CN110992980B - Hidden latent channel identification method based on edge calculation - Google Patents

Hidden latent channel identification method based on edge calculation Download PDF

Info

Publication number
CN110992980B
CN110992980B CN201911194763.3A CN201911194763A CN110992980B CN 110992980 B CN110992980 B CN 110992980B CN 201911194763 A CN201911194763 A CN 201911194763A CN 110992980 B CN110992980 B CN 110992980B
Authority
CN
China
Prior art keywords
channel
hidden
signals
mfccs
training
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201911194763.3A
Other languages
Chinese (zh)
Other versions
CN110992980A (en
Inventor
许爱东
张宇南
蒋屹新
文红
刘文洁
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Electronic Science and Technology of China
CSG Electric Power Research Institute
Original Assignee
University of Electronic Science and Technology of China
CSG Electric Power Research Institute
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by University of Electronic Science and Technology of China, CSG Electric Power Research Institute filed Critical University of Electronic Science and Technology of China
Priority to CN201911194763.3A priority Critical patent/CN110992980B/en
Publication of CN110992980A publication Critical patent/CN110992980A/en
Application granted granted Critical
Publication of CN110992980B publication Critical patent/CN110992980B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/03—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters
    • G10L25/24—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the type of extracted parameters the extracted parameters being the cepstrum
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS OR SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING; SPEECH OR AUDIO CODING OR DECODING
    • G10L25/00—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00
    • G10L25/27—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique
    • G10L25/30—Speech or voice analysis techniques not restricted to a single one of groups G10L15/00 - G10L21/00 characterised by the analysis technique using neural networks
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L63/00—Network architectures or network communication protocols for network security
    • H04L63/14—Network architectures or network communication protocols for network security for detecting or protecting against malicious traffic
    • H04L63/1408—Network architectures or network communication protocols for network security for detecting or protecting against malicious traffic by monitoring network traffic
    • H04L63/1416—Event detection, e.g. attack signature detection
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D30/00—Reducing energy consumption in communication networks
    • Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks

Abstract

The invention discloses a hidden channel identification method based on edge calculation, which is characterized in that through extracting Mel frequency cepstrum coefficients (Mel-Frequency Cepstral Coefficients, MFCCs) and adopting a machine learning method to learn coefficients of a normal communication system and hidden channel hidden communication, a hidden channel identification device is trained, and new Mel frequency cepstrum coefficients of a specific system are input into the system to realize identification of hidden channel hidden communication; the method has the advantages of low calculation complexity and high recognition accuracy for the terminal.

Description

Hidden latent channel identification method based on edge calculation
Technical Field
The invention relates to the field of security attack identification, in particular to a hidden latent channel identification method based on edge calculation.
Background
With the development of communication technology, digital voice communication is more and more widely applied, and attacks against digital voice communication are generated, and the latent channel is an attack method against digital voice communication; the latent channel is a channel which is not perceived by common people and actually exists, so that an attacker can use the channels which are not perceived and actually exist to transmit secret information, or steal information of a mobile phone, a terminal node and the like, so that the user cannot perceive that the information is stolen, attacked and transmitted.
The method comprises the steps of taking high-frequency sound waves as a carrier, utilizing signal amplitude modulation to convert a voice signal into the high-frequency sound waves, modeling an audio circuit, carrying out low-pass filtering and then restoring an original signal downwards, so that a redundant type submarine channel can be utilized to carry out short-distance transmission of voice and data information, and the transmission of the information is realized under the condition that people cannot perceive the submarine channel, namely a hidden submarine channel attack method; the recognition of hidden channel attacks is a necessity for protecting a communication system from malicious attacks, and particularly many mobile terminals, such as intelligent terminals, power terminals and the like, are equipped with voice input and voice control functions, so that hidden channel attacks for digital voice communication can carry out hidden camouflage command input, thereby launching attacks on the whole network system from the terminals; the hidden submarine channel attack method still needs a certain carrier for communication, so that the hidden submarine channel attack method can be identified through characteristic analysis of communication signals.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a hidden latent channel identification method based on edge calculation.
The method collects, trains and learns the mel frequency cepstrum coefficient of the terminal and the like under the edge computing platform, then identifies whether the hidden channel hidden communication exists for the terminal communication, and has the advantages of low computing complexity and high identification accuracy for the terminal.
A hidden channel identification method based on edge calculation comprises one or more edge calculation devices, wherein the edge calculation devices execute a hidden channel identification step, and the hidden channel identification step identifies an input channel signal and judges whether the signal is from the hidden channel.
Further, the method further comprises a step of extracting the characteristics of the mel-frequency cepstral coefficient MFCCs, wherein the mel-frequency cepstral coefficient is extracted from the input channel signal.
Further, the method also comprises a hidden channel distinguishing and training step, wherein the hidden channel distinguishing and training step trains and generates a hidden channel identification device in a machine learning mode so as to judge whether an input channel signal is from the hidden channel.
Further, the training step comprises the following sub-steps:
s1: extracting MFCCs characteristics of the channel signal;
s2: inputting the extracted MFCCs characteristics into the hidden latent channel recognition device to perform hidden latent channel recognition training;
s3: repeating the steps S1-S2 until the set recognition qualification rate is reached.
Further, the channel signals include normal communication channel signals and/or hidden channel signals.
Further, the normal channel signal MFCCs feature extraction includes the following steps:
extracting normal channel signal MFCCs characteristics and identified as D 1 ；
D 1 ＝{X 1 ,X 2 ,...,X i },
Figure BDA0002294414880000021
Wherein X is i ＝(x i,1 ,x i,2 ,...,x i,j ),j＝1,2,...,36；
The n is a set of n channel signals, the i is a set of normal communication channel signals, and each set of channel signals includes 36 waveform signals.
Further, the extracting of the characteristics of the hidden latent channel MFCCs includes the following steps:
extracting the characteristics of the known covert channel communication signal MFCCs and identified as D 2 ；
Figure BDA0002294414880000022
X u ＝(x u,1 ,x u,2 ,...,x u,j ),j＝1,2,...,36；
The n is a set of n channel signals, the u is a set of known hidden latent channel communication signals, and each set of channel signals includes 36 waveform signals.
Further, the method also comprises an MFCCs feature preprocessing step, wherein the MFCCs feature preprocessing step comprises the following steps:
from X i And X u G columns are arbitrarily taken from the channel signal set matrix formed by (1) to obtain:
Figure BDA0002294414880000023
k=1, 2,. -%, n; wherein->
Figure BDA0002294414880000024
b＝1,2,...,g；
Repeating the selection process G times to obtain a new set:
Figure BDA0002294414880000025
m＝1,2,...,G；
the said
Figure BDA0002294414880000026
Is from X i And X u Training vectors formed by g columns are arbitrarily taken from the channel signal set matrix,
Figure BDA0002294414880000027
is->
Figure BDA0002294414880000028
A training subset is formed; />
Figure BDA0002294414880000029
Is the final training set.
Further, the recognition training comprises the following sub-steps:
vector pair
Figure BDA00022944148800000210
Two by two for calculation: />
Figure BDA00022944148800000211
Giving that the weight matrix W belongs to the target d:
Figure BDA0002294414880000031
wherein C is j To attack communications for the potential channel.
The invention has the beneficial effects that: through extracting the Mel frequency cepstrum coefficient of the normal communication system and the hidden communication of the hidden channel, learning by adopting a machine learning method, training a discriminator, and realizing the identification of the hidden communication of the hidden channel; the method collects, trains and learns the mel frequency cepstrum coefficient of the terminal and the like under the edge computing platform, then identifies whether the hidden channel hidden communication exists for the terminal communication, and has the advantages of low computing complexity and high identification accuracy for the terminal.
Drawings
Fig. 1 is a flow chart of a hidden latent channel identification method based on edge calculation.
Detailed Description
In order to more clearly understand the technical features, objects and effects of the present invention, the technical solution of the present invention will be described in further detail with reference to the accompanying drawings and the random subspace k nearest neighbor machine learning algorithm (random subspace integration k-nearest neighbour, RS-KNN), but the scope of the present invention is not limited to the following description.
A hidden channel identification method based on edge calculation comprises one or more edge calculation devices, wherein the edge calculation devices execute a hidden channel identification step, and the hidden channel identification step identifies an input channel signal and judges whether the signal is from the hidden channel.
The hidden latent channel identification method as shown in fig. 1 further includes a mel-frequency cepstral coefficient MFCCs feature extraction step of extracting a mel-frequency cepstral coefficient from an input channel signal.
The channel signals include normal communication channel signals and/or hidden latent channel signals.
Extracting the characteristics of the normal channel signal MFCCs, comprising the following steps:
extracting normal channel signal MFCCs characteristics and identified as D 1 ；
D 1 ＝{X 1 ,X 2 ,...,X i },
Figure BDA0002294414880000032
Wherein X is i ＝(x i,1 ,x i,2 ,...,x i,j ),j＝1,2,...,36；
Where n is the extracted set of n channel signals including normal communication channel signals and hidden latent channel signals, and i is the set of normal communication channel signals, each set of channel signals including 36 waveform signals.
Extracting the characteristics of the MFCCs of the hidden latent channel, comprising the following steps:
extracting the characteristics of the known covert channel communication signal MFCCs and identified as D 2 ；
Figure BDA0002294414880000033
X u ＝(x u,1 ,x u,2 ,...,x u,j ),j＝1,2,...,36；
Where u is the set of known hidden latent channel communication signals.
Then through the MFCCs feature preprocessing steps, the MFCCs feature preprocessing steps are as follows:
from X i And X u G columns are arbitrarily taken from the channel signal set matrix formed by (1) to obtain:
Figure BDA0002294414880000041
k=1, 2,. -%, n; wherein->
Figure BDA0002294414880000042
b＝1,2,...,g；
Repeating the selection process G times to obtain a new set:
Figure BDA0002294414880000043
m＝1,2,...,G；
the said
Figure BDA0002294414880000044
Is from X i And X u Training vectors formed by g columns are arbitrarily taken from the channel signal set matrix,
Figure BDA0002294414880000045
is->
Figure BDA0002294414880000046
A training subset is formed; />
Figure BDA0002294414880000047
Is the final training set.
Training and generating the hidden latent channel identification device through the hidden latent channel resolution training step, specifically, the training step trains and generates a hidden channel recognition device in a machine learning mode so as to judge whether an input channel signal comes from the hidden channel.
The training step comprises the following sub-steps:
s1: extracting MFCCs characteristics of the channel signal;
s2: inputting the extracted MFCCs characteristics into the hidden latent channel recognition device to perform hidden latent channel recognition training;
s3: repeating the steps S1-S2 until the set recognition qualification rate is reached.
The recognition training comprises the following sub-steps:
vector pair
Figure BDA0002294414880000048
Two by two for calculation: />
Figure BDA0002294414880000049
Giving that the weight matrix W belongs to the target d:
Figure BDA00022944148800000410
wherein C is j To attack communications for the potential channel.
Identification training:
inputting the MFCC characteristics of the communication system to be detected in the hidden channel identification device and judging hidden channel hidden communication attack;
and extracting the MFCC characteristics of the communication system to be detected, inputting the MFCC characteristics into a trained hidden channel identification device for judgment, and judging whether hidden channel hidden communication attack occurs.
The foregoing has shown and described the basic principles and main features of the present invention and the advantages of the present invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, and that the above embodiments and descriptions are merely illustrative of the principles of the present invention, and various changes and modifications may be made without departing from the spirit and scope of the invention, which is defined in the appended claims. The scope of the invention is defined by the appended claims.

Claims (7)

1. The hidden channel identification method based on edge calculation comprises one or more edge calculation devices, and is characterized in that the edge calculation devices execute a hidden channel identification step, the hidden channel identification step identifies input channel signals and judges whether the signals come from the hidden channel; further comprising a hidden latent channel resolution training step of training and generating a hidden latent channel identification means by means of machine learning to determine whether an input channel signal is derived from said hidden latent channel, said training step comprising the sub-steps of:
s1: extracting MFCCs characteristics of the channel signal;
s2: inputting the extracted MFCCs characteristics into the hidden latent channel recognition device to perform hidden latent channel recognition training;
s3: repeating the steps S1-S2 until the set recognition qualification rate is reached.
2. The method of claim 1, further comprising a mel-frequency cepstral coefficient MFCCs feature extraction step of extracting mel-frequency cepstral coefficients from an input channel signal.
3. The edge-calculation-based hidden channel identification method of claim 1, wherein the channel signals comprise normal communication channel signals and/or hidden channel signals.
4. The hidden channel recognition method based on edge calculation according to claim 3, wherein the feature extraction for the normal channel signal MFCCs comprises the steps of:
extracting normal channel signal MFCCs characteristics and identified as D 1 ；
Figure FDA0004119891590000011
Wherein X is i ＝(x i,1 ,x i,2 ,...,x i,j ),j＝1,2,...,36；
The n is a set of n channel signals, the i is a set of normal communication channel signals, and each set of channel signals includes 36 waveform signals.
5. The method for identifying hidden channels based on edge calculation according to claim 3, wherein the feature extraction for the hidden channel signal MFCCs comprises the following steps:
extracting the characteristics of the known covert channel communication signal MFCCs and identified as D 2 ；
Figure FDA0004119891590000012
The n is a set of n channel signals, and u is a set of known hidden latent channel communication signals, each set of channel signals comprising 36 waveform signals.
6. The edge computation-based hidden channel identification method of any one of claims 4 or 5, further comprising MFCCs feature preprocessing steps of:
from X i And X u G columns are arbitrarily taken from a channel signal set matrix formed, and the following steps are obtained:
Figure FDA0004119891590000013
wherein->
Figure FDA0004119891590000014
Repeating the selection process G times to obtain a new set:
Figure FDA0004119891590000021
the said
Figure FDA0004119891590000022
Is from X i And X u Training vectors formed by g columns are arbitrarily selected from the channel signal set matrix formed by the components, and the training vectors are +.>
Figure FDA0004119891590000023
Is that
Figure FDA0004119891590000024
A training subset is formed; />
Figure FDA0004119891590000025
Is the final training set.
7. The method of edge-calculation-based hidden channel identification of claim 1, wherein the identification training comprises the sub-steps of:
vector pair
Figure FDA0004119891590000026
Two by two for calculation: />
Figure FDA0004119891590000027
Giving that the weight matrix W belongs to the target d:
Figure FDA0004119891590000028
wherein C is j To attack communications for the potential channel.
CN201911194763.3A 2019-11-28 2019-11-28 Hidden latent channel identification method based on edge calculation Active CN110992980B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911194763.3A CN110992980B (en) 2019-11-28 2019-11-28 Hidden latent channel identification method based on edge calculation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911194763.3A CN110992980B (en) 2019-11-28 2019-11-28 Hidden latent channel identification method based on edge calculation

Publications (2)

Publication Number Publication Date
CN110992980A CN110992980A (en) 2020-04-10
CN110992980B true CN110992980B (en) 2023-06-23

Family

ID=70087959

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911194763.3A Active CN110992980B (en) 2019-11-28 2019-11-28 Hidden latent channel identification method based on edge calculation

Country Status (1)

Country Link
CN (1) CN110992980B (en)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2004090886A1 (en) * 2003-04-08 2004-10-21 Koninklijke Philips Electronics N.V. Fragile audio watermark related to a buried data channel
CN101577619A (en) * 2008-05-08 2009-11-11 吴志军 Real-time speech secret communication system based on information hiding
CN105763271A (en) * 2015-12-12 2016-07-13 天津南大通用数据技术股份有限公司 Concealed channel detection method and apparatus
US10299278B1 (en) * 2017-03-20 2019-05-21 Amazon Technologies, Inc. Channel selection for multi-radio device

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2004090886A1 (en) * 2003-04-08 2004-10-21 Koninklijke Philips Electronics N.V. Fragile audio watermark related to a buried data channel
CN101577619A (en) * 2008-05-08 2009-11-11 吴志军 Real-time speech secret communication system based on information hiding
CN105763271A (en) * 2015-12-12 2016-07-13 天津南大通用数据技术股份有限公司 Concealed channel detection method and apparatus
US10299278B1 (en) * 2017-03-20 2019-05-21 Amazon Technologies, Inc. Channel selection for multi-radio device

Non-Patent Citations (6)

* Cited by examiner, † Cited by third party
Title
Pros and cons of Mel-cepstrum based audio steganalysis using SVM classification;Kraetzer C,Dittmann J.;《Information Hiding Lecture Notes on Computer Science》;20071231;全文 *
Temporal derivative-based spectrum and mel-cepstrum audio steganalysis;LIU Qingzhong,SUNG AH,QIAN Mengyu.;《IEEE Transactions on Information Forensics and Security》;20090930;363页 *
基于半监督学习的即时语音通信隐藏检测;涂山山等;《清华大学学报(自然科学版)》;20151115(第11期);1246-1249页 *
基于声信道的隐蔽信息传输关键技术;丁雪洁等;《信息安全研究》;20160205(第02期);全文 *
涂山山等.基于半监督学习的即时语音通信隐藏检测.《清华大学学报(自然科学版)》.2015,(第11期), *
隐通道传递信息机理的研究;刘志锋等;《计算机科学》;20071015(第10期);全文 *

Also Published As

Publication number Publication date
CN110992980A (en) 2020-04-10

Similar Documents

Publication Publication Date Title
CN109599109B (en) Confrontation audio generation method and system for white-box scene
Xiao et al. Secure mobile crowdsensing based on deep learning
Nie et al. Deep hierarchical network for automatic modulation classification
Wang et al. Ghosttalk: Interactive attack on smartphone voice system through power line
Gong et al. An overview of vulnerabilities of voice controlled systems
Wang et al. Deep neural networks for CSI-based authentication
CN109685200B (en) Mist computing industrial protocol construction method and system based on generation countermeasure network
CN110381509B (en) Combined authentication method and server suitable for dynamic connection scene
Arif et al. Voice spoofing countermeasure for logical access attacks detection
CN109977839A (en) Information processing method and device
KR20090089674A (en) An apparatus of sound recognition in a portable terminal and a method thereof
Kong et al. Adversarial audio: A new information hiding method and backdoor for dnn-based speech recognition models
CN114727286B (en) Threshold-free multi-attribute physical layer authentication method and related equipment
Bartusiak et al. Synthesized speech detection using convolutional transformer-based spectrogram analysis
CN110992980B (en) Hidden latent channel identification method based on edge calculation
CN104392161B (en) The device authentication based on audio physical fingerprint under the conditions of displacement
Wang et al. Specific emitter identification based on deep adversarial domain adaptation
CN110519765B (en) Cooperative physical layer authentication method and system based on received signal power
CN111161759B (en) Audio quality evaluation method and device, electronic equipment and computer storage medium
Wang et al. Supervised and semi-supervised deep neural networks for CSI-based authentication
CN115208507A (en) Privacy protection method and device based on white-box voice countermeasure sample
CN115083426A (en) High-fidelity voice desensitization method and device based on antagonistic sample generation
CN112735381B (en) Model updating method and device
CN111755014B (en) Domain-adaptive replay attack detection method and system
CN114118158A (en) Robust electromagnetic signal modulation type identification method for black reflection box detection attack

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant