Research On Covert Communication And Security Protection Method Based On Subliminal Channel Acoustic Wave

The confidentiality and security of voice communication are becoming more and more important as communication technology advances.The subliminal channel is a channel established in the public channel to achieve covert communication.In order to achieve the covert transmission of voice information,this thesis uses the nonlinear effect of the audio circuit to reveal the basic principle of the subliminal channel attack in voice communication and explore a subliminal channel covert communication method of voice communication,and through audio feature analysis,combined with machine learning algorithms,to achieve the identification of subliminal channel attacks.This article has completed the following aspects of the subliminal channel acoustic wave covert communication and security protection methods:The theoretical basis of subliminal channel acoustic wave covert communication is analyzed,and the acoustic wave modulation and nonlinear effects are modeled.An experimental platform was set up,a simple and low-cost system was designed and completed,and a subliminal channel acoustic wave covert communication system was realized on the basis of small cost and small volume.In order to study the signal characteristics of subliminal channel acoustic wave communication,an audio feature analysis scheme based on Mel Frequency Cepstral Coefficients(MFCC)is proposed.MFCC is a widely used audio feature,mainly used for voice recognition and generation.This feature has the characteristics of focusing on the auditory characteristics of the human ear.The experiment established a 36-dimensional feature audio signal model,using Matlab to analyze the feature of the audio signals,and proposed an audio feature recognition model based on MFCC,which provides a data basis for subsequent machine learning to identify malicious attacks.In order to solve the problem of malicious attack recognition,a subliminal channel acoustic wave recognition scheme based on machine learning is proposed.An discriminant model based on audio features is established,and actual audio signal data is collected to further analyze its performance.The collected data combined with multiple machine learning classification algorithms verified the feasibility of the model.With 36-dimensional audio features,the RS-KNN algorithm can achieve an average recognition rate of 99.6%.Finally,a malicious attack identification scheme under edge computing is proposed.On the basis of saving terminal's own computing resources,the subliminal channel acoustic wave malicious attack identification scheme becomes intelligent.To sum up,this thesis uses the subliminal channel acoustic wave covert communication as the application background.It mainly studies the theoretical basis of acoustic wave subliminal channel covert communication,audio feature model establishment,machine learning algorithm analysis,hardware platform implementation,experimental data verification,verified the correctness of the theoretical results.