务提供商(internet service provider, ISP)用于服务质量(quality of service, QoS)[4]和故障排除,同时也广泛用于入侵检测系统,以保护用户设备免受潜在的网络攻击[5]。因此NTC一直是网络安全研究的重要领域[6-7]
早期网络环境较简单,NTC依赖基于端口与有效载荷的方法[8],然而伴随着随机或动态端口号的使用,基于端口的技术准确率下降。Moore等[9]观的研究揭示了一个重要事实,采用IANA(internet assigned numbers authority)列表的基于端口的流量分类技术存在局限性,其准确率不高于70%;Madhukar等[10]的研究发现基于端口的技术无法识别他们所调查的流量比例高达30%~70%。
因此有效载荷的分类技术[8]被研究者们开发并用于流量分类任务,这种新技术也称为深度包检测(deep packet inspection, DPI),DPI技术克服了端口变更带来的影响。然而新应用层出不穷,导致特征库中产生大量DPI技术无法识别的未知流量特征。Gringoli等[11]对DPI方法的性能进行了评估,实验结果显示,在UNIBS数据集上,L7-filter等DPI工具的正确分类率为67.73%,而在POLITO数据集上,其正确分类率则仅为58.79%,这凸显了DPI技术面对未知流量特征时的问题。
为了解决DPI方法的问题,研究人员纷纷尝试将机器学习(machine learning, ML)算法应用于NTC领域[12-14]:如决策树[15]、随机森林、支持向量机等。罗冬梅[12]提出一种基于机器学习算法的流量分类方法,针对不同类别的在线流量样本流集合筛取出若干最近邻样本流,并计算每个样本流特征权重来确定各个特征与类别之间的相关性,并将相关性大的特征视为在线流量特征。根据得到的特征选取部分标识在线流量数据,以确定K中值聚类的起始中心来构造映射关系,从而获取未知的在线流量种类。实验结果表明该方法有很高的分类精度,且扩展性和适应能力较强。然而,随着互联网流量的激增和新型的网络流量不断涌现,仅仅依赖人工分析来提取网络流量特征已经无法满足需求。
为了克服传统方法和ML方法在分类任务上的的局限性,深度学习(deep learning, DL)[16]应运而生。DL作为ML的前沿分支,其优势在于能够自动从大规模数据中提取特征[17]。这种端到端NTC的架构大幅减少了人工参与特征提取的繁琐过程,并且凭借其出色的特征识别能力能够揭示人类分析难以察觉的复杂特征,从而实现更一致和准确的分类效果[18]。宋继红等[19]通过改进现有模型,引入Inception模块,利用不同尺寸和数量的卷积核来提取和重构特征,显著提升了流量分类的精度;李道全等[20]提出一种基于一维卷积神经网络(convolutional neural networks, CNN)的模型,其在F1、召回率等多个指标上均优于经典ML方法;王勇等[21]提出了一种将数据进行归一化处理后转化为灰度图片作为CNN的输入,然后基于LeNet-5深度CNN构建了一个模型,该模型能够自主学习流量特征,从而实现对网络流量的高效分类的方法;孔镇等[22]则设计了一种新的流量特征扩展方式,通过将详细的特征信息转化为图像格式,并采用CNN对代表流数据的图像进行细粒度分类;于帅等[23]另辟蹊径,提出了一种基于深度特征融合的流量分类策略。通过对原始的统计特征进行融合处理,随后应用CNN对转换后的灰度图像进行分类;马继烨等[24]通过数据归一化、数据增强和标签加噪等预处理步骤,结合Res2Net深度残差神经网络,构建了一种能够容忍噪声的深度神经网络模型,有效提升了流量分类的准确性和鲁棒性。何迎利等[25]针对网络流量具有强烈的非线性和不确定性导致其预测精度不高的问题,在传统时序序列预测模型的基础上设计实现了一种局部上下文信息增强的注意力机制,这种机制能够从微观层面深入解析时间序列的内在结构,显著提升了预测模型对局部变化的感知能力,进而提升网络流量预测精度。
然而使用CNN进行流量分类时仍然会存在众多问题,如数据集样本数不足或类别比例不均衡、分类模型构建不合适、超参数选取欠佳、数据预处理不当、训练不充分或过度训练等因素都会影响分类任务的效果。针对上述问题提出改进方法,以期改善NTC任务的效果。
针对上述使用CNN进行NTC任务存在的问题,在现有工作的基础上,提出一种新的基于数据增强与CNN-Optuna-Attention改进进行NTC的方法,该方法主要贡献如下。
(1)针对单一CNN模型分类效果不佳等问题,本文提出一种在现有CNN模型[21]的基础上加以改进的方法,将CNN与注意力机制(attention mechanism, AM)相结合,通过接收全连接层的输出作为输入,计算输入特征的权重并乘以原始特征,实现对关键特征的加强,有助于提高模型对关键信息的捕捉能力,从而提高网络流量分类的准确性。
(2)针对网络流量数据的特征数无法准确转换为灰度图像的问题,提出一种对特征进行无关字段替换和均值填充的改进方式。根据流量特征构建矩阵时采取将每行数据除缺失值和布尔型值外的特征进行均值计算,并在末尾进行八次均值填充以适应CNN的输入特性的方式进行改进。
(3)针对现有数据集中网络流量类别比例差距过大或不均衡可能导致模型过拟合于高比例类别,以及小比例类别数据难以识别分类的问题,通过对流量数据进行旋转、缩放、镜像[24]等操作从而对小比例类别流量数据进行数据增强。该方式可以使各类别数据比例差距缩小。
(4)针对模型超参数组合的选择不佳导致分类准确率不高的问题,提出一种五折交叉验证划分数据集,并且基于贝叶斯优化进行超参数组合优化的方法[26],该方法可以更充分地对模型的未知超
参数组合的性能进行评估,从而比较不同模型参数下的评价结果,进行模型超参数优化,确定最终的模型。
本文方法流程图如图1所示。
实验采用两个真实数据集。Moore数据集[9]是由专业用户采用抽样算法得到的377 526 个网络数据样本,并根据应用类型分为12 种类别。例如应用类型为WWW的具体数据格式如图2所示。Moore流量数据集统计信息如表1所示。为了能够更好地验证本文所提方法的性能,同样使用了ISCX non-VPN数据集[23]进行分类,使用其中4 个网络应用类型,分别是视频通话(video)、语音通话(audio)、文字聊天(chat)、IP语音(VoIP)。ISCX non-VPN流量数据集统计信息如表2所示。
为保证Moore数据集特征的有效性,将数据集中每一条网络流样本的249 个特征依次读取,替换特征中的无关特征字段,具体替换形式为
Y→(1.0)
N→(0.0)
?→(0.0)
由于二维CNN的输入格式为数值矩阵,因此需要将预处理好的特征进行矩阵化处理,对于Moore数据集具有的249 位特征而言,选择构建一个16×16的矩阵。由于特征维数少于矩阵元素个数,故而需要进行填充操作。不同于文献[21]采取0填充的方式,采取将每行数据除标签、缺失值和布尔型值外的特征进行均值计算,在矩阵的末位进行8 次均值填充的方式,上述改进不仅可以满足CNN的输入特性。而且有利于减少信息损失、增加模型的鲁棒性。
将填充后归一化的矩阵中每个元素作为像素点映射到灰度图像中,Moore数据集中每种应用类型对应的灰度图如图3所示。对于ISCX non-VPN数据集而言,采用文献[23]的图像处理方式,将原始特征通过特征融合扩展到21 616 个特征,规整到0~255,按照特征对应一个像素点映射到灰度图像中,ISCX non-VPN每种应用类型对应的灰度图如图4所示。
Moore数据集作为经典网络流量数据集虽然被研究人员广泛使用,但仔细观察表1可知,Moore数据集的377 526 个网络流量数据样本中,INTERACTIVE的类别样本数在整个数据集中只有110 个,占比仅有0.028%,GAMES类别样本数则更少,仅有8 个,占比0.002%。
为了防止Moore数据集中网络流量类别比例差距过大或不均衡可能导致模型过拟合于高比例类别,以及难以识别分类小比例类别数据的问题发生,选择使用ImageDataGenerator进行实验设计,ImageDataGenerator作为Keras深度学习框架中的一个图像数据生成器,它在深度学习中常用于数据增强。ImageDataGenerator通过随机变换、裁剪、缩放、旋转、平移等操作来增加图像数据的多样性,从而扩充数据量。这种方法可以帮助模型更好地泛化到不同的数据分布,提高模型的性能和鲁棒性。
设计的ImageDataGenerator生成器的详细参数如表3所示。通过对数据集中小比例类别进行旋转、平移、翻转、剪切、缩放操作从而达到扩充数据集的效果。
通过对Moore数据集进行分析,本文中使用表3设计的ImageDataGenerator对INIERACTIVE和GAMES类别进行个性化数据增强,具体样本数量变化如表4所示,对INIERACTIVE和GAMES类别进行数据增强的图像如图5所示,该图展示了两种类别数据增强后的图像。
针对超参数组合选择不当可能会导致分类模型出现欠拟合或过拟合的问题,提出了一种五折交叉验证与基于贝叶斯优化的Optuna相结合的超参数组合优化方式[26]。通过使用Optuna生成超参数组合,利用五折交叉验证方法进行模型训练从而优化超参数组合。
使用五折交叉验证对模型超参数组合进行优化,其核心思想如图6所示,将原始训练集划分为5 份进行5 次实验,选择其中4 份数据用于训练,剩余1 份用于评价效果,重复5 次后保证每份数据都扮演过训练集和验证集的角色,验证集在训练过程中用于检验模型的训练情况,确定合适的超参数组合。最后得到模型的5 次评价结果,取均值作为该超参数组合下的模型的最终评价结果,从而可以比较不同超参数下模型的评价结果,进行模型超参数优化。
相比于以往研究中超参数优化使用网格搜索算法遍历给定的超参数组合来优化模型,使用Optuna进行超参数组合的选择。Optuna方法较网格搜索方法的优势如表5所示。
网格搜索方法需要研究者提前给出需要超参数遍历的空间,获取的超参数取值都是在给定的准确值内进行迭代选择的,对于提出的基于数据增强与AM改进的CNN分类模型来讲,其超参数组合是无法准确手动定义的。且需确定学习率、丢弃率、批次大小、迭代次数的复杂超参数组合,并不是简单、单一的超参数,使用网格搜索需要经过嵌套循环遍历,无疑增加了运行时间。因此选择Optuna方法进行超参数优化。
采用Optuna与五折交叉验证相结合的方式进行超参数组合优化,首先使用Optuna定义超参数组合的范围,在范围内限制30 次超参数组合进行实验,并且30 种不同的超参数组合都经过五折交叉验证来获取其平均准确率。为了防止模型过拟合,本文使用早停机制防止过度迭代,利用Optuna指定的学习率、丢弃率、批次大小的搜索空间如表6所示。
由实验所得30 种超参数组合的准确率得分可知,采用学习率、丢弃率、迭代次数、批次大小设置值为
learning_rate=0.000 484 496 235 947
dropout_rate=0.062 456 633 270 408
batch_size=256
防止分类模型的超参数选择不佳可能会影响分类准确率的问题发生。
CNN作为DL的重要领域之一,其广泛运用于计算机视觉以及自然语言处理方面。一维CNN适用于序列数据或语言数据,二维CNN擅长处理图像或音频频谱图等数据,三维CNN适用于视频或体积图像等数据。根据Moore和ISCX non-VPN的数据集可知,针对Moore和ISCX non-VPN数据集转换的二维图像数据,使用二维CNN进行分类是最合适的选择。因此本文模型在二维CNN的基础上进行研究。
然而,传统的CNN模型在处理复杂图像时,往往难以捕捉到图像中的关键信息。AM允许模型在处理信息时,能够动态地聚焦于输入数据的某些部分,同时忽略其他不相关的信息。这种机制模拟了人类处理复杂信息时的注意力分配过程,有助于模型更好地捕捉关键特征,提高任务的准确性。本文通过引入AM来增强模型对关键特征的关注,提高分类性能。
本文中提出的模型采用序贯模型(sequential)结构,主要包括卷积层、池化层、全连接层以及自定义的AM层。CNN-Optuna-Attention模型具体架构如图7所示。CNN-Optuna-Attention模型网络参数结构如表7所示。
(1)卷积层。卷积层作为CNN的核心模块,用于提取输入图像的局部特征。本文模型使用两个卷积层,第一个卷积层有8 个3×3的卷积核,输入形状为16×16×1,匹配二维图像的格式,第二个卷积层有16 个3×3的卷积核。每个卷积层后都使用了ReLU激活函数,以增强模型的非线性表达能力。特征图大小计算方式如下。
假设输入体积大小为H1×W1×D1(高度×宽度×通道数),卷积核数量为K,卷积核大小为F×F,步长为S,零填充大小为P可得
H2= H 1 - F + 2 P S+1
W2= W 1 - F + 2 P S+1
D2=K
式中:H2为输出高度;W2为输出宽度;D2为输出通道数。
(2)池化层。池化层用于对卷积层输出的特征图进行下采样,以减少计算量并提取更重要的特征。本文模型使用两个最大池化层,池化窗口大小为2×2,步长为2。
(3)全连接层。连接层用于将卷积层和池化层提取的特征整合起来并进行分类。每个全连接层后都使用了ReLU激活函数,并添加了Dropout层以防止过拟合。
(4)自定义AM层。注意力层用于增强模型对关键特征的关注。该层应用于第二个全连接层之后,接受全连接层的输出作为输入,通过计算输入特征的权重并乘以原始特征,实现对关键特征的加强。本文AM层具体实现细节如下。
设输入张量为X,且
X∈RN×T×D
式(10)中:N为批次大小(batch_size);T为序列长度;D为特征维度。
① 权重和偏置初始化:
权重矩阵为W,且
W∈RD×1
偏置向量为b,且
b∈RT×1
② 计算能量:
对于输入张量X中的每一个元素xntd(其中n表示批次索引,t表示时间步索引,d表示特征索引),其对应的能量ent计算公式为
ent∈tanh( d = 1 Dxntdwd+bt)
式(13)中:wd为权重矩阵W的第d个元素;bt为偏置向量b的第t个元素。
③ 计算注意力权重:
注意力权重ant可以通过对能量ent应用softmax函数得到,即
ant= e x p ( e n t ) k = 1 T e x p ( e n k )
④ 应用注意力权重并求和:
最终的输出yn是输入张量X与注意力权重a的加权和,沿着时间步t进行求和,即
yn= t = 1 Tant( d = 1 Dxntd)
式(15)中:yn为批次中第n个样本的输出。
(5)输出层。输出层用于输出最终的分类结果。本文模型使用softmax激活函数,输出层有12/4 个神经元,对应于Moore和ISCX non-VPN数据集的流量类别数,将全连接层的输出转换为概率分布,从而进行多分类任务。
实验设备方面,实验采用配备CPU为i5-8300H,4核2.30 GHz,内存为8 GB,GPU为NVIDIA GeForce GTX 1050 的计算机和Windows 11的操作系统。编程方面,选择Python作为编程语言并使用PyCharm作为开发工具。在数据处理方面,采用Numpy和Pandas库,深度学习框架选择TensorFlow。
为了评估本文所提方法的分类性能,本文中使用准确率A、精确率P以及F1分数这3个指标进行性能评估。将预测结果用符号定义为真正例(TP)、假正例(FP)、假反例(FN)、真反例(TN),如表8所示。
(1)准确率A:准确率作为分类问题最基本的评估指标,其表示模型正确预测的样本数量与所有预测样本数量之间的比率,该指标用于评估模型的整体性能,准确率越高代表分类效果越好,其表达式为
A= T P + T N T P + T N + F P + F N
(2)精确率P:精确率能够有效评估分类模型对每一个类别的分类好坏,其表示在所有被模型判定为正例的样本中,实际上的确是正例的样本所占的比例,其表达式为
P= T P T P + F P
(3)F1分数:F1分数是精确率P和召回率R的调和平均数,其表示为
F1= 2 P R P + R
R= T P T P + F N
将本文设计的方法在Moore数据集上的实验结果与BP神经网络方法、文献[21]分类方法、使用默认超参数且没有引入AM机制的本文模型方法(简称对比方法1),以及基于条件生成对抗网络[27]实现数据增强且引入本文AM机制的方法(简称对比方法2)做对比。将本文方法在ISCX non-VPN数据集上的实验结果与文献[15,22-23]的方法作对比。通过表9给出的Moore数据集下5 种方法的准确率对比可知,本文所提方法在准确率方面要优于其他对比方法。图8图9分别给出了不同方法在Moore数据集上的精确率和F1分数的对比柱状图表,经仔细分析可知本文所提方法的改进模型要较优于BP神经网络和文献[21]方法,并且通过本文所以方法与对比方法1相比,验证了本文所提方法在超参数优化以及五折交叉验证方法上的有效性,通过本文所提方法与对比方法2相比,验证了本文所提方法在数据增强方面的优势。表10给出了ISCX non-VPN数据集下4种方法的准确率对比。可以看出本文所提方法与模型在准确率方面要优于文献[15,22-23]。图10图11分别给出了不同方法在ISCX non-VPN数据集上的精确率和F1分数的对比柱状图表。由图表可知,本文所提模型与优化方法的精确率与F1分数优于文献[5,22]的方法,同是采用相同特征融合方式处理的ISCX non-VPN数据集,与文献[23]相比在聊天类别方面略低,但在视频、语音、IP语音均比其他方法效果更好。
为了更为直观地展示本文所提方法的分类性能,图12给出了本文所提方法在Moore与ISCX non-VPN数据集上的混淆矩阵。通过以上两个数据集的模型测试结果对比。进一步验证了本文中所使用的数据增强方法、超参数优化方法和引入AM改进的CNN分类模型较其他方法的有效性。
针对当前使用DL进行NTC任务存在的问题,提出了一种基于数据增强技术和CNN-Optuna-Attention改进的网络流量分类方法,该方法首先对Moore与ISCX non-VPN数据集进行恰当的预处理,然后使用不同的方式转换为二维图像,通过分析Moore数据集的数据量可知其类别分布存在很大的问题,因此采取ImageDataGenerator对Moore数据集进行个性化的数据增强,扩充了GAMES和INTERACTIVE的类别样本。其次使用五折交叉验证和Optuna对分类模型的超参数组合进行优化,选择出适合分类模型的超参数,最后将上述两个数据集转换出的图像、超参数组合提供给本文经AM改进的CNN分类模型,实现高效准确的分类。
通过与其他文献的研究方法对比可知,本文所提方法在准确率、精确率、F1分数上都有着提高。在未来的工作中,将根据以下几个方面进行深入研究:①适当改变CNN分类模型的结构,比较网络流量分类的效果。②尝试利用其他数据增强方法对Moore数据集进行数据扩充,比较数据增强后生成图像的效果好坏。③在进行超参数优化时将Optuna框架中的n_trials的取值适当的增大,增加超参数组合的种类,将更多模型中需要考虑的超参数加入到超参数优化的范围内,通过该方式获取分类模型的多种类型超参数组合的结果。
  • 山东理工大学科技博士项目(4041422007)
参考文献 引证文献
排序方式:
[1]
中国互联网信息中心CNNIC. 第53次《中国互联网络发展状况统计报告》[EB/OL]. (2024-03-22) [2024-06-18]. https://www.cnnic.net.cn/n4/2024/0321/c208-10962.html. https://www.cnnic.net.cn/n4/2024/0321/c208-10962.html
China Internet Information Center CNNIC. The 53rd statistical report on the internet development in China[EB/OL]. (2024-03-22) [2024-06-18] https://www.cnnic.net.cn/n4/2024/0321/c208-10962.html. https://www.cnnic.net.cn/n4/2024/0321/c208-10962.html
[2]
郑春艳. 浅谈计算机网络发展现状[J]. 现代计算机, 2022, 28(21): 77-79.
Zheng Chunyan. Introduction to the current status of computer network development[J]. Modern Computer, 2022, 28(21): 77-79.
[3]
Azab A, Khasawneh M, Alrabaee S, et al. Network traffic classification: techniques, datasets, and challenges[J]. Digital Communications and Networks, 2022, 10(3): 676-692.
[4]
Munarko A. Determinants of service quality: efficiency and workspace[J]. AKADEMIK: Jurnal Mahasiswa Humanis, 2022, 2(3): 136-144.
[5]
任成刚. 大数据下的计算机网络安全技术分析[J]. 网络安全和信息化, 2023, (11): 148-150.
Ren Chenggang. Analysis of computer network security technology under big data[J]. Network Security and Informatization, 2023, (11): 148-150.
[6]
Zhao J, Jing X, Yan Z, et al. Network traffic classification for data fusion: a survey[J]. Information Fusion, 2021, 72: 22-47.
[7]
杨宇, 闫钰, 申芳, 等. 基于机器和深度学习的入侵检测综述[J]. 科学技术与工程, 2023, 23(18): 7607-7621.
Yang Yu, Yan Yu, Shen Fang, et al. A review of intrusion detection based on machine and deep learning[J]. Science Technology and Engineering, 2023, 23(18): 7607-7621.
[8]
Shahraki A, Abbasi M, Taherkordi A, et al. Active learning for network traffic classification: a technical study[J]. IEEE Transactions on Cognitive Communications and Networking, 2021, 8(1): 422-439.
[9]
Moore A W, Zuev D. Internet traffic classification using Bayesian analysis techniques[C]// Proceedings of the 2005 ACM Sigmetrics International Conference on Measurement and Modeling of Computer Systems. New York: ACM, 2005: 50-60.
[10]
Madhukar A, Williamson C. A longitudinal study of P2P traffic classification[C]// 14th IEEE International Symposium on Modeling, Analysis, and Simulation. Monterey: IEEE, 2006: 179-188.
[11]
Gringoli F, Salgarelli L, Dusi M, et al. GT: picking up the truth from the ground for internet traffic[J]. Computer Communication Review, 2009, 39: 12-18.
[12]
罗冬梅. 网络协议流不平衡环境下基于机器学习算法的在线流量分类方法[J]. 科学技术与工程, 2017, 17(28): 103-107.
Luo Dongmei. Online traffic classification method based on machine learning algorithm in network protocol flow imbalance environment[J]. Science Technology and Engineering, 2017, 17(28): 103-107.
[13]
王瑞敏, 汤云凯, 郝高鑫, 等. 基于机器学习的网络流量分类研究[J]. 山西电子技术, 2024(2): 104-106, 119.
Wang Ruimin, Tang Yunkai, Hao Gaoxin, et al. Research on network traffic classification based on machine learning[J]. Shanxi Electronic Technology, 2024(2): 104-106, 119.
[14]
于治平, 刘彩霞, 刘树新, 等. 基于机器学习的网络流量分类综述[J]. 信息工程大学学报, 2023, 24(4): 447-453, 483.
Yu Zhiping, Liu Caixia, Liu Shuxin, et al. A review of network traffic classification based on machine learning[J]. Journal of Information Engineering University, 2023, 24(4): 447-453, 483.
[15]
Perna G. Online classification of RTC traffic[C]// 2021 IEEE 18th Annual Consumer Communications & Networking Conference (CCNC). Las Vegas: IEEE, 2021: 1-6.
[16]
张超群, 易云恒, 周文娟, 等. 基于深度学习与数据增强技术的小样本岩石分类[J]. 科学技术与工程, 2022, 22(33): 14786-14794.
Zhang Chaoqun, Yi Yunheng, Zhou Wenjuan, et al. Small-sample rock classification based on deep learning and data enhancement techniques[J]. Science Technology and Engineering, 2022, 22(33): 14786-14794.
[17]
Matsuo Y, Lecun Y, Sahani M, et al. Deep learning, reinforcement learning, and world models[J]. Neural Networks, 2022, 152: 267-275.
[18]
顾玥, 李丹, 高凯辉. 基于机器学习和深度学习的网络流量分类研究[J]. 电信科学, 2021, 37(3): 105-113.
Gu Yue, Li Dan, Gao Kaihui. Research on network traffic classification based on machine learning and deep learning[J]. Telecommunications Science, 2021, 37(3): 105-113.
[19]
宋继红, 王桐桐. 基于特征融合卷积神经网络的加密流量分类方法[J]. 信息与电脑(理论版), 2021, 33(2): 49-51.
Song Jihong, Wang Tongtong. An encrypted traffic classification method based on feature fusion convolutional neural network[J]. Information and Computer Science (Theoretical Edition), 2021, 33(2): 49-51.
[20]
李道全, 王雪, 于波, 等. 基于一维卷积神经网络的网络流量分类方法[J]. 计算机工程与应用, 2020, 56(3): 94-99.
Li Daoquan, Wang Xue, Yu Bo, et al. A network traffic classification method based on one-dimensional convolutional neural network[J]. Computer Engineering and Applications, 2020, 56(3): 94-99.
[21]
王勇, 周慧怡, 俸皓, 等. 基于深度卷积神经网络的网络流量分类方法[J]. 通信学报, 2018, 39(1): 14-23.
Wang Y, Zhou Huiyi, Feng Hao, et al. Network traffic classification method based on deep convolutional neural network[J]. Journal on Communications, 2018, 39(1): 14-23.
[22]
孔镇, 董育宁. 一种基于深度学习的网络流量细粒度分类方法[J]. 南京邮电大学学报(自然科学版), 2021, 41(3): 100-108.
Kong Zhen, Dong Yuning. A deep learning-based approach for fine-grained classification of network traffic[J]. Journal of Nanjing University of Posts and Telecommunications (Natural Science Edition), 2021, 41(3): 100-108.
[23]
于帅, 董育宁, 邱晓晖. 一种基于深度特征融合的网络流量分类方法[J]. 南京邮电大学学报(自然科学版), 2022, 42(3): 82-89.
Yu Shuai, Dong Yuning, Qiu Xiaohui. A network traffic classification method based on deep feature fusion[J]. Journal of Nanjing University of Posts and Telecommunications (Natural Science Edition), 2022, 42(3): 82-89.
[24]
马继烨, 朱国胜, 卫操, 等. 一种噪声容忍的网络流量分类方法[J]. 计算机科学, 2023, 50(S2): 775-781.
Ma Jiye, Zhu Guosheng, Wei Cao, et al. A noise-tolerant network traffic classification method[J]. Computer Science, 2023, 50(S2): 775-781.
[25]
何迎利, 胡光宇, 张浩, 等. 基于局部信息增强注意力机制的网络流量预测[J]. 科学技术与工程, 2023, 23(30): 13014-13022.
He Yingli, Hu Guangyu, Zhang Hao, et al. Network traffic prediction based on local information-enhanced attention mechanism[J]. Science Technology and Engineering, 2023, 23(30): 13014-13022.
[26]
李艳辉, 王衍萌. 基于数据增强技术与CNN-BiLSTM-Attention的油田注水流量预测及效果[J]. 科学技术与工程, 2023, 23(32): 13896-13902.
Li Yanhui, Wang Yanmeng. Oilfield water injection flow prediction and effect based on data enhancement technique with CNN-BiLSTM-Attention[J]. Science Technology and Engineering, 2023, 23(32): 13896-13902.
[27]
金麒旻. 生成对抗网络的质量提升技术研究[D]. 北京: 北京邮电大学, 2023.
Jin Qimin. Research on quality improvement technology of generative adversarial network[D]. Beijing: Beijing University of Posts and Telecommunications, 2023.
2025年第25卷第15期
PDF下载
385
146
引用本文
BibTeX
文章信息
doi: 10.12404/j.issn.1671-1815.2404568
  • 接收时间:2024-06-18
  • 首发时间:2025-07-09
  • 出版时间:2025-05-28
补充材料
相关文章
文章信息
作者
出版历史
  • 收稿日期:2024-06-18
  • 修回日期:2025-11-17
基金
山东理工大学科技博士项目(4041422007)
作者信息
    山东理工大学计算机科学与技术学院, 淄博 255049

通讯作者:

* 崔鑫(1972—),女,汉族,山东淄博人,博士,副教授。研究方向:下一代互联网技术、网络安全、网络大数据、无线传感网。E-mail:
参考文献
分享链接
https://castjournals.cast.org.cn/joweb/kxjsygc/CN/10.12404/j.issn.1671-1815.2404568
分享至
全文二维码

扫描看全文

引用本文
BibTeX
本文的引用情况
2种不同金属材料的力学参数

Family
属数
Number of
genus
种数
Number of
species
占总种数比例
Percentage of
total species (%)

Genus
种数
Number of
species
占总种数比例
Percentage of total
species (%)
鹅膏菌科Amanitaceae 2 11 5.26 鹅膏菌属 Amanita 10 4.78
小菇科 Mycenaceae 2 12 5.74 丝盖伞属 Inocybe 5 2.39
多孔菌科 Polyporaceae 8 14 6.70 蜡蘑属 Laccaria 5 2.39
红菇科 Russulaceae 3 23 11.00 小皮伞属 Marasmius 6 2.87
小菇属 Mycena 11 5.26
光柄菇属 Pluteus 5 2.39
红菇属 Russula 17 8.13
栓菌属 Trametes 5 2.39
关闭全屏