Tongxin xuebao (May 2023)
Non-specific TVLA method based on two-sample KS test
Abstract
Test vector leakage assessment (TVLA) is prone to “false negative” when the power consumption sample size is small.To address this issue, it was found that for non-specific TVLA, when the power consumption sample size changes, the test statistic t-values obtained at the leakage sampling points in the power trace vary accordingly, while the t-values at the non-leakage sampling points do not significantly vary.Therefore, when there is leakage, the distributions of the t-values obtained under different sample sizes will be different.Based on this, it was proposed to implement non-specific TVLA under different sample sizes and perform two-sample KS test on the obtained t-value sequences to evaluate whether there was leakage.Verifications were carried out based on unprotected-aligned simulation power consumption, protected-aligned power consumption dataset DPA Contest v4_2 and protected-non-aligned self-collected power consumption respectively.The results showed that the sample size required by the proposed method on the aligned simulation power consumption and DPA Contest v4_2 was reduced by at most 46.1% and 39.0% respectively.And after the alignment, the required sample size of the proposed method on the self-collected power consumption is also smaller than that of other schemes, with a maximum reduction of 29.4%.Therefore, the proposed method can effectively reduce the probability of “false negative” when the power consumption sample size is small.
