2023年5月JournalonCommunicationsMay2023第44卷第5期通信学报Vol.44No.5基于双样本KS检验的非特定TVLA方法郑震,严迎建,蔡爵嵩,刘燕江(信息工程大学密码工程学院,河南郑州450001)摘要:测试向量泄露评估(TVLA)在能耗样本量较少时易出现“假阴性”错误。针对该问题进行了理论推导,发现对非特定TVLA,能量迹中存在泄露的采样点处得到的检验统计量t值随能耗样本量变化而变化,非泄露点处的t值则无显著变化,因此当存在泄露时不同能耗样本量下得到的t值序列的分布不同。据此提出在不同样本量下实施非特定TVLA并对得到的t值序列实施双样本KS检验以评估泄露。分别在无防护对齐仿真能耗数据、加防护对齐能耗数据集DPAContestv4_2和加防护非对齐自测能耗数据上进行了验证,结果表明在对齐的仿真能耗数据和DPAContestv4_2数据集上所提方法检测出泄露所需样本量较其他方法均有所减小,最多分别减小了46.1%和39.0%;在非对齐的自测能耗数据进行对齐处理后,所提方法所需能耗样本量较其他方法同样有所减小,最多减小了29.4%。因此所提方法能够有效减小能耗样本量较小时出现“假阴性”错误的概率。关键词:侧信道;测试向量泄露评估;假阴性;双样本KS检验中图分类号:TP309文献标志码:ADOI:10.11959/j.issn.1000−436x.2023101Non-specificTVLAmethodbasedontwo-sampleKStestZHENGZhen,YANYingjian,CAIJuesong,LIUYanjiangDepartmentofCryptogramEngineering,InformationEngineeringUniversity,Zhengzhou450001,ChinaAbstract:Testvectorleakageassessment(TVLA)isproneto“falsenegative”whenthepowerconsumptionsam-plesizeissmall.Toaddressthisissue,itwasfoundthatfornon-specificTVLA,whenthepowerconsumptionsamplesizechanges,theteststatistict-valuesobtainedattheleakagesamplingpointsinthepowertracevaryac-cordingly,whilethet-valuesatthenon-leakagesamplingpointsdonotsignificantlyvary.Therefore,whenthereisleakage,thedistributionsofthet-valuesobtainedunderdifferentsamplesizeswillbedifferent.Basedonthis,itwasproposedtoimplementnon-specificTVLAunderdifferentsamplesizesandperformtwo-sampleKStestontheobtainedt-valuesequencestoevaluatewhethertherewasleakage.Verificationswerecarriedoutbasedonun-protected-alignedsimulationpowerconsumption,protected-alignedpowerconsumptiondatasetDPAContestv4_2andpr...