Computational methods and technical means of processing signals of side electromagnetic emanation

Abstract

The aim of this work is to develop a method to improve the quality of the reconstructed image based on the signals of side electromagnetic fields using post-image processing. To do this, the paper considers the problem of analyzing the side electromagnetic radiation from video displays and ways to solve it. The analysis of the side radiation of electromagnetic waves from information transmission cables, including the HDMI video interface, has been carried out. Due to the 10-bit noise-resistant encoding of video information for digital data transmission interfaces, signal analysis and image restoration are most difficult. Since this encoding expands the bandwidth for side electromagnetic radiation and leads to a nonlinear display of the observed signal and a decrease in the intensity of radiation from the display pixels. Therefore, hardware and software complexes for analyzing analog interfaces receive fuzzy reconstructed images when analyzing digital interfaces. The proposed solution to the problem is to transform the reconstructed image into the original one by training the model on a convolutional neural network. Despite its effectiveness, this approach requires careful mathematical analysis of spurious emissions. This paper will consider this aspect. The complex for conducting experiments is based on an accessible software-defined radio device. The main criterion for the effectiveness of image restoration and improvement after receiving side electromagnetic signals is the Char Error Rate. This indicator has been reduced by more than 60% compared to image restoration without post-processing. To assess fault tolerance, methods are proposed to reduce the probability of image restoration using the complexes described analogues. The results obtained are of practical importance for laboratory studies aimed at assessing data security in various general-purpose systems. In future studies, it is planned to train the model on an updated dataset with other neural network analogues in order to optimize the process of predicting variables in the regression model.

Keywords

• computing systems and their elements, information security, side electromagnetic radiation, video interface, software-defined radio devices, convolutional neural networks, regularization.