结合深度学习和自适应的业务动态访问控制研究及应用分析

打开文本图片集

中图分类号：TN915.08-34；TP311 文献标识码：A 文章编号：1004-373X（2025）16-0050-05

Research and application analysis of business dynamic access control based on deep learning and adaptive technology

LIWei¹，ZHANGJinjin²，ZHANGXiaoyan1 （1.China MobileCommunicationsGroup Shaanxi Co.，Ltd.，Xi’an71oo61，China; 2.Collegeof Armament Scienceand Technology，Xi'an Technological University，Xi'an 710o21，China）

Abstract：Inordertosolve theproblem thattraditionalaccess control meansare dificulttomeet therequirementsof dynamicassessmentandresponse，amethodofzero-trustsecurityaccessintegratingdeeplearningandadaptivedynamicaccess control（DA-ZeroTrust）isproposed.The vectorizedrepresentationofuserbehaviors isconstructedandthetemporal dependency relationshipsofinteractionsequenesareexploedsoastealizteontiuosaemetofuserbehaviorsandteection ofabnormaluseridentities.TheMarkovdecisionprocessisusedtoevaluatethevalueofaccessbehavior，soastorealizethe adaptiveallcationofdynamicaccesscontrol permissions.Theexperimentalresultsshowthatthismethodcan ffectively overcomekeytechnicaldificultiessuchasuser encoding，semanticfeatureextractionandabnormalbehaviordetection，andis able to quickly detect and respond to abnormal behaviors.

Keywords：accesscontrol;zerotrust;deeplearning；adaptivedynamicaccesscontrol;anomalydetection;Markovdecision process;access behavior value

0 引言

传统的网络安全防御机制将所有恶意攻击和安全风险都归咎于外部人侵，这实际上暗含了对内部网络环境中人员、设备、系统和应用的信任。（剩余6383字）