基于机器学习的电源电压比较实时检测功率分析攻击.pdf

1、Machine Learning-Based Real-Time Detection of Power Analysis Attacks Using Supply Voltage ComparisonsNan Wang,Ruichao Liu,Yufeng Shan,Yu Zhu,Song Chen2025.1.21Presenter:RuiChao LiuBackground01Experimental Result04Research Objectives02Conclusion05contentsDetection Method03BackgroundResearch Objective

2、sDetection MethodExperimental ResultConclusion Side-channel attacks steal sensitive information by collecting physical data,with power analysis attack(PAA)being one of them,focusing on power supply voltage curve.3BackgroundExamples of Power Analysis AttacksExamples of Side-Channel Attacks Introducti

3、on PAABackgroundResearch ObjectivesDetection MethodExperimental ResultConclusionPAA ModelBackgroundTargeting the pins inside the package responsible for powering the target encryption core,to obtain a favorable power consumption curve for analysis,it is necessary to insert an attack resistor at the

4、chip package location.PAA ModelBackgroundResearch ObjectivesDetection MethodExperimental ResultConclusionCircuit ResponseBackgroundThe attack resistor will cause a voltage drop across the nodes on the power grid,resulting in the maximum at the affected node.Circuit Response of PAABackgroundResearch

5、ObjectivesDetection MethodExperimental ResultConclusion1.Insert constant current source and use low metal layer routing to mitigate PAA.2.Add noise to power curve to lower data-power correlation via random switching capacitor distribution,confusing time-domain waveforms.6BackgroundNo countermeasures

6、 against attacksProtectionBackgroundResearch ObjectivesDetection MethodExperimental ResultConclusion1.Capture power curves,detect intrusions with real-time linear classifier analysis.2.Use ring oscillators to identify abnormal voltage changes in the power grid caused by the.7BackgroundNoise resistan