A physical layer key enhancement method combining clustering and wavelet transform

Abstract

The application discloses a physical layer key enhancement scheme combined with clustering and wavelet transform and belongs to the technical field of wireless channel physical layer key generation; the physical layer key enhancement scheme comprises the following steps: step S1: a sender Alice and a receiver Bob transmit probe sequences to each other; step S2: amplitude information in a received sequence is extracted as channel characteristic values at a receiving end, and the channel characteristic values are pretreated by combining a wavelet transform with a K-means clustering algorithm; and step S3: the pretreated channel characteristic values are subjected to equal-probability quantization to generate a key. The application utilizes the characteristic that noise information is distributed in high-frequency wavelet coefficients in the wavelet transform, effectively filters out noise information in the characteristic values by combining a K-means clustering method, retains effective information, and thus the purpose of reducing a key inconsistency rate is achieved, and the reliability of the key is enhanced.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of wireless channel physical layer key generation, and particularly relates to a physical layer key enhancement method combining clustering and wavelet transform. BACKGROUND

[0002] The micro wireless channel physical layer key generation technology is a technology for generating a key based on channel characteristics by utilizing the reciprocity of a channel in a half-duplex (TDD) mode, and mainly includes five steps of channel sounding, preprocessing, quantization, key negotiation and privacy enhancement. The technology extracts a key by utilizing the characteristics of a wireless channel, realizes encryption of communication content of a legal communication party, solves the difficulty of key management in a traditional mode, reduces the difficulty of key distribution and update, avoids the process of key distribution, and solves the security problem caused by the open characteristic of a wireless channel.

[0003] The preprocessing stage is an important part of the physical layer key generation technology, and mainly completes the processing of characteristic values extracted in the channel sounding stage by the communication parties, removes the influence of a noise environment on the characteristic values, and enhances the reliability of the key generated in the subsequent quantization stage.

[0004] Clustering is a process of classifying and organizing data members similar in some aspects in a data set. K-means clustering is the most famous partition clustering algorithm, and is the most widely used among all clustering algorithms due to its simplicity and efficiency. Given a data point set and a required number of clusters K, the K-means algorithm repeatedly divides the data into various clusters according to a distance function.

[0005] Wavelet transform is a time-frequency analysis method of a signal, has the characteristics of multi-resolution analysis, and has the ability to represent local characteristics of a signal in both time and frequency domains, is a time-frequency localization analysis method in which a time window and a frequency window can be changed, and is very suitable for analyzing non-stationary signals and extracting local characteristics of signals. At present, it has become an effective method for removing signal noise.

[0006] In the prior art, most preprocessing methods based on wavelet transform use a unified threshold denoising scheme, the threshold is obtained when the dimensionality tends to infinity under the joint distribution of multi-dimensional independent normal variables, and is an optimal solution based on a minimum maximum estimation algorithm. The defect is that the same threshold is used for processing and denoising in multi-layer wavelet decomposition, and the denoising effect cannot be best. SUMMARY

[0007] The application provides a physical layer key enhancement method combining clustering and wavelet transform.

[0008] The application aims at realizing a physical layer key enhancement method combining clustering and wavelet transform, characterized in that the physical layer key enhancement scheme comprises the following steps:

[0009] Step S1: the sender Alice and the receiver Bob send each other a probe sequence;

[0010] Step S2: extract the amplitude information in the received sequence as the channel characteristic value at the receiving end, and pre-process the channel characteristic value through wavelet transform combined with K-means clustering algorithm;

[0011] Step S3: equal-probability quantize the pre-processed channel characteristic value to generate a key.

[0012] Preferably, the wavelet transform in step S2 is a time-frequency domain analysis method, which has the characteristic of multi-resolution analysis and decomposes the signal into wavelet functions of different times at multiple scales.

[0013] Preferably, the pre-processing of the channel characteristic value through wavelet transform combined with K-means clustering algorithm in step S2 specifically comprises the following steps:

[0014] Step S2-1: divide the channel characteristic value matrix into sub-blocks , recombine the by stacking them by columns, and obtain the recombined matrix .

[0015] Step S2-2: perform multi-level wavelet decomposition on the recombined matrix to obtain the approximation coefficient and the wavelet coefficient of each layer.

[0016] Step S2-3: construct the feature quantity of the first layer wavelet coefficient sequence.

[0017] Step S2-4: generate a sub-sequence , , , based on the wavelet coefficient sequence.

[0018] Step S2-5: calculate the standard deviation of each sub-sequence , and construct another feature quantity , of the wavelet coefficient sequence.

[0019] Step S2-6: Use the K-means clustering algorithm to process the sample set composed of two features of the wavelet coefficient sequence. The wavelet coefficients are classified into two classes: one with more noise and one with less noise. The wavelet coefficients from the class with less noise are retained, while those from the other class are filtered out, resulting in the processed first-level wavelet coefficients. ;

[0020] Step S2-7: From the second layer to the... Repeat steps S2-3 to S2-6 for the wavelet coefficients of the layer to obtain the processed wavelet coefficients. ;

[0021] Step S2-8: Combine the processed wavelet coefficients and approximation coefficients from each layer and perform wavelet reconstruction to obtain the preprocessed channel eigenvalue matrix. .

[0022] Preferably, the sub-block in step S2-1 They are respectively and , Greater than the coherent bandwidth, Greater than the coherence time.

[0023] Preferably, in steps S2-7, the second layer to the first layer... The wavelet coefficients of the layer are repeated in steps S2-3 to S2-6 to construct wavelet coefficient subsequences. The standard deviation of each subsequence is selected as the evaluation parameter of the degree of dispersion. The standard deviations of the wavelet coefficient sequence and the subsequences are selected as the sample set. The K-means clustering algorithm is used to classify the sample set, and the wavelet coefficients are divided into two categories with more and less noise.

[0024] Preferably, the wavelet coefficients obtained in steps S2-7 are processed. J represents the number of wavelet decomposition layers, and J uses 5 layers.

[0025] Preferably, in step S2-2, the recombined matrix Perform multi-level wavelet decomposition to obtain approximate coefficients. and wavelet coefficients at each level The specific operation is as follows:

[0026] Wavelet decomposition uses wavelet functions and scaling functions. conduct middle Layer decomposition, represented as:

[0027]

[0028] in, Representing scale The approximation coefficients on the upper level contain macro information, and the useful signal is the main part, The scale The wavelet coefficients on the lower level contain detail information, and the noise is the main part.

[0029] Preferably, the wavelet function is obtained by translating and scaling a mother wavelet; the wavelet function is represented as:

[0030]

[0031] wherein, is the set of all integers, a constant norm is maintained, independent of the scale ; and is used to parameterize the time or spatial position, is used to parameterize the frequency or scale.

[0032] The present application has the following advantages: 1. By using the feature that the noise information is distributed in the high-frequency wavelet coefficients in the wavelet transform, the noise information in the feature values is effectively filtered out and the effective information is retained by combining the K-means clustering method, so that the purpose of reducing the key inconsistency rate is achieved, and the reliability of the key is enhanced.

[0033] 2. By classifying the sample set composed of two feature quantities of the wavelet coefficient sequence using the K-means clustering algorithm, two categories containing more and less noise are obtained, the wavelet coefficients of each layer are processed separately, the noise is further effectively removed, and the effective information is retained, so that the reliability of the key is enhanced. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 is the overall flowchart of the present application.

[0035] Figure 2 is the partial flowchart of the present application.

[0036] Figure 3 is the comparison chart of the key inconsistency rates generated by different schemes. DETAILED DESCRIPTION

[0037] The present application is further described below in conjunction with the drawings.

[0038] As Figure 1 shown, a physical layer key enhancement method combining clustering and wavelet transform, characterized in that the physical layer key enhancement scheme comprises the following steps:

[0039] Step S1: the sender Alice and the receiver Bob send each other probe sequences;

[0040] Step S2: extracting the amplitude information in the received sequence as the channel feature value at the receiving end, and pre-processing the channel feature value through wavelet transform combined with K-means clustering algorithm;

[0041] The pre-processing of the channel feature value through wavelet transform combined with K-means clustering algorithm specifically includes the following steps:

[0042] Step S2-1: dividing the channel feature value matrix into sub-blocks , and recombining the sub-blocks after stacking them by column, to obtain a recombined matrix .

[0043] Dividing the channel feature value matrix into sub-blocks , and the sub-blocks are and , is greater than the coherence bandwidth, is greater than the coherence time.

[0044] Step S2-2: performing multi-level wavelet decomposition on the recombined matrix to obtain an approximation coefficient and wavelet coefficients of each layer.

[0045] The specific operation of performing multi-level wavelet decomposition on the recombined matrix to obtain an approximation coefficient and wavelet coefficients of each layer is as follows:

[0046] The wavelet decomposition is performed through a wavelet function and a scaling function , and is represented as:

[0047]

[0048] wherein represents the approximation coefficient on the scale , contains macro information, and the useful signal accounts for the majority, represents the wavelet coefficient on the scale , contains detailed information, and the noise accounts for the majority.

[0049] wherein the wavelet function is obtained through a mother wavelet through translation and scale stretching; and the wavelet function is represented as:

[0050] ​​​

[0051] wherein, is the set of all integers, maintains a constant norm, independent of the scale is used to parameterize time or spatial position, is used to parameterize frequency or scale.

[0052] Step S2-3: constructing a feature quantity of the first layer wavelet coefficient sequence ;

[0053] Step S2-4: generating a sub-sequence based on the wavelet coefficient sequence, , , , ;

[0054] Step S2-5: calculating the standard deviation of each sub-sequence and constructing another feature quantity of the wavelet coefficient sequence , ;

[0055] Step S2-6: classifying the sample set composed of the two feature quantities of the wavelet coefficient sequence by using the K-means clustering algorithm, obtaining two classes containing more and less noise, selecting the wavelet coefficients in the class containing less noise, filtering out the wavelet coefficients in the other class, and obtaining the processed first layer wavelet coefficients ;

[0056] Step S2-7: repeating steps S2-3 to S2-6 for the second layer to the layer wavelet coefficients to obtain the processed wavelet coefficients ; J is the number of layers of wavelet decomposition, and J is 5 layers;

[0057] Step S2-8: combining the processed wavelet coefficients of each layer with the approximation coefficient array to perform wavelet reconstruction, thereby obtaining the preprocessed channel feature value matrix .

[0058] Step S3: performing equal-probability quantization on the preprocessed channel feature values to generate a key.

[0059] The main purpose of the application is to enhance the reliability of the key, only performing three steps of channel detection, preprocessing, and quantization; the step schematic diagram is shown as Figure 1 ;

[0060] ​​The simulation verification of the present application is carried out, the simulation adopts Jacks multipath channel model, and specific channel parameters are shown in Table 1. The amplitude information in the channel response is extracted as a characteristic value. The selected wavelet function in the wavelet transform is db4, and 5-layer wavelet decomposition is carried out.

[0061] Table 1 simulation parameter setting table:

[0062]

[0063] In Table 2, the standard deviation of the difference between the characteristic values of the channels of the two communication parties before and after preprocessing is studied. It can be seen that after the preprocessing scheme of the present application, the standard deviation of the difference between the characteristic values of the channels is greatly reduced. Therefore, the keys generated by the two communication parties will have a lower inconsistency rate.

[0064] Table 2 standard deviation of the difference between the characteristic values of the channels of the two communication parties:

[0065] Scheme Before pre-processing Uniform threshold denoising pre-processing Pre-processing of the invention Standard deviation 0.367 0.158 0.073

[0066] Figure 3 The inconsistency rate of the keys generated by different schemes under different signal-to-noise ratios is studied. In the figure, the inconsistency rate of the keys generated by quantization without preprocessing, by the uniform threshold denoising scheme and by the present application scheme is compared. From the simulation, it can be concluded that the inconsistency rate of the keys generated by quantization after preprocessing has decreased significantly. At the same time, compared with the uniform threshold denoising scheme, the keys generated by the joint K-means clustering and wavelet transform key enhancement scheme proposed by the present application have further improved the inconsistency rate of the keys and enhanced the reliability of the keys.

[0067] Working principle: through the joint clustering and wavelet transform physical layer key enhancement scheme of the channel characteristic values, the K-means clustering is used to filter out the wavelet coefficients with more noise in each layer of the multi-layer wavelet transform, so that the two communication parties obtain reconstructed characteristic values with lower standard deviation of the difference between the channel characteristic values, and at the same time, compared with the uniform threshold denoising scheme, the keys with higher reliability are obtained.

[0068] The above only describes the embodiments of the present application and is not used to limit the present application. For those skilled in the art, the present application can have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the scope of the claims of the present application.

Claims

1. A physical layer key enhancement method combining clustering and wavelet transform, characterized in that: The physical layer key enhancement method comprises the following steps: Step S1: the sender Alice and the receiver Bob send each other a probe sequence; Step S2: extracting the amplitude information in the received sequence as a channel characteristic value at the receiving end, and pre-processing the channel characteristic value through wavelet transform combined with a K-means clustering algorithm; Step S2-1: divide the channel feature value matrix ;​​​​​ Step S2-2: recombining the matrix performing multi-level wavelet decomposition to obtain approximation coefficients and wavelet coefficients of each layer ; Step S2-3: constructing a feature quantity of the 1st layer wavelet coefficient sequence ; Step S2-4: generating a sub-sequence based on the sequence of wavelet coefficients, , , , ; Step S2-5: Calculate the standard deviation of each sub-sequence and construct another feature quantity of the wavelet coefficient sequence , ;​ Step S2-6: using K-means clustering algorithm to classify the sample set composed of the two characteristic quantities of the wavelet coefficient sequence Classifying to obtain two categories containing more and less noise, selecting the wavelet coefficients in the category containing less noise, filtering the wavelet coefficients in the other category to obtain the processed first layer wavelet coefficients ; Step S2-7: Wavelet coefficient repetition of the second layer to the first layer Step S2-7: Wavelet coefficient repetition of the second layer to the first layer Step S2-7: Wavelet coefficient repetition of the second layer to the first layer Step S2-8: combine the processed wavelet coefficients of each layer with the approximation coefficients to perform wavelet reconstruction, thereby obtaining a pre-processed channel feature value matrix ; Step S3: generating a key by equi-probability quantization of the pre-processed channel characteristic value.

2. The physical layer key enhancement method combining clustering and wavelet transform according to claim 1, characterized in that: The wavelet transform in step S2 is a time-frequency domain analysis method, which has the characteristic of multi-resolution analysis, and decomposes the signal into wavelet functions of different times at multiple scales.

3. The method of claim 1, wherein: The sub-block in the step S2-1 respectively and , greater than a coherence bandwidth, greater than a coherence time.

4. The physical layer key enhancement method based on joint clustering and wavelet transform according to claim 1, characterized in that: The step S2-7 of repeating the wavelet coefficient of the second layer to the The steps S2-3 to S2-6 of repeating the wavelet coefficient of the second layer to the 5. The method of claim 1, wherein: The wavelet coefficients obtained in the step S2-7 after processing J is the number of layers of wavelet decomposition, and J is 5 layers.

6. The physical layer key enhancement method based on joint clustering and wavelet transform according to claim 1, characterized in that: The step S2-2 recombines the matrix The multi-level wavelet decomposition is performed to obtain the approximation coefficients and the wavelet coefficients of each layer The specific operation is as follows: The wavelet decomposition is performed by a wavelet function and a scaling function is performed in layer decomposition, denoted as ； where denotes the approximation coefficients at scale contain macro information, with the useful signal being the dominant part, denotes the wavelet coefficients at scale contain detail information, with the noise being the dominant part.

7. The method of claim 6, wherein: The wavelet function is obtained by translating and scaling a mother wavelet; the wavelet function is represented as: ； in, It is the set of all integers. Maintain a constant norm, relative to the scale Irrelevant; Used to parameterize time or spatial location. Used to parameterize frequency or scale.

Images