Tongxin xuebao (Dec 2021)
Side channel leakage assessment with the Bartlett and multi-classes F-test
Abstract
In order to solve the problem that when test vector leakage assessment （TVLA） technology is used for side channel leakage detection， the mean difference between the two groups of power consumption samples （fixed plaintext and random plaintext） is small， and the t-test may miss detection and lead to false negative evaluation.The Bartlett and multi-classification F-test side channel leakage assessment method （Bartlett-F test） was proposed.In the Bartlett-F test， the Bartlett-test was used to the power samples with greater variance difference than mean difference to solve the problem of missing detection， and the multi-classification F-test was used to the power samples with greater mean difference than variance difference to solve the problem of false negative evaluation.In the test， there is leakage if the P-value is less than the threshold.The experimental results show that when the mean difference is less than the variance difference， the sample size required by Bartlett-test is 1.5×104 when the P-value of Bartlett-test is less than the threshold， while the sample size required by t-test is larger.When the variance difference is less than the mean difference， the sample size required by t-test is 2.0×102 when the P-value of t-test is less than the threshold， while the sample size required by F-test is only 1/10 of t-test.Therefore， Bartlett-F test can solve the problems of TVLA technology in leak detection.
