SIDE-CHANNEL INFORMATION LEAK DETECTION WITH WAVELET TRANSFORMATION

Subject of study. The paper presents the results of the experimental study of a wavelet-based pre-processing that was used to increase the amount of statistically-discoverable information leakage in a side-channel signal. The studied signal was acquired via an electromagnetic channel of the ARM-Cortex M4F processor. The signal was recorded at the distance of 1 mm and 3 cm. During signal collection, the Advanced Encryption Standard (AES) algorithm was executed on the board. Method. After the acquisition, electromagnetic signal is processed using a discrete wavelet transform with shrinkage of the detail coefficients. The influence of various shrinkage thresholds, mother wavelet, and wavelet decomposition level is studied. After processing of the signal records, the leakage analysis is performed using the Test Vector Leakage Assessment (TVLA), a method based on the Welch statistical test. The obtained estimates are used to compare wavelet transform pre-processing with the leakage estimate of the original signal. Main Results. The signal processed by a wavelet transform shows higher values of the statistical test, that means more confidence in the presence of an information leakage. The universal threshold and zeroing detail coefficients increase the value of the t-criterion by 1.4 times. The third level of decomposition shows the highest result for all wavelet functions. The discrete wavelet of Meyer shows the best result in all experiments. Symlets and Coiflets also show stable results in the experiments. Practical Relevance. Statistical methods, such as the Welch statistical test, can detect leakage without costly and time-consuming stages such as research and attacks. The wavelet transform and processing of the received signal increases the informational components of the signal providing close-to-real statistical signal profiles. Pre-processing based on wavelet transform also allows for leakage detection on fewer signal records.