Large-scale access pilot pollution attack confrontation method based on channel decontamination and communication system

Abstract

The invention discloses a channel decontamination-based large-scale access pilot pollution attack confrontation method and a communication system. The method comprises the following steps of: constructing an uplink multiple-input multiple-output communication model based on a millimeter wave authorization-free mMTC network, and constructing a three-dimensional transmission model and a channel model in an angle-time delay-time slot domain; based on the constructed three-dimensional model, converting a PCA confrontation problem into a multi-dimensional sparse signal recovery problem based on a multi-dimensional dictionary; designing a sparse adaptive matching pursuit algorithm based on multiple dictionaries to recover a CVR matrix by using the sparsity of a millimeter wave channel in a virtual angle domain and a time delay domain and the time correlation of a legal LU active state; and finally, channel reconstruction of all LUs is realized according to activity estimation, PCA detection and channel decontamination criteria. According to the method, the accuracy of user channel estimation and data decoding is improved, the performance of resisting pilot pollution attacks is improved, and the communication security is ensured.

Description

Technical Field

[0001] The present invention belongs to the field of random access and data transmission in a large-scale access network, and in particular relates to a large-scale access pilot pollution attack countermeasure method based on channel decontamination and a communication system. Background Art

[0002] The unlicensed access mechanism can avoid a large amount of signaling overhead and access delay generated by devices and base stations during the random access process, providing efficient access technology support for massive Machine Type Communications (mMTC) networks and improving communication efficiency. Under the unlicensed access mechanism, devices can send data directly to base stations without waiting for base station authorization. Since the mechanism lacks identity authentication for access devices and only distinguishes users by identifying the received pilot sequence, the mMTC network is vulnerable to pilot contamination attack (PCA). PCA refers to an attacker sending the same pilot sequence as a legitimate user (LU) to impersonate the LU, thereby accessing the base station and polluting the LU's channel information with noise, thereby deteriorating the base station's activity detection of legitimate users, channel estimation, and the accuracy of subsequent data reception. How to design a PCA countermeasure solution for unlicensed mMTC networks based on channel decontamination is a technical problem that needs to be solved.

[0003] Among the existing studies on channel decontamination and PCA confrontation, most of them are aimed at PCA confrontation scenarios under single-user and single-attacker conditions. For example, the Chinese patent with publication number CN109905203A provides a collaborative vector security transmission solution to resist PCA, which counters PCA through collaborative fusion detection between the receiver and the third-party trusted node to achieve secure transmission. However, in the unlicensed mMTC network, there are a large number of legitimate users and potential attackers at the same time. The base station needs to identify the legitimate users and attackers who are in the signal sending state at this time among many users, that is, complete the user activity detection, and then perform subsequent attack detection; at the same time, the implementation of this solution relies on the collaboration of third-party trusted nodes. In the unlicensed mMTC network, there are high requirements for the ability of third-party nodes to process a large number of user signals at the same time. This makes the solution less applicable in the unlicensed mMTC network scenario. Therefore, in the unlicensed mMTC network, it is challenging to design a PCA confrontation solution that can simultaneously identify legitimate user activity detection, PCA detection, and channel decontamination. Summary of the invention

[0004] In order to solve the problems existing in the prior art, the present invention provides a method for countering large-scale access pilot pollution attacks based on channel decontamination. By utilizing the joint characteristics of the millimeter wave channel in the angle-delay-time slot domain, the legitimacy of the access device is judged, and the channel of the attacked LU is decontaminated and reconstructed to counter PCA.

[0005] In order to achieve the above object, in a first aspect, the present invention provides a method for countering a large-scale access pilot pollution attack based on channel decontamination, comprising the following steps:

[0006] A millimeter-wave uplink multi-input multi-output system model is built based on the unlicensed mMTC network. The transmission characteristics of legitimate users and attackers in the angle-delay-time slot domain are jointly considered to construct a three-dimensional sparse CVR matrix H and a three-dimensional received signal Y. A three-dimensional transmission model with sparsity in the angle-delay domain and time correlation in the time slot domain is obtained.

[0007] Based on the constructed three-dimensional transmission model, the channel decontamination problem against pilot contamination attacks can be transformed into the following multi-dimensional sparse signal recovery problem:

[0008]

[0009] Where t represents the time slot index, T represents the total number of time slots included in one transmission, and O {：，t} =vec(Y {：，:，t} ) represents the vectorized observation signal, Φ represents the measurement tensor, Ψ k Represents a dictionary, and A set of atomic indices representing dictionaries, Where i1 and i2 represent the column index and row index of the dictionary Ψ1, j1 and j2 represent the column index and row index of the dictionary Ψ2, N represents the total number of users in the system, D represents the number of channel delay taps, M represents the number of base station antennas, and R represents the total number of users in the system. delay , R AoA represents the number of delays and angles in the dictionary; H represents the estimated CVR matrix, H {j，t} represents the value of the estimated CVR matrix indexed at time slot t, j;

[0010] The multidimensional sparse signal recovery problem is solved by the sparse adaptive matching pursuit algorithm based on the multidimensional dictionary, and the estimation result of the CVR matrix H is obtained.

[0011] Estimation results based on CVR matrix H Combine user activity detection, PCA detection and Xinda decontamination criterion to achieve PCA confrontation and obtain the estimated set of legitimate active users PCA attacker collection The legitimate user channel matrix after decontamination reconstruction

[0012] Furthermore, the transmission system includes a base station with M antennas, N LUs with single antennas, and multiple PCA attackers. Assume that each LU generates sporadic traffic and has a fixed position, and the PCA attacker can randomly launch attacks and has a variable position. The LU index set is represented as The base station antenna index set is expressed as Each LU is assigned a pilot sequence s n , with a length of C; assume that multiple PCA attackers attack by sending the same pilot sequence as the active LU to pollute the channel information of the LU; the pilot signals received by the base station from all active LUs and PCA attackers in time slot t are represented by the channel model Y t =G t S+G′ t S+W t Indicates that G t and G′ t They represent the channel matrices of LU and attacker, S represents the pilot matrix, and W t Indicates noise.

[0013] Furthermore, the channel model Y t =G t S+G′ t S+W t In the above example, the channel matrix is ​​expressed as The two-dimensional indicator variable and Respectively represent the activation state and the attacked state of LUn, and They represent the channel matrix of LUn in time slot t and the channel matrix of the attacker attacking LUn in time slot t respectively; the channel information is composed of D channel taps, that is, Represents the channel information at tap d, using the geometric channel model Indicates that, where L represents the number of scatterers, represents the channel gain of the lth path of LUn in time slot t, θ n，l represents the physical AoA of the lth path, a(θ n，l ) represents the antenna array response, τ n，l represents the delay of path l, T s represents the sampling period, f(τ) represents the sampling pulse; CVR is introduced to describe the sparse characteristics of the millimeter wave channel, then Expressed as Where B represents the virtual AoA dictionary, f[d] represents the delay dictionary, represents the CVR matrix of LUn in time slot t. The CVR matrix has sparse characteristics in the angle-delay domain. Correspondingly, for an attacker attacking LUn, the channel information Expressed as in, Denotes the CVR matrix of an attacker attacking LUn at time slot t.

[0014] Furthermore, a three-dimensional transmission model Y consisting of multiple time slots is constructed {：，：，t} =BH {：，：，t} FS+W t , t = 1, ..., T, where Y {：，：，t} =Y t represents the received signal at time slot t, represents the unified CVR matrix of time slot t, where represents the CVR of LUn in the attacked, safe, and idle states at time slot t, where and Represents the set of LUs that are active and attacked in time slot t; the activation state of the LU remains unchanged within T time slots. For user LUn in a safe state, the CVR has time correlation.

[0015] Furthermore, a sparse adaptive matching pursuit algorithm based on a multidimensional dictionary is used to solve the multidimensional sparse signal recovery problem, including: estimating the CVR matrix under fixed sparsity through matching projection and residual update steps, and updating the sparsity through iterative adaptive updating to obtain the final estimation result closest to the true CVR matrix H Among them, the matching projection step is expressed as:

[0016]

[0017] Among them, O res Represents the residual observation component; the residual update steps are:

[0018]

[0019] Where Γ represents the currently estimated matrix support set, and Represents the tensor Φ and the current estimation result The deformation matrix, is the Kronecker product of dictionaries.

[0020] Furthermore, when updating the sparsity through iterative adaptive updating, at a fixed sparsity Signal recovery is performed based on the output results of the inner layer. to update.

[0021] Furthermore, based on the estimation results of the CVR matrix H Implementing joint user activity detection, PCA detection and Xinda decontamination includes the following steps:

[0022] According to the estimated CVR matrix of each user The power of is used to detect user activity and obtain the estimated LU activity indicator vector:

[0023]

[0024] in, Represents the power threshold, and obtains the estimated active LU set

[0025] Perform PCA detection to screen the effective path set in, represents the estimated CVR of LUn at time slot t, Represents the power threshold, and the estimated LU attack state indicator vector is obtained:

[0026]

[0027] And get the estimated attacked LU set

[0028] Channel decontamination is performed, and the time correlation of the LU active state is used to decontaminate and reconstruct the CVR matrix of LUn in time slot t to obtain the decontaminated CVR matrix of LUn in time slot t:

[0029]

[0030] in, based on Get the decontaminated CVR matrix of all N LUs in time slot t And the channel information matrix

[0031] In a second aspect, the present invention can provide a communication system, including a base station with M antennas, N LUs with single antennas, and multiple PCA attackers; each LU generates sporadic traffic and has a fixed position, and the PCA attacker can randomly launch attacks and has a variable position; the LU index set is represented as The base station antenna index set is expressed as Each LU is assigned a pilot sequence s n , with a length of C; assume that multiple PCA attackers attack by sending the same pilot sequence as the active LU to pollute the channel information of the LU; the pilot signals received by the base station from all active LUs and PCA attackers in time slot t are represented by the channel model Y t =G t S+G′ tS+W t Indicates that G t and G′ t They represent the channel matrices of LU and attacker, S represents the pilot matrix, and W t Represents noise; communication is performed based on the above-mentioned large-scale access pilot pollution attack countermeasure method based on channel decontamination.

[0032] The present invention can also provide a computer device, including a processor and a memory, the memory is used to store a computer executable program, the processor reads the computer executable program from the memory and executes it, and when the processor executes the computer executable program, it can implement the above-mentioned large-scale access pilot pollution attack countermeasure method based on channel decontamination.

[0033] A computer-readable storage medium is also provided, in which a computer program is stored. When the computer program is executed by a processor, the above-mentioned large-scale access pilot pollution attack countermeasure method based on channel decontamination can be implemented.

[0034] Compared with the prior art, the present invention has at least the following beneficial effects: the present invention establishes a three-dimensional channel model in the angle-delay-time slot domain of the millimeter wave unlicensed mMTC network, and utilizes the sparsity of the millimeter wave channel in the virtual angle domain and the delay domain and the time correlation of the LU activity; secondly, based on the established model, a PCA resistance method based on channel decontamination is proposed, which can jointly perform user activity identification, PCA detection and channel decontamination to ensure the security of mMTC network communication; it can effectively resist large-scale access pilot pollution attacks to ensure the security of mMTC network communication; through sparse signal processing and multi-dimensional dictionary matching, wireless resources can be used more efficiently, and through channel decontamination and PCA confrontation, the system's robustness to malicious attacks is enhanced, and the stability of the network is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 The invention provides a millimeter wave unlicensed mMTC network angle-delay-time slot three-dimensional channel model.

[0036] Figure 2 The graph is a curve showing the change of the normalized mean square error of channel estimation with the signal-to-noise ratio under the method of the present invention and the comparative scheme.

[0037] Figure 3 The graphs are the variation curves of the data decoding bit error rate with the signal-to-noise ratio under the method of the present invention and the comparative scheme. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0039] Embodiment 1, the present invention provides a method for countering large-scale access pilot pollution attacks based on channel decontamination, comprising the following steps: the system constructs a millimeter wave multi-input multi-output system for an unlicensed mMTC network, the system comprising a base station with M antennas, N single-antenna LUs, and multiple PCA attackers. Assume that the LU generates sporadic traffic and has a fixed position, while the attacker's position is variable. Let and Represents the LU index set and the base station antenna index set.

[0040] Considering the geometric channel model with L scatterers, the channel between the base station and LUn at time slot t is expressed as:

[0041]

[0042] in represents the channel gain of the lth path of LUn in time slot t, θ n，l represents the physical angle of arrival (AoA) of the lth path, a(θ n，l ) represents the antenna array response, τ n，l represents the delay of path l, T s represents the sampling period, and f(τ) represents the sampling pulse. Assuming that the position of the scatterer remains unchanged, then θ n，l Also remain fixed. Due to the high path loss and blocking sensitivity of mmWave communication, the mmWave channel is sparse in both the angular domain and the delay domain. By introducing the channel virtual representation (CVR) to describe the mmWave channel, the actual physical channel is mapped to the virtual AoA domain and the delay domain. Define the virtual AoA dictionary With Delay Dictionary Where R AoA and R delay represents the number of quantized virtual AoA and delay, φ r and v r′ represents the rth discrete virtual AoA and the r′th discrete delay. Channel and its CVR The relationship is expressed as:

[0043]

[0044] Since there are only L paths in the channel model, when R AoA R delay >> When L, CVR matrix It has coefficient property, i.e., angle-delay domain sparsity.

[0045] In the mmWave unlicensed mMTC network, each LU is assigned a pilot sequence. The length is C. Assume that multiple PCA attackers launch attacks by sending the same pilot sequence as LU. The signals received by the base station from all active LUs and PCA attackers at time slot t are expressed as:

[0046]

[0047] in and represents the set of active LUs and attacked LUs in time slot t, represents the channel of LUn in time slot t, represents the channel of the attacker attacking LUn′ at time slot t, represents the pilot signal of LUn at D time delays, s n [d] means s n The d-th displacement, represents noise, G t and G′ t represents the channel matrix of all active LUs and all attackers in time slot t. Define the activity indicator variable and attack status indicator variable in or 0 means that LUn is active or inactive in time slot t, Or 0 indicates whether LUn is attacked by a PCA attacker in time slot t. Therefore, the active LU set of time slot t can be obtained The set of attacked LUs in time slot t The channel matrix of all active LUs in slot t Channel matrix of all attackers

[0048] In order to disrupt the communication between the base station and the LU, the PCA attacker may monitor the channel state of the LU and launch an attack when the LU is active. In this way, the base station will estimate the channel state information of the active LU based on the contaminated pilot signal, which will seriously reduce the accuracy of the channel estimation, and the base station may not be able to correctly decode the data sent by the active LU, resulting in communication failure. Therefore, assuming that the attacker attacks the LU in the active state, that is, Define active and unattacked LUns The state of LU is defined as "safe" state, then in each time slot, the state of LU may be any of "inactive", "safe" and "attacked". In addition, since LU generates sporadic business, the collection and are very small, and the channel matrix G t and G′ t is a sparse matrix.

[0049] Angle-Delay-Time Slot Three-Dimensional Millimeter Wave Channel Model

[0050] Defining a unified CVR matrix To represent the CVR of LUn in different states at time slot t. The specific definition is:

[0051]

[0052] in, represents the CVR matrix of the attacker who attacks LUn in time slot t. Consider that each transmission consists of T consecutive time slots, and the CVR matrix of T time slots is Combined, a unified CVR three-dimensional matrix can be obtained in Three-dimensional matrix H n It has sparseness in both virtual angle domain and time delay domain, such as Figure 1 As shown. In addition, in actual mMTC networks, user activities may show implementation-related characteristics, and active users will transmit in adjacent time slots with high probability. Therefore, the temporal correlation of user activity is considered by assuming that the activity of each LU remains unchanged in consecutive T time slots, that is, Since the positions of all LUs and scatterers are fixed, the physical AoA and virtual AoA of the LUs also remain constant. In addition, it is observed that the path delay of the mmWave channel varies much slower than the path gain, even though the path gain The path delay τ varies greatly between different time slots, but the path delay τ between consecutive time slots n，l Therefore, when LUn is in a safe state, that is, like Figure 1 LUn2 in the time slots, we can get the CVR matrix of LUn in T time slots. have the same support set, i.e.

[0053]

[0054] in, express The support set of is the set of non-zero element indices. However, since the attacker randomly initiates PCA in each time slot and the attacker's position is variable, the attacked LU will not show time correlation. That is, for the LU in the attacked state (like Figure 1 LUn3), Varies with t.

[0055] Based on the unified CVR matrix, the received signal Y t It is expressed as:

[0056] Y t =BH t FS+W t

[0057] in, represents the delay dictionary matrix, The received signal Y of T time slots t Combined, we get a three-dimensional received signal where Y {：，：，t} =Y t . And convert the single-slot transmission model to the following three-dimensional transmission model:

[0058] Y {：，：，t} =BH {：，：，t} FS+W t , t=1,…,T

[0059] Among them, H {：，：，t} =H t Based on the constructed three-dimensional transmission model, the PCA adversarial problem can be converted into a sparse CVR matrix In order to resist PCA, the channel of the attacked LU needs to be decontaminated and reconstructed. Because at this time You need to eliminate the attacker's And reply true Most of the existing solutions only utilize the sparsity of the millimeter wave channel in the virtual angular domain, but in the virtual angular domain, the CVRs of the attacker and the attacked LU are likely to overlap, that is, and Therefore, the present invention expands the sparsity of the millimeter wave channel in the angle-delay-time slot domain, as shown in the figure, to enhance This will significantly reduce the CVR overlap probability between the attacker and the LU, thereby effectively improving the performance of channel reconstruction.

[0060] Constructing the multidimensional sparse signal recovery problem

[0061] According to the constructed three-dimensional transmission model, the three-dimensional sparse matrix H can be represented by two independent dictionaries, namely the virtual AoA dictionary B and the delay dictionary F, and the pilot matrix S serves as the measurement matrix. Therefore, the PCA adversarial problem can be expressed as a multidimensional sparse recovery problem based on multidimensional dictionaries. Specifically, the dictionary set {Ψ k} k＝1，2 ,in The observation matrix is ​​defined as Among them, {：，t} =vec(Y {：，：，t} ), vec(·) represents the column vectorization of the matrix. In the case of multiple dimensions, the measurement matrix can be constructed in the form of a tensor, expressed as To obtain Φ, first define the matrix And Defining an index collection and represents {Ψ k} k＝1，2 The row index and column index of the dictionary atom in . Furthermore, the multidimensional sparse signal recovery problem can be formulated as:

[0062]

[0063] When the unified CVR matrix H is recovered by the above formula, channel decontamination can be achieved by eliminating the CVRX′ composed of the channel components of all attackers and reconstructing the true CVRX of all LUs.

[0064] The multidimensional sparse signal recovery problem is solved by a multidimensional dictionary-based sparsity adaptive matching pursuit algorithm. In order to solve the constructed multidimensional sparse signal recovery problem, a multidimensional dictionary based sparsity adaptive matching pursuit (MD-SAMP) algorithm is designed. The algorithm is designed based on the multidimensional orthogonal matching pursuit (MOMP) algorithm, which can effectively utilize the multidimensional sparse structure of the CVR matrix H to solve the problem. The main process of the MD-SAMP algorithm consists of two iterative parts: matching projection and residual update. In each iteration, the algorithm calculates the inner product, i.e., projection, of the dictionary atom and the residual observation component to select the atom with the maximum projection value. Then, the residual projection component is updated based on the screened atomic projection.

[0065] Specifically, the matching projection steps of the constructed multidimensional problem are:

[0066]

[0067] Among them, res represents the residual observation component. This maximization step requires calculating all possible indices of the dictionary atoms This will result in a relatively high computational complexity. To reduce the complexity, the optimization of index j=(j1, j2) can be transformed into a historical solution based on Iterative optimization process, in which j1 and j2 are optimized separately. In this way, the optimization problem can be transformed into

[0068]

[0069] Where k, k′=1, 2, k≠k′. After the matching projection is completed, the selected atom index j=(j1, j2) is added to the estimated support set Γ of H. Then, the residual observation component can be updated according to the following formula:

[0070]

[0071] in is the estimated result of H, which can be obtained by least squares estimation, that is, represents the matrix inversion, and Represents Φ and The deformation, Represents the Kronecker product of dictionaries.

[0072] In the MOMP algorithm, the above iterative process is repeated until the size of the estimated support set Γ is equal to the sparsity of the matrix H, that is, the size of the support set supp(H). However, in actual scenarios, the size of supp(H), that is, the number of channel paths of all active LUs and attackers, cannot be perfectly known, which limits the practical application of the MOMP algorithm. In order to solve this problem, the designed MD-SAMP algorithm performs two nested iterative processes to solve the sparse signal recovery problem by adaptively determining the sparsity. The algorithm process is shown in Table 1. The inner loop is fixed at the sparsity. The signal is restored under the condition that the outer loop adjusts the sparsity based on the output result of the inner loop. to update.

[0073] Let l represent the iteration index of the inner loop. In the lth inner loop, the support set is first estimated, as shown in the fifth row of Table 1. By matching the projection, the support set Select atomic index, and get the set Γ l , and the (l-1)th iteration result Γ l-1 With Γ lThen, we get the set Λ. Then, we perform the support set correction step, as shown in the sixth row of Table 1. We select s indexes from Λ so that their corresponding projection values ​​are the largest, forming the set Among them, Ξ(·, s) represents the largest value among the selected parameters. An indexed collection of elements. The signal estimation and residual update steps are performed, see the seventh and eighth rows of Table 1. After that, the iterative index is updated and the (l+1)th inner loop begins.

[0074] When the termination criteria After satisfying, we get the sparsity The support set below and the estimated CVR matrix Next, the sparsity update step is performed in the outer loop, see row 11 of Table 1, and the sparsity is estimated Finally, when the termination criterion When satisfied, this means that the sparsity has been overestimated, where Represents the background noise power. Therefore, the entire iteration process has been completed at this point, and the reconstructed CVR matrix is ​​obtained

[0075] Table 1 MD-SAMP algorithm

[0076]

[0077] Combined user activity detection, PCA detection and channel decontamination

[0078] Output results based on MD-SAMP algorithm The PCA confrontation problem can be solved by combining user activity detection, PCA detection and channel decontamination. First, user activity detection is performed. The power of , we get the estimated activity indicator variable:

[0079]

[0080] in, represents the CVR power threshold of the active LU. Then, the estimated active LU set can be obtained Secondly, PCA detection is performed. Since the real channel of each LU only contains L paths, if the attacker attacks LUn in time slot t, the number of paths will be greater than L. Therefore, the valid path set is first screened out by the path power size. in represents the power threshold of a single path in CVR. Then, the estimated attack status indicator variable is determined by the following formula:

[0081]

[0082] Where |·| represents the number of elements in the set, and the estimated attacked LU set can be obtained Finally, channel decontamination is performed. Since the CVR of the active LU has time correlation, while the CVR of the attacker does not satisfy this characteristic, we can compare the path sets in T time slots. To distinguish the contaminated coefficients in the CVR of the attacked LU. Specifically, first obtain the path intersection of LUn in T time slots Secondly, from n Select L paths with the largest power as the legal paths of LUn. For the remaining paths, that is, It is considered as the contaminated path brought by the attacker and needs to be eliminated. Therefore, the channel decontamination of LUn can be expressed as:

[0083]

[0084] in represents the CVR matrix of LUn after decontamination at time slot t. The decontamination of all LUs in T time slots can be obtained The channel coefficients estimated by LU can be further expressed by get.

[0085] The present invention evaluates the system performance of the proposed solution through simulation, and the system parameter settings are shown in Table 2:

[0086] Table 2 Simulation parameters

[0087]

[0088]

[0089] In order to measure the performance of the designed scheme, comparative experiments were conducted with the following compressed sensing algorithms: orthogonal matching pursuit (OMP) algorithm, block sparsity adaptive subspace pursuit (BSASP) algorithm, MOMP algorithm, and Oracle-LS algorithm. Among them, OMP and BSASP algorithms are classic greedy pursuit algorithms for single-dimensional sparse signal recovery, which only recover the CVR matrix in the angle domain; MOMP algorithm and the designed MD-SAMP algorithm both recover CVR in the angle-delay-time slot domain, but MOMP algorithm can only be implemented under the premise of known sparsity; Oracle-LS algorithm provides a performance upper bound for other algorithms. This algorithm assumes that the real active LU set and the attacked LU set, as well as the path positions of all users are perfectly known by the base station. In addition, in order to evaluate the impact of PCA on channel estimation performance and data reception performance, experimental results of the designed scheme without PCA detection and channel decontamination are given.

[0090] Figure 2 , Figure 3 The performance simulation results of the normalized mean square error of the CVR matrix recovery of all LUs and the bit error rate of the data sent by the active LU and decoded by the base station as the signal-to-noise ratio changes are shown. The normalized mean square error is defined as where X and represents the true CVR and estimated CVR of all LUs in time slot T; the bit error rate is defined as where b n，t and Represents the data bits sent by LUn and the bits decoded by the base station. Figure 2 , Figure 3It can be seen that the performance of all schemes improves with the increase of SNR, and the designed MD-SAMP algorithm is significantly better than OMP and BSASP algorithms, and is very close to MOMP algorithm. Since OMP and BSASP algorithms only reconstruct the CVR matrix in the angle domain, the CVRs of the attacker and the attacked LU will overlap highly under a limited number of discretized virtual AoAs, which makes it difficult to eliminate the contaminated coefficients in the CVR of the attacked LU, which seriously deteriorates the performance of the receiver for channel estimation and data decoding. On the other hand, MOMP and the designed MD-SAMP algorithm reconstruct the CVR matrix in the angle-delay-time slot domain. In the expanded channel domain, the sparsity of the CVR of each transmitter will be significantly enhanced, thereby significantly reducing the overlap probability of the CVR of each LU and the attacker, thereby improving the channel reconstruction and data decoding performance. However, the MOMP algorithm assumes that the true sparsity of the CVR is perfectly known by the base station, which cannot be realized in practice. In order to solve this problem, the designed MD-SAMP algorithm realizes signal recovery by adaptively adjusting the sparsity, and can achieve performance close to that of the MOMP algorithm. In addition, by comparing with the purple line without considering PCA detection and channel decontamination, it can be seen that PCA will seriously degrade the performance of channel estimation and data decoding. Since the present invention designs the criteria for PCA detection and channel decontamination, the proposed scheme can effectively resist PCA and improve system performance. However, the performance of the Oracle-LS algorithm is better than that of the proposed MD-SAMP algorithm. This is because the Oracle-LS algorithm assumes that the number and position of the paths in the CVR matrix are known, which is impossible to achieve in practice.

[0091] In summary, the present invention proposes a large-scale access pilot contamination attack countermeasure method based on channel decontamination, which makes full use of the sparsity and correlation of the millimeter wave channels of legitimate users and attackers in the angle-delay-time slot domain, and obtains higher channel recovery and data decoding performance. Simulation results show that compared with the existing PCA countermeasure scheme, the PCA countermeasure method based on channel decontamination of the present invention effectively reduces the mean square error of channel estimation, improves the accuracy of data decoding, and has better system performance.

[0092] Embodiment 2, the present invention provides a communication system, including a base station with M antennas, N LUs with single antennas, and multiple PCA attackers; each LU generates sporadic traffic and has a fixed position, and the PCA attacker can randomly launch attacks and has a variable position; the LU index set is represented as The base station antenna index set is expressed as Each LU is assigned a pilot sequence s n, with a length of C; assume that multiple PCA attackers attack by sending the same pilot sequence as the active LU to pollute the channel information of the LU; the pilot signals received by the base station from all active LUs and PCA attackers in time slot t are represented by the channel model Y t =G t S+G′ t S+W t Indicates that G t and G′ t They represent the channel matrices of LU and attacker, S represents the pilot matrix, and W t Represents noise; communication is performed based on the above-mentioned large-scale access pilot pollution attack countermeasure method based on channel decontamination.

[0093] In addition, the present invention can also provide a computer device, including a processor and a memory, the memory is used to store a computer executable program, the processor reads part or all of the computer executable program from the memory and executes it, and when the processor executes part or all of the computer executable program, it can implement the large-scale access pilot pollution attack countermeasure method based on channel decontamination described in the present invention.

[0094] On the other hand, the present invention provides a computer-readable storage medium, in which a computer program is stored. When the computer program is executed by a processor, the method for countering large-scale access pilot pollution attacks based on channel decontamination described in the present invention can be implemented.

[0095] The computer device may be a laptop computer, a desktop computer or a workstation.

[0096] The processor of the present invention may be a central processing unit (CPU), a digital signal processor (DSP), an application specific integrated circuit (ASIC) or a readily available field programmable gate array (FPGA).

[0097] The memory of the present invention may be an internal storage unit of a laptop computer, a desktop computer or a workstation, such as a memory or a hard disk; or an external storage unit, such as a mobile hard disk or a flash memory card.

[0098] Computer-readable storage media may include computer storage media and communication media. Computer storage media include volatile and non-volatile, removable and non-removable media implemented by any method or technology for storing information such as computer-readable instructions, data structures, program modules or other data. Computer-readable storage media may include: read-only memory (ROM), random access memory (RAM), solid-state drive (SSD) or optical disk, etc. Among them, random access memory may include resistance random access memory (ReRAM) and dynamic random access memory (DRAM).

Claims

1. A method for countering large-scale access pilot pollution attacks based on channel decontamination, characterized in that: The following steps are involved: A millimeter-wave uplink multi-input multi-output system model is built based on the unlicensed mMTC network. The transmission characteristics of legitimate users and attackers in the angle-delay-time slot domain are jointly considered to construct a three-dimensional sparse CVR matrix H and a three-dimensional received signal Y. A three-dimensional transmission model with sparsity in the angle-delay domain and time correlation in the time slot domain is obtained. Based on the constructed three-dimensional transmission model, the channel decontamination problem against pilot contamination attacks can be transformed into the following multi-dimensional sparse signal recovery problem: Where t represents the time slot index, T represents the total number of time slots included in one transmission, and O {：，t} =vec(Y {：，：，t} ) represents the vectorized observation signal, Φ represents the measurement tensor, Ψ k Represents a dictionary, and A set of atomic indices representing dictionaries, Where i1 and i2 represent the column index and row index of the dictionary Ψ1, j1 and j2 represent the column index and row index of the dictionary Ψ2, N represents the total number of users in the system, D represents the number of channel delay taps, M represents the number of base station antennas, and R represents the total number of users in the system. delay , R AoA represents the number of delays and angles in the dictionary; H represents the estimated CVR matrix, H {j，t} represents the value of the estimated CVR matrix indexed at time slot t, j; The multidimensional sparse signal recovery problem is solved by the sparse adaptive matching pursuit algorithm based on the multidimensional dictionary, and the estimation result of the CVR matrix H is obtained. Estimation results based on CVR matrix H Combine user activity detection, PCA detection and Xinda decontamination criterion to achieve PCA confrontation and obtain the estimated set of legitimate active users PCA attacker collection The legitimate user channel matrix after decontamination reconstruction 2. The method for countering large-scale access pilot pollution attacks based on channel decontamination according to claim 1 is characterized in that: The transmission system includes a base station with M antennas, N single-antenna LUs and multiple PCA attackers. Assume that each LU generates sporadic traffic and has a fixed position, and the PCA attacker can launch attacks randomly and has a variable position. The LU index set is expressed as The base station antenna index set is expressed as Each LU is assigned a pilot sequence s n , with a length of C; suppose that multiple PCA attackers attack by sending the same pilot sequence as the active LU to pollute the channel information of the LU; The pilot signals received by the base station from all active LUs and PCA attackers in time slot t are modeled by channel model Y t =G t S+G′ t S+W t Indicates that G t and G′ t They represent the channel matrices of LU and attacker, S represents the pilot matrix, and W t Indicates noise.

3. The method for countering large-scale access pilot pollution attacks based on channel decontamination according to claim 2 is characterized in that: The channel model Y t =G t S+G′ t S+W t In the above example, the channel matrix is ​​expressed as The two-dimensional indicator variable and Respectively represent the activation state and the attacked state of LUn, and They represent the channel matrix of LUn in time slot t and the channel matrix of the attacker attacking LUn in time slot t respectively; the channel information is composed of D channel taps, that is, Represents the channel information at tap d, using the geometric channel model Indicates that, where L represents the number of scatterers, represents the channel gain of the lth path of LUn in time slot t, θ n，l represents the physical AoA of the lth path, a(θ n，l ) represents the antenna array response, τ n，l represents the delay of path l, T s represents the sampling period, f(τ) represents the sampling pulse; CVR is introduced to describe the sparse characteristics of the millimeter wave channel, then Expressed as Where B represents the virtual AoA dictionary, f[d] represents the delay dictionary, represents the CVR matrix of LUn in time slot t. The CVR matrix has sparse characteristics in the angle-delay domain. Correspondingly, for an attacker attacking LUn, the channel information Expressed as in, Denotes the CVR matrix of an attacker attacking LUn at time slot t.

4. The method for countering large-scale access pilot pollution attacks based on channel decontamination according to claim 2, characterized in that: The three-dimensional transmission model composed of multiple time slots is constructed as follows: {：，：，t} =BH {：,：，t} FS+W t , t = 1, ..., T, where Y {：，：，t} =Y t represents the received signal at time slot t, represents the unified CVR matrix of time slot t, where represents the CVR of LUn in the attacked, safe, and idle states at time slot t, where and represents the set of LUs that are active and attacked in time slot t; the activation state of LU remains unchanged within T time slots. For user LUs in a secure state n , the CVR has time correlation.

5. The method for countering large-scale access pilot pollution attacks based on channel decontamination according to claim 1, characterized in that: The multidimensional sparse signal recovery problem is solved by using a sparse adaptive matching pursuit algorithm based on a multidimensional dictionary, including: estimating the CVR matrix under fixed sparsity through matching projection and residual update steps, and updating the sparsity through iterative adaptive updating to obtain the final estimation result closest to the true CVR matrix H. Among them, the matching projection step is expressed as: Among them, O res Represents the residual observation component; the residual update steps are: Where Γ represents the currently estimated matrix support set, and Represents the tensor Φ and the current estimation result The deformation matrix, is the Kronecker product of dictionaries.

6. The method for countering large-scale access pilot pollution attacks based on channel decontamination according to claim 5, characterized in that: When the sparsity is updated through iterative adaptive updating, signal recovery is performed under a fixed sparsity s, and the sparsity s is updated based on the output result of the inner layer.

7. The method for countering large-scale access pilot pollution attacks based on channel decontamination according to claim 1, characterized in that: Estimation results based on CVR matrix H Implementing joint user activity detection, PCA detection and Xinda decontamination includes the following steps: According to the estimated CVR matrix of each user The power of is used to detect user activity and obtain the estimated LU activity indicator vector: in, Represents the power threshold, and obtains the estimated active LU set Perform PCA detection to screen the effective path set in, represents the estimated CVR of LUn at time slot t, Represents the power threshold, and the estimated LU attack state indicator vector is obtained: And get the estimated attacked LU set Channel decontamination is performed, and the time correlation of the LU active state is used to decontaminate and reconstruct the CVR matrix of LUn in time slot t to obtain the decontaminated CVR matrix of LUn in time slot t: in, based on Get the decontaminated CVR matrix of all N LUs in time slot t And the channel information matrix 8. A communication system, characterized in that: It includes a base station with M antennas, N single-antenna LUs, and multiple PCA attackers; each LU generates sporadic traffic and has a fixed position, and the PCA attacker can launch attacks randomly and has a variable position; the LU index set is represented as The base station antenna index set is expressed as Each LU is assigned a pilot sequence s n , with a length of C; suppose that multiple PCA attackers attack by sending the same pilot sequence as the active LU to pollute the channel information of the LU; The pilot signals received by the base station from all active LUs and PCA attackers in time slot t are modeled by channel model Y t =G t S+G′ t S+W t Indicates that G t and G′ t They represent the channel matrices of LU and attacker, S represents the pilot matrix, and W t Indicates noise; Communication is performed based on the large-scale access pilot pollution attack countermeasure method based on channel decontamination as described in any one of claims 1-6.

9. A computer device, characterized in that: It includes a processor and a memory, the memory is used to store a computer executable program, the processor reads the computer executable program from the memory and executes it, and when the processor executes the computer executable program, it can implement the large-scale access pilot pollution attack countermeasure method based on channel decontamination as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that: A computer program is stored in a computer-readable storage medium. When the computer program is executed by a processor, the method for countering large-scale access pilot pollution attacks based on channel decontamination as described in any one of claims 1 to 7 can be implemented.

Images