The objectified procedure and a technology for assessing the state of complex noise speech information protection

Abstract

White noise generators are typically applied for the systems of active interference in the design of systems for protecting verbal information. The level of speech information protection against leaking through acoustic and vibrational channels is determined by employing the appropriate normative method and technology. However, using the multi-channel methods for interception of language information, as well as modern methods for processing digital phonograms (wavelet transform, correlation analysis, etc.), allow the intruder to gain unauthorized access. Attempts to apply generators of speech-like noise based on the use of white noise (and its colored clones), reverberation methods, as well as the Language Choir method and some others, do not resolve the set task. In the framework of this research, we have proposed a technique to overcome these difficulties. It is based on using speech-like interference generators the type of scrambler and applying the objectivized method and a technology to assess the degree of verbal information protection at the border of controlled zone. The objectivized method combines methods for determining the criteria for speech residual intelligibility (methods by Pokrovski and Speech Intelligibility Index), methods for filtering complex noise acoustic signals (wavelet transforms, phonemic correlation analysis and others) and the method for comparing a test signal at the point of location of the signal source and at the border of controlled zone. That makes it possible to improve reliability of the resulting estimate of the level of linguistic signal protection against leaking via acoustic and vibration channels beyond the controlled zone’s borders. In order to investigate the level of protection of speech information at different types of the interference and at different ratios of signal/interference, we have designed a simulation model of the experiment. A technology has been proposed to synthesize test signals based on random phonograms and/or phonograms of articulation tables, recorded by voice narrators, and on mixing different types of interference noise at the assigned ratios of signal to interference. Our research was performed in the block «Wavelet 1-D» from the programming environment Matlab. It was established that when applying a noise interference, the type of white noise, and at the ratios of interference to signal of 20...24 dB, the proposed procedure improves the residual intelligibility of test signal from W≤10 % to W≈40…60 %.