A multi-attribute physical layer authentication method and device, a terminal and a storage medium

Abstract

The application provides a multi-attribute physical layer authentication method, device, terminal and storage medium. The method comprises: extracting a plurality of physical layer attributes of a received signal as a feature data set; performing bottom-up layering on the feature data set to obtain an l-layer data set; performing top-down weighted clustering on the l-layer data set, and determining the number K of clustering centers of the feature data set based on an average contour coefficient; determining the clustering centers of the feature data set; and judging the user type corresponding to each clustering center based on the distance between the clustering center of the feature data set and a reference vector; wherein the user type comprises a legitimate user and a fraudulent user. The application performs hierarchical clustering on the feature data set, can find high-quality clustering centers without using training samples, thereby reducing the complexity of the early preparation work; meanwhile, the application also adjusts the number of clustering centers in the case of insufficient prior knowledge through the contour coefficient, thereby improving the adaptability of the physical layer authentication.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of communication security, and in particular to a multi-attribute physical layer authentication method and device, a terminal and a storage medium. BACKGROUND

[0002] The main principle of physical layer authentication is to use physical layer attributes such as channel state information (CSI), received signal strength (RSS), angle of arrival (AoA) and carrier frequency offset (CFO) of the communication link and the device as identification signatures for authentication. These physical layer attributes are directly related to the communication device and the corresponding physical environment, so it is difficult to simulate and predict them. However, physical layer attributes can be incompletely estimated or damaged by noise, resulting in unreliable physical layer authentication.

[0003] In recent years, many studies have begun to use multi-observation and multi-attribute to enhance the robustness of physical layer authentication. Since multi-observation or multi-attribute provides additional identifiability for the legitimate transceiver, the probability of an attacker simulating all attributes is extremely low, and such methods can achieve more accurate authentication results. Currently, there are two methods for combining multiple physical layer attributes for identity authentication. The first is a non-parametric method, which specifically establishes a multi-attribute authentication model through training samples. The second is a parametric method, which specifically combines multiple physical layer attributes by setting a form of system expression. The parametric method does not require a large number of training samples and has the advantage of small computational burden, but the choice of system expression has a great impact on the authentication result and has poor adaptability. The non-parametric method does not require pre-setting of the authentication model and has better adaptability, but it requires a large number of training samples due to the increase in the number of attributes and the correlation between attributes, and the preparation work is complex. Therefore, the current physical layer authentication method has the problem of being difficult to balance good effect and low complexity. SUMMARY

[0004] Embodiments of the present application provide a multi-attribute physical layer authentication method, device, terminal and storage medium to solve the problem of the difficulty of balancing good effect and low complexity in the physical layer authentication method.

[0005] In a first aspect, embodiments of the present application provide a multi-attribute physical layer authentication method, comprising:

[0006] extracting a plurality of physical layer attributes of the received signal as a feature data set S(t) = (a ij ) n×m ; wherein n is the number of samples, m is the number of physical layer attributes, and aij represents the jth physical layer attribute of the ith sample in the feature data set;

[0007] performing bottom-up hierarchical clustering on the feature data set by a K-means clustering algorithm to obtain an l-layer data set;

[0008] performing top-down weighted clustering on the l-layer data set to divide the feature data set into k clusters;

[0009] iteratively calculating an average silhouette coefficient of the samples based on the k clusters, and determining a number K of cluster centers of the feature data set based on the average silhouette coefficient;

[0010] determining the cluster centers of the feature data set based on the number K of cluster centers of the feature data set and the cluster centers of the k clusters;

[0011] judging a user type corresponding to each cluster center based on a distance between the cluster center and a reference vector; wherein the user type includes a legitimate user and a fraudulent user.

[0012] In a possible implementation, the bottom-up hierarchical clustering on the feature data set by the K-means clustering algorithm to obtain the l-layer data set comprises:

[0013] taking the feature data set as an A-layer data set, and letting A=1; wherein A is a count value;

[0014] performing clustering on the A-layer data set by the K-means clustering algorithm to obtain M A clusters;

[0015] updating a target function; the target function is:

[0016]

[0017] wherein M A represents a number of clusters obtained by clustering the A-layer data set, n A,k represents a number of samples in the kth cluster obtained by clustering the A-layer data set, represents the qth sample in the kth cluster obtained by clustering the A-layer data set, c A,k represents a centroid of the kth cluster obtained by clustering the A-layer data set;

[0018] taking the centroids of the M A clusters as an A+1-layer data set;

[0019] letting A+1, and jumping to the step of “performing clustering on the A-layer data set by the K-means clustering algorithm to obtain M A clusters” until F is less than a preset thresholdA or A = L, an l-layer dataset is obtained; wherein, L is a preset maximum layer number.

[0020] In a possible implementation, before the l-layer dataset is obtained by performing bottom-up layering on the feature dataset through the K-means clustering algorithm, the method further includes:

[0021] Each sample is normalized by a normalization processing formula to obtain a normalized matrix The normalization processing formula is:

[0022]

[0023]

[0024]

[0025] wherein, a ij represents the jth physical layer attribute of the ith data sample in the feature dataset, a ij represents a ij ;

[0026] A correlation coefficient matrix R = (r m×m ) is obtained by calculating correlation coefficients between the physical layer attributes based on a correlation coefficient formula; the correlation coefficient formula is:

[0027]

[0028] wherein, r ij represents a correlation coefficient between the ith physical layer attribute and the jth physical layer attribute, a ti and a tj covariance;

[0029] An eigenvector of the correlation coefficient matrix is calculated, and m principal components are obtained based on the eigenvector:

[0030]

[0031] wherein, y m represents the mth principal component, u nm represents an nth element in the mth eigenvector, and Z n represents an nth vector in the normalized matrix;

[0032] A variance contribution rate of each principal component is calculated based on a variance contribution rate formula; the variance contribution rate formula is:

[0033]

[0034] wherein, b j denotes the variance contribution rate of the jth principal component, λ k denotes the eigenvalue of the kth principal component;

[0035] The cumulative variance contribution rate of the m principal components is calculated based on a cumulative variance contribution rate formula, and the number of selected principal components is determined based on the cumulative variance contribution rate. The cumulative variance contribution rate formula is:

[0036]

[0037] wherein, α p denotes the cumulative variance contribution rate of the p principal components, and p denotes the number of selected principal components;

[0038] The p principal components with the largest variance contribution rate are selected to obtain the de-correlated feature data set;

[0039] Correspondingly, the K-means clustering algorithm is used to perform bottom-up hierarchical clustering on the feature data set to obtain an l-layer data set, including:

[0040] The K-means clustering algorithm is used to perform bottom-up hierarchical clustering on the de-correlated feature data set to obtain an l-layer data set.

[0041] In a possible implementation, the l-layer data set is subjected to top-down weighted clustering to divide the feature data set into k clusters, including:

[0042] Let B = 1; wherein, B is a count value;

[0043] Randomly select k data in the B-layer data set as the centroids;

[0044] Assign each sample in the B-layer data set to the nearest centroid to obtain k clusters, and update the centroids based on the current k clusters;

[0045] Let B = B-1, and jump to the step of “assigning each sample in the B-layer data set to the nearest centroid to obtain k clusters, and updating the centroids based on the current k clusters”, until B = 1, to obtain the final k clusters.

[0046] In a possible implementation, the average silhouette coefficient of the samples is iteratively calculated based on the k clusters, and the number K of clustering centers of the feature data set is determined based on the average silhouette coefficient, including:

[0047] Let C = k; wherein, C is a count value;

[0048] Calculate the intra-class cohesion and inter-class separation of each sample in the C clusters;

[0049] Calculate a silhouette coefficient of each sample based on the intra-class cohesion and the inter-class separation of each class;

[0050] Calculate an average silhouette coefficient corresponding to C based on the silhouette coefficients of each sample;

[0051] When C = 2, obtain k-1 average silhouette coefficients, and jump to the step of calculating the intra-class cohesion and the inter-class separation of each sample in C clusters until C = 2;

[0052] Take C corresponding to the average silhouette coefficient closest to 1 as the cluster center number K of the feature data set.

[0053] In a possible implementation, determining the cluster centers of the feature data set based on the cluster center number K of the feature data set and the cluster centers of the k clusters comprises:

[0054] Select two centroid centers with the closest Euclidean distance from the centroid centers of each cluster, calculate a mean center point of the two centroid centers as a new centroid center, and remove the two centroid centers, repeat the step until the number of centroid centers is equal to the cluster center number K of the feature data set, and obtain the cluster centers of the feature data set.

[0055] In a possible implementation, determining the user type corresponding to each cluster center based on the distance between the cluster center and the reference vector comprises:

[0056] Calculate the distance between each cluster center and the reference vector;

[0057] Determine the cluster center closest to the reference vector as a legitimate sample, and determine the other cluster centers as fraudulent samples.

[0058] In a second aspect, an embodiment of the present application provides a multi-attribute physical layer authentication device, comprising:

[0059] A feature extraction module configured to extract a plurality of physical layer attributes of a received signal as a feature data set S(t) = (a ij ) n×m ; wherein n is the number of samples, m is the number of physical layer attributes, and a ij represents the jth physical layer attribute of the ith data sample in the feature data set;

[0060] A clustering and layering module configured to perform bottom-up layering on the feature data set by using a K-means clustering algorithm to obtain an l-layer data set;

[0061] A weighted clustering module configured to perform top-down weighted clustering on the l-layer data set to divide the feature data set into k clusters;

[0062] The number determination module is configured to calculate average silhouette coefficients of the samples based on the k clusters, and determine the number K of clustering centers of the feature data set based on the average silhouette coefficients.

[0063] The center selection module is configured to determine the clustering centers of the feature data set based on the number K of clustering centers of the feature data set and the clustering centers of the k clusters.

[0064] The user judgment module is configured to judge the user types corresponding to each clustering center based on distances between the clustering centers of the feature data set and reference vectors, wherein the user types include legitimate users and fraudulent users.

[0065] In a third aspect, an embodiment of the present application provides a terminal, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the steps of the method of the first aspect or any possible implementation manner of the first aspect when executing the computer program.

[0066] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executable on a processor to implement the steps of the method of the first aspect or any possible implementation manner of the first aspect.

[0067] The embodiment of the present application provides a multi-attribute physical layer authentication method, device, terminal and storage medium, and the method comprises the following steps: ij n×m ; wherein n is the number of samples, m is the number of physical layer attributes, and a ij represents the jth physical layer attribute of the ith sample in the feature data set; the feature data set is hierarchically clustered from bottom to top through a K-means clustering algorithm to obtain an l-layer data set; the l-layer data set is weighted clustered from top to bottom, and the feature data set is divided into k clusters; average silhouette coefficients of the samples are iteratively calculated based on the k clusters, and the number K of clustering centers of the feature data set is determined based on the average silhouette coefficients; the clustering centers of the feature data set are determined based on the number K of clustering centers of the feature data set and the clustering centers of the k clusters; and the user types corresponding to each clustering center are judged based on distances between the clustering centers of the feature data set and reference vectors; wherein the user types include legitimate users and fraudulent users. The embodiment of the present application can find high-quality clustering centers without using training samples, thereby reducing the complexity of the preliminary preparation work; meanwhile, the number of clustering centers is adjusted based on the silhouette coefficients in the case of insufficient prior knowledge, thereby improving the adaptability of the physical layer authentication. BRIEF DESCRIPTION OF DRAWINGS

[0068] ​In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained from these drawings without creative labor.

[0069] Figure 1 is the application scenario diagram of the multi-attribute physical layer authentication method provided by an embodiment of the present application.

[0070] Figure 2 is the implementation flowchart of the multi-attribute physical layer authentication method provided by an embodiment of the present application.

[0071] Figure 3 is the structure schematic diagram of the layer data set provided by an embodiment of the present application.

[0072] Figure 4 is the structure schematic diagram of the multi-attribute physical layer authentication device provided by an embodiment of the present application.

[0073] Figure 5 is the schematic diagram of the terminal provided by an embodiment of the present application. DETAILED DESCRIPTION

[0074] In the following description, for the purpose of explanation and not limitation, specific details are set forth, such as particular system configurations, techniques, etc., in order to provide a thorough understanding of the embodiments of the present application. However, it will be apparent to those skilled in the art that the present application can be practiced in other embodiments that depart from these specific details. In other instances, detailed descriptions of well-known systems, devices, circuits, and methods are omitted so as not to obscure the description of the present application with unnecessary detail.

[0075] In order to make the objects, technical solutions and advantages of the present application clearer, the following will be described by specific embodiments in conjunction with the drawings.

[0076] Figure 1 is the application scenario diagram of the multi-attribute physical layer authentication method provided by an embodiment of the present application. As shown in Figure 1 , Alice, Bob and Eve are used to represent different entities. Alice is a legitimate sender, which sends messages to the intended receiver Bob, while x fraudulent attackers (Eve1, …, Eve x ) intend to pretend to be Alice and send fraudulent signals in other communication time slots.

[0077] In wireless communication, users' information is vulnerable to attackers, mainly due to the open broadcast nature of wireless signal propagation, the intermittent nature of communication, and the standardized transmission scheme. A spoofing attacker can monitor or tamper with the private information of a legitimate user by disguising his own MAC address and other identity information as that of a legitimate sender, which poses a great threat to the security of wireless communication. Therefore, a suitable authentication scheme to authenticate users is essential.

[0078] Currently, the identity authentication of wireless terminals mainly adopts a key-based identity authentication scheme, and existing research mainly focuses on secure and efficient key generation, distribution and management. However, the key-based identity authentication scheme has its own limitations. In a decentralized heterogeneous network such as a multi-access edge computing network and a fog computing network, the password authentication has the problem of key distribution and management difficulty. This problem leads to excessive delay, which is unbearable for delay-sensitive communication, such as vehicle networks. In addition, for devices with limited computing power and battery life, such as the Internet of Things (IoT), the computational overhead of password authentication is not desirable. Therefore, physical layer authentication (PLA) is proposed as an alternative solution to key-based identity authentication due to its small amount of computation, small network overhead, and small energy consumption.

[0079] Based on the communication model shown in FIG. 1, the present application adopts a general multi-attribute-based authentication model, in which the identification signature of the received signal at time t is composed of m physical layer attributes, which is modeled as a multi-dimensional random variable as follows: Figure 1

[0080] A t =[a t1 ,…,a tm ] T

[0081] where a tm represents the mth physical layer attribute of the received signal at time t.

[0082] In the ideal model, it can be assumed that all users are static, and the correlation of all users based on the channel attribute is the same. Eve x can generate any waveform signal to disguise as Alice. The position of Eve x is unknown, but it is more than a wavelength away from Alice. In addition, Bob can also extract various physical layer attributes from the received signal.

[0083] The proposed physical layer authentication scheme includes the following two stages.

[0084] ​1) Initialization phase: Alice broadcasts one or more pilot signals to request access to Bob. Then, Bob infers the identification signature from the received signals as a reference vector A r = [a r1 ,...,a rm ] T .

[0085] 2) Authentication phase: Alice sends useful signals to communicate with Bob, while all Eves x pretend to be Alice and send fraudulent signals to Bob in the absence of Alice's signals. Then, Bob estimates the identification signature A t of the received signals and compares it with A r to confirm the legitimacy of the received signals.

[0086] The similarity of A t and A r is represented by d(A t ,A r ). The smaller d(A t ,A r ) is, the more similar A t and A r are, and the greater the probability that the received signal comes from Alice, and vice versa. Therefore, the physical layer authentication is defined as:

[0087]

[0088] where τ represents the threshold value. H0 represents that the received signal comes from the legitimate user Alice, and H1 represents that the received signal comes from the fraudulent user Eve.

[0089] Referring to Figure 2 , which shows the implementation flowchart of the multi-attribute physical layer authentication method provided by the embodiments of the present application, which is described in detail as follows:

[0090] Step 201, extract a plurality of physical layer attributes of the received signal as a feature data set S(t) = (a ij ) n×m ; wherein n is the number of samples, m is the number of physical layer attributes, and a ij represents the jth physical layer attribute of the ith sample in the feature data set.

[0091] In the embodiment, the physical layer attributes for authentication can include channel-based attributes and PUF (physically unclonable function)-based attributes. Specifically, the channel-based attributes have strong identification ability for users, such as CSI, RSS, AoA, etc., and therefore, as many channel attributes as possible should be selected in a static scenario. The PUF-based attributes are stable in time-varying channels (such as CFO and IQI) and are preferred in dynamic situations.

[0092] In the embodiment, a sliding time window can be constructed to filter the received signals in a period of time, and sampling the received signals at a certain frequency can obtain a plurality of samples. Then, a plurality of physical layer attributes of each sample are extracted, and finally, a feature data set S(t) = (a ij ) n×m .

[0093] In step 202, the feature data set is hierarchically divided from bottom to top by the K-means clustering algorithm to obtain an l-layer data set.

[0094] In the embodiment, the final clustering result of the traditional K-means clustering algorithm is greatly affected by the initial centroid. The more the number of samples in the feature data set, the lower the probability that the K-means clustering algorithm finds a high-quality initial centroid in the feature data set by random means. Based on the idea of hierarchical clustering, the embodiment takes the centroid of a group of samples close to each other as a representative sample, and on this basis, the entire feature data set can be simplified into a smaller data set composed of several representative samples with similar distribution to the original data set. From this perspective, a hierarchical structure of the original data set can be constructed. As shown in FIG. 2, the number of the final hierarchy is defined as l, where the first layer is the original feature data set, and each subsequent layer is composed of a smaller data set of the previous layer. Figure 3

[0095] The embodiment divides the original feature data set multiple times to obtain a higher probability of high-quality initial centroids in the final l-layer data set, which can retain the advantage of fast clustering speed of the K-means clustering algorithm while solving the problem of difficulty in determining high-quality centroids by randomly selecting initial centroids. In addition, the high-quality centroids obtained in the embodiment help to accelerate the convergence speed in the subsequent clustering stage, and compared with multiple clustering, the hierarchical step in the embodiment only needs to be executed once, and the amount of calculation and time are less than the algorithm using the idea of multiple clustering. Therefore, compared with the traditional K-means method, the clustering method provided by the embodiment converges faster.

[0096] In step 203, the l-layer data set is weighted clustered from top to bottom to divide the feature data set into k clusters.​

[0097] In this embodiment, among the hierarchical results obtained in step 202, the probability of the l-th layer data set containing high-quality centroids is the highest compared to other layer data sets. Therefore, when clustering, the l-th layer data set is used as the clustering starting point, and clustering is performed layer by layer in a top-down manner, which can obtain a better clustering result for the feature data set.

[0098] Step 204 : Iteratively calculate the average silhouette coefficient of the samples based on the k clusters, and determine the number K of cluster centers of the feature data set based on the average silhouette coefficient.

[0099] In this embodiment, since the application scenario of this application does not require prior information, and the clustering algorithm is not trained based on a pre-acquired deceptive attacker training set, it is necessary to automatically determine the number of cluster sets. The silhouette coefficient is a method for evaluating the clustering effect. For the K-means clustering algorithm, different numbers of cluster centers K will lead to different clustering effects. The average silhouette coefficient combines the two factors of cluster cohesion and separation, and can evaluate the clustering results corresponding to different numbers of cluster centers K, thereby adaptively selecting the number of cluster centers K according to the sample structure to obtain the optimal clustering result.

[0100] Step 205 : Determine the cluster center of the feature data set based on the number K of cluster centers of the feature data set and the cluster centers of the k clusters.

[0101] In this embodiment, the k clusters obtained in step 203 represent a relatively optimal clustering result that can be obtained when the number of cluster centers K is uncertain. The k cluster centers corresponding to the k clusters are the relatively optimal cluster centers selected during the weighted clustering process. After the number of cluster centers K is determined, if the number of cluster centers K is less than the number of cluster centers k already obtained, the k cluster centers can be screened and merged to reduce the number of cluster centers to K, thereby achieving the optimal final clustering result.

[0102] Step 206 : Determine the user type corresponding to each cluster center based on the distance between the cluster center of the feature data set and the reference vector; wherein the user type includes a legitimate user and a fraudulent user.

[0103] In the embodiment, due to time variation of the wireless channel and estimation error, even if all users are static, the physical layer properties of the received signals can still fluctuate greatly, resulting in unreliable physical layer authentication. The embodiment can convert the physical layer authentication into cluster set authentication after clustering the obtained samples, and also convert the similarity between the samples and the reference vector into the distance between the samples and the reference vector. The clustering result of the samples can reflect the statistical characteristics of a specific user. Before clustering, the higher the similarity between the sample and the reference vector, the higher the probability that the user corresponding to the sample is a legitimate user. Correspondingly, after clustering, the closer the distance between the sample and the reference vector, the higher the probability that the user corresponding to the sample is a legitimate user.

[0104] In a possible implementation, the feature data set is hierarchically divided from bottom to top by the K-means clustering algorithm to obtain an l-layer data set, including:

[0105] The feature data set is taken as an A-layer data set, and A is set to 1; wherein A is a count value;

[0106] The A-layer data set is clustered by the K-means clustering algorithm to obtain M A clusters;

[0107] The target function is updated; the target function is:

[0108]

[0109] wherein M A represents the number of clusters obtained by clustering the A-layer data set, n A,k represents the number of samples in the k th cluster obtained by clustering the A-layer data set, represents the q th sample in the k th cluster obtained by clustering the A-layer data set, c A,k represents the centroid of the k th cluster obtained by clustering the A-layer data set;

[0110] The centroids of the M A clusters are taken as an A+1-layer data set;

[0111] A is set to A+1, and the step of clustering the A-layer data set by the K-means clustering algorithm to obtain M A clusters is jumped to, until F is less than a preset threshold A , or A is equal to L, to obtain the l-layer data set; wherein L is a preset maximum number of layers.

[0112] In the embodiment, in order to guarantee the effect and efficiency of the hierarchical feature data set, the Euclidean distance between each sample in each data set after hierarchical is controlled in a reasonable range, and the embodiment determines whether to continue hierarchical by presetting the maximum number of layers L and the target function. The Euclidean distance is used to measure the similarity between sample A t and sample A r , and the calculation formula of the Euclidean distance is as follows:

[0113]

[0114] Wherein, a tl represents the lth physical layer attribute of sample A t , and p represents the total number of physical layer attributes. When hierarchical is performed to a certain stage, if the target function is less than the preset threshold A , it indicates that the Euclidean distance between each data in the latest obtained data set is too small, and hierarchical does not need to be continued.

[0115] In addition, in the hierarchical process, the sample should also satisfy the following constraint conditions:

[0116]

[0117] (2) M l-1 > M l ≥ K

[0118] Wherein, constraint (1) indicates that each sample belongs to only one cluster, constraint (2) indicates that the setting of M l needs to be greater than the number of cluster centers K and less than the number of samples n l of the lth layer, so as to prevent the existence of empty set.

[0119] In a possible implementation, before obtaining the lth layer data set by performing bottom-up hierarchical on the feature data set through the K-means clustering algorithm, the method further comprises:

[0120] standardizing each sample by a standardization processing formula to obtain a standardized matrix The standardization processing formula is as follows:

[0121]

[0122]

[0123]

[0124] Wherein, a ij represents the jth physical layer attribute of the ith data sample in the feature data set, represents a ij after standardization processing;

[0125] The correlation coefficient between each physical layer attribute is calculated based on a correlation coefficient formula, and a correlation coefficient matrix R=(r ij ) m×m The correlation coefficient formula is:

[0126]

[0127] wherein, r ij represents the correlation coefficient between the i-th physical layer attribute and the j-th physical layer attribute, represents the covariance of a ti and a tj ;

[0128] The eigenvectors of the correlation coefficient matrix are calculated, and m principal components are obtained based on the eigenvectors:

[0129]

[0130] wherein, y m represents the m-th principal component, u nm represents the n-th element in the m-th eigenvector, and Z n represents the n-th vector in the standardized matrix;

[0131] The variance contribution rate of each principal component is calculated based on a variance contribution rate formula; the variance contribution rate formula is:

[0132]

[0133] wherein, b j represents the variance contribution rate of the j-th principal component, and λ k represents the eigenvalue of the k-th principal component;

[0134] The cumulative variance contribution rate of the m principal components is calculated based on a cumulative variance contribution rate formula, and the number of selected principal components is determined based on the cumulative variance contribution rate; the cumulative variance contribution rate formula is:

[0135]

[0136] wherein, α p represents the cumulative variance contribution rate of the p principal components, and p represents the number of selected principal components;

[0137] The p principal components with the largest variance contribution rate are selected, and the de-correlated feature dataset is obtained;

[0138] Correspondingly, the feature dataset is hierarchically divided from bottom to top by the K-means clustering algorithm, and l layers of datasets are obtained, including:

[0139] The l-layer data set is obtained by performing bottom-up layering on the de-correlated feature data set through a K-means clustering algorithm.

[0140] In this embodiment, clustering of the samples is to automatically divide the identification signatures in the received signals into different clusters, and the signatures in the same cluster can be regarded as samples from the same user, and thus authentication of the identification signatures can be converted into authentication of the clusters. The sample clustering needs to be implemented based on the similarity between the identification signatures, however, the correlation between the samples can bring exponential growth of the calculation complexity to the clustering algorithm. In this embodiment, principal component analysis is performed on the feature data set, the attributes can be reconstructed through a dimension reduction method, the m-dimensional correlated identification signatures are converted into p-dimensional uncorrelated identification signatures, and thus the subsequent clustering process can be simplified. The dimension reduction and de-correlation of the multi-dimensional attributes can remove the noise and redundant information in the data, greatly reduce the algorithm complexity, and avoid the problem of indistinct differentiation caused by incomplete estimation.

[0141] In a possible implementation, the l-layer data set is subjected to top-down weighted clustering, and the feature data set is divided into k clusters, including:

[0142] Let B = 1, where B is a count value;

[0143] Randomly selecting k data in the B-layer data set as the centroids;

[0144] Assigning each sample in the B-layer data set to the nearest centroid to obtain the k clusters, and updating the centroids based on the current k clusters;

[0145] Let B = B-1, and jump to the step of assigning each sample in the B-layer data set to the nearest centroid to obtain the k clusters, and updating the centroids based on the current k clusters, until B = 1, and the final k clusters are obtained.

[0146] In this embodiment, the number k of the initial centroids should be selected to be a relatively large value, to ensure that k > K after the number K of the clustering centers of the feature data set is determined, so that K high-quality centroids can be selected from the final k centroids, and meanwhile the number k of the initial centroids should be less than or equal to the number of the samples in the l-layer data set. First, the initial centroids are selected in the l-layer data set, and the probability of selecting high-quality centroids is the largest; then, in a top-down manner, the samples in the B-layer data set are assigned to the k clusters, and the centroids are continuously adjusted, so that the centroids can be corrected in the clustering process, and the final k clusters correspond to the k optimal centroids.

[0147] After each sample in the B-layer data set is assigned to the nearest centroid, the centroid can be updated through the following formula:

[0148]

[0149] wherein O k denotes the centroid of the kth cluster, n′ B,k denotes the number of samples in the kth cluster in the Bth data set, w′ q denotes the qth sample in the kth cluster in the Bth data set; when B>1, r i denotes the number of samples in the B-1th data set represented by w′ q , when B=1, r i =1.

[0150] In a possible implementation, the average silhouette coefficient of the samples is iteratively calculated based on the k clusters, and the number K of cluster centers of the feature data set is determined based on the average silhouette coefficient, comprising:

[0151] Let C=k; wherein C is a count value;

[0152] The intra-class cohesion and inter-class separation of each sample in the C clusters are calculated;

[0153] The silhouette coefficient of each sample is calculated based on the respective intra-class cohesion and inter-class separation;

[0154] The average silhouette coefficient corresponding to C is calculated based on the silhouette coefficient of each sample;

[0155] C-1 is obtained, and the step of calculating the intra-class cohesion and inter-class separation of each sample in the C clusters is skipped until C=2, thereby obtaining k-1 average silhouette coefficients;

[0156] The C corresponding to the average silhouette coefficient closest to 1 is taken as the number K of cluster centers of the feature data set.

[0157] In this embodiment, the intra-class cohesion is defined as follows:

[0158] Let sample A i be finally clustered into cluster C1, and the average distance a(i) of A i to all other samples in C1 be a i , then the intra-class cohesion of A

[0159] The inter-class separation is defined as follows:

[0160] Let sample A i be finally clustered into cluster C1, and the average distance b(i) of A i to all samples in other clusters C m be b im , then b(i)=min{b i2 ,b i3 ,...,bik} is the inter-class separation degree of A i .

[0161] The profile coefficient of the sample data is defined as follows:

[0162] If the intra-class cohesion degree of A i is a(i), and the inter-class separation degree is b(i), then the profile coefficient of A i is:

[0163]

[0164] wherein s(i) ∈ [-1, 1], and the closer s(i) is to 1, the more reasonable the clustering of A i is.

[0165] The profile coefficient of the clustering result is defined as follows:

[0166] If the profile coefficient of A i is s(i), then the average of the profile coefficients of all data is the profile coefficient of the clustering result. The closer the value of s(i) is to 1, the more reasonable the entire clustering result is.

[0167] When the data sample s(t) is clustered by K-means, the profile coefficient of the clustering result based on different cluster center quantity K can be calculated. By comparing the profile coefficients, the value of K that makes the entire clustering result most reasonable can be finally determined.

[0168] In a possible implementation, the cluster centers of the feature data set are determined based on the cluster center quantity K of the feature data set and the cluster centers of the k clusters, and the method comprises the following steps:

[0169] selecting two centroid points with the closest Euclidean distance from the centroid points of each cluster, calculating the mean center point of the two centroid points as a new centroid point, and removing the two centroid points, repeating the step until the number of the remaining centroid points is equal to the cluster center quantity K of the feature data set, and obtaining the cluster centers of the feature data set.

[0170] In the embodiment, each time the step of "selecting two centroid points with the closest Euclidean distance from the centroid points of each cluster, calculating the mean center point of the two centroid points as a new centroid point, and removing the two centroid points" is performed, the two clusters closest to each other in the current cluster can be merged, and the number of the current centroid points is reduced by 1. Repeating the step can reduce the number of the centroid points to the optimal cluster center quantity K, and as many high-quality clusters as possible are retained.

[0171] In a possible implementation, the distance between the clustering center of the feature data set and the reference vector is used to determine the user type corresponding to each clustering center, including:

[0172] The distance between each clustering center and the reference vector is calculated.

[0173] The clustering center closest to the reference vector is determined as a legitimate sample, and other clustering centers are determined as fraudulent samples.

[0174] In the embodiment, the clustering center O k The statistical characteristics of the identification signature of each user can be reflected, and the user can be stably observed based on the clustering center. The clustering center O k can be used as a representative identification signature of each cluster, and after clustering, the clustering center O k can be compared with the reference vector A t to make a decision:

[0175] Δ k = d (O k , A t )

[0176] The A t closest to the clustering center of the legitimate user is determined as a legitimate sample, and other clustering centers are determined as fraudulent samples. In addition, the fraudulent cluster can be regarded as a group of signals from a fraudulent user, and the total number of fraudulent clusters is the number of fraudulent users.

[0177] In a specific embodiment, the multi-attribute physical layer authentication method provided by the application can be implemented by computer program code. In order to facilitate understanding, Table 1 is a natural language description of the computer program code of the multi-attribute physical layer authentication method provided by the application.

[0178] Table 1

[0179]

[0180]

[0181] The multi-attribute physical layer authentication method provided by the embodiment of the application includes: extracting a plurality of physical layer attributes of a received signal as a feature data set S (t) = (a ij ) n×m ; wherein n is the number of samples, m is the number of physical layer attributes, and a ija i,j represents the jth physical layer attribute of the ith sample in the feature data set; the K-means clustering algorithm is used to perform bottom-up layering on the feature data set to obtain an l-layer data set; the l-layer data set is subjected to top-down weighted clustering to divide the feature data set into k clusters; the average silhouette coefficient of the samples is iteratively calculated based on the k clusters, and the number K of cluster centers of the feature data set is determined based on the average silhouette coefficient; the cluster centers of the feature data set are determined based on the number K of cluster centers of the feature data set and the cluster centers of the k clusters; the user type corresponding to each cluster center is determined based on the distance between the cluster center of the feature data set and the reference vector; wherein the user type includes a legitimate user and a fraudulent user. The embodiment performs hierarchical clustering on the feature data set, can find high-quality cluster centers without using training samples, thereby reducing the complexity of the preliminary preparation work; meanwhile, the number of cluster centers is adjusted based on the silhouette coefficient in the case of insufficient prior knowledge, thereby improving the adaptability of the physical layer authentication.

[0182] It should be understood that the size of the serial number of each step in the above embodiment does not mean the order of execution, and the execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiment of the application.

[0183] The following is a device embodiment of the application, and for details not described in detail, reference can be made to the corresponding method embodiments described above.

[0184] Figure 4 The structure of the multi-attribute physical layer authentication device provided by the embodiment of the application is shown, only the part related to the embodiment of the application is shown for the convenience of description, and the details are as follows:

[0185] As shown in Figure 4 , a multi-attribute physical layer authentication device 4 includes:

[0186] A feature extraction module 41 is configured to extract a plurality of physical layer attributes of a received signal as a feature data set S(t) = (a ij ) n×m ; wherein n is the number of samples, m is the number of physical layer attributes, a ij represents the jth physical layer attribute of the ith sample in the feature data set;

[0187] A clustering and layering module 42 is configured to perform bottom-up layering on the feature data set by using the K-means clustering algorithm to obtain an l-layer data set;

[0188] A weighted clustering module 43 is configured to perform top-down weighted clustering on the l-layer data set to divide the feature data set into k clusters;

[0189] The quantity determining module 44 is configured to calculate average profile coefficients of the samples based on the k clusters, and determine the number K of cluster centers of the feature data set based on the average profile coefficients.

[0190] The center selecting module 45 is configured to determine the cluster centers of the feature data set based on the number K of cluster centers of the feature data set and the cluster centers of the k clusters.

[0191] The user determining module 46 is configured to determine a user type corresponding to each cluster center based on a distance between the cluster center of the feature data set and a reference vector, wherein the user type includes a legitimate user and a fraudulent user.

[0192] In a possible implementation, the clustering and layering module 42 is specifically configured to:

[0193] take the feature data set as an A-th layer data set, and let A = 1, wherein A is a count value;

[0194] cluster the A-th layer data set by using a K-means clustering algorithm to obtain M A clusters;

[0195] update a target function, wherein the target function is:

[0196]

[0197] wherein M A represents a number of clusters obtained by clustering the A-th layer data set, n A,k represents a number of samples in a k-th cluster obtained by clustering the A-th layer data set, represents a q-th sample in the k-th cluster obtained by clustering the A-th layer data set, c A,k represents a centroid of the k-th cluster obtained by clustering the A-th layer data set;

[0198] take the centroids of the M A clusters as an (A+1)-th layer data set;

[0199] let A+1, and jump to the step of "clustering the A-th layer data set by using a K-means clustering algorithm to obtain M A clusters", until F is less than a preset threshold A , or A = L, to obtain an L-th layer data set, wherein L is a preset maximum number of layers.

[0200] In a possible implementation, the multi-attribute physical layer authentication apparatus 4 further includes a standardization module 40.

[0201] The standardization module 40 is configured to:

[0202] Before the feature dataset is layered from bottom to top using the K-means clustering algorithm to obtain the l-layer dataset, each sample is standardized using the standardization formula to obtain the standardized matrix The standardization formula is:

[0203]

[0204]

[0205]

[0206] Among them, a ij represents the jth physical layer attribute of the i-th data sample in the feature dataset, represents a after normalization ij ;

[0207] The correlation coefficients between the physical layer attributes are calculated based on the correlation coefficient formula to obtain the correlation coefficient matrix R = (r ij ) m×m ; The correlation coefficient formula is:

[0208]

[0209] Among them, r ij represents the correlation coefficient between the i-th physical layer attribute and the j-th physical layer attribute, Indicates a ti and a tj covariance of

[0210] Calculate the eigenvectors of the correlation coefficient matrix and obtain m principal components based on the eigenvectors:

[0211]

[0212] Among them, y m represents the mth principal component, u nm represents the nth element in the mth eigenvector, Z n represents the nth vector in the normalized matrix;

[0213] The variance contribution rate of each principal component is calculated based on the variance contribution rate formula; the variance contribution rate formula is:

[0214]

[0215] Among them, b j represents the variance contribution rate of the jth principal component, λ k represents the eigenvalue of the kth principal component;

[0216] The cumulative variance contribution rate of the m principal components is calculated based on a cumulative variance contribution rate formula, and the number of selected principal components is determined based on the cumulative variance contribution rate. The cumulative variance contribution rate formula is:

[0217]

[0218] wherein, α p represents the cumulative variance contribution rate of the p principal components, and p represents the number of selected principal components;

[0219] The p principal components with the largest variance contribution rate are selected to obtain the de-correlated feature data set.

[0220] Correspondingly, the K-means clustering algorithm is used to perform bottom-up layering on the feature data set to obtain an l-layer data set, including:

[0221] The K-means clustering algorithm is used to perform bottom-up layering on the de-correlated feature data set to obtain an l-layer data set.

[0222] In a possible implementation, the l-layer data set is weighted clustered from top to bottom to divide the feature data set into k clusters, including:

[0223] Let B = l; wherein, B is a count value;

[0224] Randomly select k data in the B-layer data set as the centroids;

[0225] Each sample in the B-layer data set is assigned to the nearest centroid to obtain k clusters, and the centroids are updated based on the current k clusters;

[0226] B = B-1, and jump to the step of “assigning each sample in the B-layer data set to the nearest centroid to obtain k clusters, and updating the centroids based on the current k clusters”, until B = 1, and finally obtain the k clusters.

[0227] In a possible implementation, the number determination module 44 is specifically configured to:

[0228] Let C = k; wherein, C is a count value;

[0229] Calculate the intra-class cohesion and inter-class separation of each sample in the C clusters;

[0230] Calculate the silhouette coefficient of each sample based on the intra-class cohesion and the inter-class separation;

[0231] Calculate the average silhouette coefficient corresponding to C based on the silhouette coefficient of each sample;

[0232] C-1, and jump to the step of calculating the intra-class cohesion and inter-class separation of each sample in C clusters until C=2, to obtain k-1 average silhouette coefficients;

[0233] The C corresponding to the average silhouette coefficient closest to 1 is taken as the cluster center number K of the feature data set.

[0234] In a possible implementation, the center selection module 45 is specifically configured to:

[0235] Two centroid centers with the closest Euclidean distance are selected from the centroid centers of the clusters, a mean center point of the two centroid centers is calculated as a new centroid center, and the two centroid centers are removed, and the step is repeated until the number of centroid centers is equal to the cluster center number K of the feature data set, to obtain the cluster centers of the feature data set.

[0236] In a possible implementation, the user judgment module 46 is specifically configured to:

[0237] The distance between each cluster center and the reference vector is calculated.

[0238] The cluster center closest to the reference vector is determined as a legitimate sample, and the other cluster centers are determined as fraudulent samples.

[0239] The multi-attribute physical layer authentication device provided by the embodiment of the application comprises: a feature extraction module configured to extract a plurality of physical layer attributes of a received signal as a feature data set S(t) = (a ij ) n×m ; wherein n is the number of samples, m is the number of physical layer attributes, a ij represents the jth physical layer attribute of the ith data sample in the feature data set; a clustering and layering module configured to perform bottom-up layering on the feature data set by using a K-means clustering algorithm to obtain an l-layer data set; a weighted clustering module configured to perform top-down weighted clustering on the l-layer data set to divide the feature data set into k clusters; a number determination module configured to iteratively calculate the average silhouette coefficient of the samples based on the k clusters, and determine the cluster center number K of the feature data set based on the average silhouette coefficient; a center selection module configured to determine the cluster centers of the feature data set based on the cluster center number K of the feature data set and the cluster centers of the k clusters; and a user judgment module configured to determine the user type corresponding to each cluster center based on the distance between the cluster center of the feature data set and a reference vector; wherein the user type comprises a legitimate user and a fraudulent user. The embodiment performs hierarchical clustering on the feature data set, can find high-quality cluster centers without using training samples, and thus reduces the complexity of the preliminary preparation work; meanwhile, the number of cluster centers is adjusted in the case of insufficient prior knowledge by using the silhouette coefficient, and thus the adaptability of the physical layer authentication is improved.

[0240] Figure 5 is a schematic diagram of a terminal provided by an embodiment of the present application. As shown in Figure 5 , the terminal 5 of this embodiment comprises a processor 50, a memory 51, and a computer program 52 stored in the memory 51 and executable on the processor 50. The processor 50 implements the steps in each of the above multi-attribute physical layer authentication method embodiments when executing the computer program 52, such as steps 201 to 206 shown in Figure 2 . Alternatively, the processor 50 implements the functions of each module / unit in each of the above device embodiments when executing the computer program 52, such as the functions of modules / units 40 to 46 shown in Figure 4 .

[0241] By way of example, the computer program 52 can be segmented into one or more modules / units stored in the memory 51 and executed by the processor 50 to complete the present application. The one or more modules / units can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program 52 in the terminal 5. For example, the computer program 52 can be segmented into modules / units 40 to 43 shown in Figure 4 .

[0242] The terminal 5 can be a desktop computer, a notebook computer, a palm computer, a cloud server, and other computing devices. The terminal 5 can include, but is not limited to, the processor 50 and the memory 51. Those skilled in the art can understand that Figure 5 the terminal 5 is merely an example and does not constitute a limitation on the terminal 5, and can include more or fewer components than those shown, or combine certain components, or different components, for example, the terminal can also include an input / output device, a network access device, a bus, etc.

[0243] The processor 50 can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic components, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0244] The memory 51 can be an internal storage unit of the terminal 5, such as a hard disk or a memory of the terminal 5. The memory 51 can also be an external storage device of the terminal 5, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, and the like equipped on the terminal 5. Further, the memory 51 can also include both the internal storage unit and the external storage device of the terminal 5. The memory 51 is used to store the computer program and other programs and data required by the terminal. The memory 51 can also be used to temporarily store data that has been output or is to be output.

[0245] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above functional units and modules is exemplified, and in actual application, the above functions can be completed by different functional units and modules according to needs, that is, the internal structure of the apparatus is divided into different functional units or modules to complete all or part of the functions described above. Each functional unit and module in the embodiment can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit, and the integrated unit can be realized in the form of hardware or in the form of software functional unit. In addition, the specific names of each functional unit and module are only for the convenience of mutual distinction, and do not limit the protection scope of the present application. The specific working process of the units and modules in the above system can refer to the corresponding process in the foregoing method embodiments, which will not be described here.

[0246] In the above embodiments, the description of each embodiment has its own emphasis, and the parts not described or recorded in detail in a certain embodiment can refer to the relevant description of other embodiments.

[0247] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be realized in electronic hardware or in combination of computer software and electronic hardware. Whether the functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0248] In the embodiments of the present application, it should be understood that the disclosed apparatus / terminal and method can be implemented in other manners. For example, the embodiments of the apparatus / terminal described above are merely schematic, and the division of the modules or units is merely logical function division, and there can be another division manner in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between the units can be indirect couplings or communication connections through some interfaces, devices or units, and can be electrical, mechanical or in other forms.

[0249] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0250] In addition, each functional unit in the various embodiments of the present application can be integrated into a processing unit, or each unit can be a physically independent unit, or two or more units can be integrated into a unit. The integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0251] The integrated module / unit, if implemented in the form of a software functional unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, all or part of the flow of the above-mentioned embodiment methods can be completed by a computer program instructing related hardware, and the computer program can be stored in a computer readable storage medium. When the processor executes the computer program, the steps of each multi-attribute physical layer authentication method embodiment can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable file or some intermediate form. The computer readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the computer readable medium can include appropriate contents according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer readable medium does not include electrical carrier signals and telecommunication signals.

[0252] The above-described embodiments are only used to illustrate the technical solutions of the present application, but not limit the present application; although the present application has been described in detail with reference to the foregoing embodiments, those ordinarily skilled in the art should understand: the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced equivalently; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A multi-attribute physical layer authentication method, characterized in that: include: Extract multiple physical layer attributes of the received signal as feature data sets ;in, is the sample size, is the number of physical layer attributes, Indicates the first The first sample physical layer attributes; The feature data set is layered from bottom to top using the K-means clustering algorithm to obtain Layer dataset; Regarding the The layer data set is weighted clustered from top to bottom, and the feature data set is divided into clusters; Based on the Iteratively calculating the average silhouette coefficient of the samples for each cluster, and determining the number K of cluster centers of the feature data set based on the average silhouette coefficient; Based on the number of cluster centers K of the feature data set and the The cluster centers of the clusters determine the cluster centers of the feature data set; Determining the user type corresponding to each cluster center based on the distance between the cluster center of the feature data set and the reference vector; wherein the user type includes legitimate users and fraudulent users; The feature data set is layered from bottom to top using the K-means clustering algorithm to obtain Layer datasets, including: The feature dataset is taken as Layer dataset, and let ;in, is the count value; The K-means clustering algorithm is used to Clustering of the layer data set, we get clusters; Update the objective function; the objective function is: in, Indicates the The number of clusters obtained by clustering the layer data set, Indicates the The first layer of data sets is clustered The number of samples in a cluster, Indicates the The first layer of data sets is clustered The first in the cluster samples, Indicates the The first layer of data sets is clustered The centroid of each cluster; The The centroid of the cluster is the Layer dataset; Will , and jump to "Use K-means clustering algorithm to cluster the first Clustering of the layer data set, we get Clusters" step until Less than the preset threshold ,or When Layer dataset; where The preset maximum number of layers.

2. The multi-attribute physical layer authentication method according to claim 1, wherein: The feature data set is layered from bottom to top using the K-means clustering algorithm to obtain Before layering the data set, the method further includes: Each sample is standardized by the standardization formula to obtain the standardized matrix ; The standardization formula is: in, Indicates the first The first data sample Physical layer attributes, Indicates the standardized ; Based on the correlation coefficient formula, the correlation coefficient between each physical layer attribute is calculated to obtain the correlation coefficient matrix ; The correlation coefficient formula is: in, Indicates the physical layer attributes and The correlation coefficient of the physical layer attributes, express and covariance of Calculate the eigenvector of the correlation coefficient matrix and obtain Principal components: in, Indicates the principal components, Indicates the The first feature vector elements, Indicates the first vectors; The variance contribution rate of each principal component is calculated based on the variance contribution rate formula; the variance contribution rate formula is: in, Indicates the The variance contribution of the principal components, Indicates the The eigenvalues ​​of the principal components; The cumulative variance contribution rate formula is used to calculate the The cumulative variance contribution rate of the principal components is calculated, and the number of principal components to be selected is determined based on the cumulative variance contribution rate; the cumulative variance contribution rate formula is: in, express The cumulative variance contribution rate of the principal components, Indicates the number of principal components selected; Select The principal component with the largest variance contribution rate is obtained to obtain the feature data set after decorrelation; Accordingly, the feature data set is layered from bottom to top using the K-means clustering algorithm to obtain Layer datasets, including: The decorrelated feature dataset is layered from bottom to top using the K-means clustering algorithm to obtain Layer dataset.

3. The multi-attribute physical layer authentication method according to claim 1, wherein: The above mentioned The layer data set is weighted clustered from top to bottom, and the feature data set is divided into clusters, including: make ;in, is the count value; In the Randomly select from the layer data set The data is used as the centroid; The first Each sample in the layer data set is assigned to the centroid closest to the sample, and the k clusters, and based on the current Update the centroid of each cluster; Will , and jump to " Each sample in the layer data set is assigned to the centroid closest to the sample, and the k clusters, and based on the current Cluster update centroid" step until When the final clusters.

4. The multi-attribute physical layer authentication method according to claim 1, wherein: Based on the Iteratively calculating the average silhouette coefficient of the samples for each cluster, and determining the number K of cluster centers of the feature data set based on the average silhouette coefficient, including: make ;in, is the count value; calculate The intra-class cohesion and inter-class separation of each sample in a cluster; Calculate the silhouette coefficient of each sample based on the cohesion within each class and the separation between each class; Calculation of silhouette coefficient based on each sample The corresponding average silhouette coefficient; Will , and jump to "Calculation The step of "the cohesion within each cluster and the separation between each cluster" is repeated until When The average silhouette coefficient; The average silhouette coefficient closest to 1 corresponds to The number K of cluster centers of the feature data set.

5. The multi-attribute physical layer authentication method according to claim 1, wherein: The number of cluster centers K based on the feature data set and the The cluster centers of the clusters are determined to determine the cluster centers of the feature data set, including: Select two centroids with the closest Euclidean distance from the centroids of each cluster, calculate the mean center point of the two centroids as the new centroids, and remove the two centroids. Repeat this step until the number of centroids is equal to the number K of cluster centers of the feature data set, and obtain the cluster center of the feature data set.

6. The multi-attribute physical layer authentication method according to any one of claims 1 to 5, characterized in that: The determining of the user type corresponding to each cluster center based on the distance between the cluster center of the feature data set and the reference vector includes: Calculate the distance between each cluster center and the reference vector; The cluster center closest to the reference vector is determined as a legitimate sample, and the other cluster centers are determined as fraudulent samples.

7. A multi-attribute physical layer authentication device, characterized in that: include: Feature extraction module, used to extract multiple physical layer attributes of the received signal as feature data sets ;in, is the sample size, is the number of physical layer attributes, Indicates the first The first data sample physical layer attributes; The clustering and stratification module is used to stratify the feature data set from bottom to top using the K-means clustering algorithm to obtain Layer dataset; Weighted clustering module for the The layer data set is weighted clustered from top to bottom, and the feature data set is divided into clusters; A quantity determination module is used to determine the quantity of Iteratively calculating the average silhouette coefficient of the samples for each cluster, and determining the number K of cluster centers of the feature data set based on the average silhouette coefficient; Center selection module, for clustering center number K based on the feature data set and the The cluster centers of the clusters determine the cluster centers of the feature data set; A user judgment module, configured to judge the user type corresponding to each cluster center based on the distance between the cluster center of the feature data set and the reference vector; wherein the user type includes a legitimate user and a fraudulent user; The clustering and stratification module is specifically used for: The feature dataset is taken as Layer dataset, and let ;in, is the count value; The K-means clustering algorithm is used to Clustering of the layer data set, we get clusters; Update the objective function; the objective function is: in, Indicates the The number of clusters obtained by clustering the layer data set, Indicates the The first layer of data sets is clustered The number of samples in a cluster, Indicates the The first layer of data sets is clustered The first in the cluster samples, Indicates the The first layer of data sets is clustered The centroid of each cluster; The The centroid of the cluster is the Layer dataset; Will , and jump to "Use K-means clustering algorithm to cluster the first Clustering of the layer data set, we get Clusters" step until Less than the preset threshold ,or When Layer dataset; where The preset maximum number of layers.

8. A terminal comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images