SPEECH-LIKE INTERFERENCE RESISTANT TO NOISE REDUCTION METHODS AS A RESULT OF SCRAMBLING OF THE PROTECTED SPEECH

Abstract

The aim of this article is to study the possibility of creating a speech-like interference (SLI) based on scrambling of the protected speech signal that is resistant to noise cleaning, restoration and reconstruction of distorted speech through processing of its spectrogram images, including the use of the speaker’s voice database. Methods: applied systems analysis, digital spectral-time analysis, digital processing of signals and images, image analysis of sonograms Results: it is proposed to use as the main indicator of the security of speech information the concept of potential speech intelligibility, methods for assessing potential intelligibility are determined. An algorithm of compensationtype digital noise cleaning using image analysis-synthesis technology has been developed. The requirements for the created SLI are formed, consisting in changing the spetral-temporal envelope of speech, removing («weakening») traces of the harmonic structure of speech, equalizing the capacities of the protected speech signal and the SLI masking it. The approaches to the formation of an SLI resistant to the procedures of digital noise cleaning and reconstruction of distorted sonograms are described, proposals are presented on the methodology for evaluating the effectiveness of interference through the indicator of potential intelligibility. The results obtained will allow to better understand the work of attacking tools, develop methods and means of effectively countering possible threats, and expand the capabilities of existing speech information protection systems. The scientific novelty consists in the development of criteria for evaluating the effectiveness of speech-like interference, taking into account the capabilities of the intruder to restore the intelligibility of distorted speech messages, proposed methods for the formation of scrambler-type SLI that meet such criteria, substantiated proposals to improve existing methods for assessing and monitoring the security of speech information from its leakage through technical channels.