深度学习的用户数据自监督安全防御

打开文本图片集

引用格式：喻佳.深度学习的用户数据自监督安全防御[J].现代电子技术，2025，48（20）：30-34.

关键词：用户数据；深度学习；自监督；安全防御；编码器网络；异常攻击识别；数据风险等级评估中图分类号：TN912.2-34；TP393 文献标识码：A 文章编号：1004-373X（2025）20-0030-05

Abstract：Inorder tocope withcomplex andever-changingdataattack paterns，process datastreams inreal-time，and enhance securitydefensecapabilities，a methodofuserdata self-supervisedsecuritydefensebasedondeep learning is proposed. Theuserdatasecuritydefensemodelisconstructed，theencodernetworkanddecodernetwork inthevariationautoencoderare usedforthedataprocessingbycombiningdeplearningandself-supervisedlearningtechnology，soastoidentifyuserdata abnormaldefense，calculateuserdatastandarddeviation，evaluatedatarisklevel，andimplementuserdatasecuritydefense accordingtotheresultsofdatarisk assessment.Bytaking thestudentachievementdata fromthe teaching managementof a colegeiProvinceasthebasicdataset，thedefensefectoftheproposedmethodisdetected.Theexperimentalresults demonstratethatthismethodcaneffectivelyhandlestudentuserdataunderlow，medium，andhighatack intensities，ensuring theintegrityof studentachievementdata.Undervarying amountsof abnormaldata，thedefenseratecanremainabove96%，with a dataleakageriskbelow 1.67% .The securitylevel ishigh，and the fluctuation range of defense capabilityisless than 2% . The proposed method can contribute to the intelligent development in the field of data security defense.

Keywords：userdata;deeplearning;self-monitoring;securitydefense;encodernetwork;abnormalatack identification;data risk level assessment

0 引言

用户数据作为数字化生态的核心，能够记录与用户相关的数据，为人们提供精准的数字化信息。（剩余5437字）