结合特征降维和NGO-CNN-BiLSTM的招考智慧平台网络异常流量检测方法

打开文本图片集

关键词：招考智慧平台；特征降维；卷积神经网络；双向长短期记忆网络；北方苍鹰优化算法DOI：10.15938/j. jhust. 2025.02.007中图分类号：TP391 文献标志码：A 文章编号：1007-2683（2025）02-0064-09

Abstract：Inviewofthechalengesbroughtbycomputervirusesandnetwork maliciousattckstotheoperationandmaintenanceof therecruitmentintelligentplatfomsystem，thispaperproposesamethodtodetectabnoaltraffcofthercruitmentsmartplatfor basednfeaturedimensionalityreductioncombinedwithNorthernGoshawkOptimization（NGO）tooptimizeconvolutional neural network（CNN）andBi-directionalLong Short-Trm Memory（BiLSTM）.Byusing theKernelPrincipal ComponentAnalysis（KPCA） methodtoreducethedimensionalityofthenonlinearinformationcontaindinthedataset，therduceddataisusedasinputforthdep learningnetworkmodel.Then，theconvolutionkernelofCNNisoptimizedbyNGOtoobtaintheoptimalconvolutionkerel，andthe abnormatraffcwasdetectedbyBiLSTM.TheCIC-IDS-2017datasetisusedtoanalyzethetrainingandtestsamplesoftheitrusion detectionnetworkmodel，andtheacuracyandrainingtimeareimprovedcomparedwithother methods，whichconfirmsthefeasibility and effectiveness of this method.

Keywords：recruitmentintellgence platfom；featuredimensionalityeduction;ConvlutionalNeuralNetwork；Bi-directioalLong Short-Term Memory； Northern Goshawk Optimization

0 引言

招考智慧平台具有更高的开放性，而开放性必然导致潜在的安全威胁，如何提升招考智慧平台的入侵检测系统的流量数据分析处理能力，并提高网络攻击检测效率、异常数据分类精度和入侵检测的自动化水平，是招考智慧平台网络安全面临的核心和关键问题。（剩余13349字）