IRS-assisted high-speed rail communication system physical layer security optimization method and system

Abstract

The invention discloses an IRS-assisted high-speed rail communication system physical layer security optimization method and system, and the method comprises the steps: constructing an IRS-assisted high-speed rail millimeter wave MIMO physical layer security system model, and obtaining the expression of each cascade channel in combination with a Saleh-Valenzuela channel model with a time-varying characteristic; the speed of the train is predicted by using a PSO-LSTM neural network method through the expression of each cascade channel, Doppler frequency shift compensation is carried out, and a high-speed rail millimeter wave MIMO physical layer safety system model after Doppler frequency shift compensation is obtained; a high-speed rail millimeter wave MIMO physical layer security system model after Doppler frequency shift compensation is used, and based on IRS phase shift constraint and base station maximum power constraint, joint optimization is carried out on a base station beam forming matrix and IRS phase shift by taking minimization of a system secrecy outage probability as a target. According to the invention, the security of the high-speed rail communication system is effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of high-speed rail communication technology, and in particular to an IRS-assisted high-speed rail communication system physical layer security optimization method and system. Background Art

[0002] With the rapid development of high-speed railways, their information and intelligence levels are constantly improving, which puts higher requirements on the data transmission rate and security of wireless communication systems. In high-speed railway wireless communication systems, traditional communication systems face severe challenges due to the Doppler effect, signal attenuation and complex channel environment caused by the high-speed movement of trains. Especially in terms of physical layer security, that is, how to use the characteristics of wireless channels to ensure the security of communications without relying on high-level encryption, is a hot topic and difficulty in current research.

[0003] In high-speed railway wireless communication systems, Intelligent Reflecting Surface (IRS) technology is considered to be a promising technology for improving spectrum efficiency and transmission rate. IRS consists of a large number of low-cost passive reflective units, which can reshape the channel environment and alleviate the problem of rapid fluctuations in received signal strength caused by the Doppler effect. However, existing research on IRS-assisted wireless communication systems mainly focuses on performance analysis such as spectrum efficiency and channel capacity, and relatively little research on physical layer security.

[0004] Therefore, in order to further ensure passenger privacy and driving safety, it is imperative to study IRS-assisted high-speed rail safety communication. Considering the characteristics of high-speed rail scenarios such as high speed and complex terrain, the present invention proposes a physical layer security optimization method and system for IRS-assisted high-speed rail communication systems. Summary of the invention

[0005] In order to solve the technical problems existing in the above-mentioned prior art, the present invention proposes a physical layer security optimization method and system for an IRS-assisted high-speed railway communication system, which enhances the performance of high-speed railway communication in the presence of eavesdroppers by optimizing the base station transmit beamforming and the IRS phase shift matrix, thereby improving the secure transmission capability of railway wireless communications.

[0006] On the one hand, to achieve the above-mentioned purpose, the present invention provides an IRS-assisted high-speed railway communication system physical layer security optimization method, comprising:

[0007] Construct a high-speed rail millimeter-wave MIMO physical layer security system model based on IRS assistance, and combine it with the Saleh-Valenzuela channel model with time-varying characteristics to obtain the expressions of each cascade channel in the high-speed rail communication system;

[0008] Through the expressions of the cascade channels at each level, the PSO-LSTM neural network method is used to predict the train speed, obtain the Doppler frequency shift value, perform Doppler frequency shift compensation, and obtain the high-speed rail millimeter wave MIMO physical layer safety system model after Doppler frequency shift compensation;

[0009] Using the high-speed rail millimeter-wave MIMO physical layer security system model after Doppler frequency shift compensation, based on IRS phase shift constraints and base station maximum power constraints, the base station beamforming matrix and IRS phase shift are jointly optimized with the goal of minimizing the probability of system confidentiality interruption.

[0010] On the other hand, to achieve the above-mentioned purpose, the present invention also provides an IRS-assisted high-speed railway communication system physical layer security optimization system, comprising: a model construction module, an expression construction module, a speed prediction module, an optimization problem construction module, a multivariable decoupling module and an optimal solution calculation module;

[0011] The model building module is used to add IRS to the high-speed rail communication model to build a high-speed rail millimeter wave MIMO physical layer security system model based on IRS assistance;

[0012] The expression building module is used to obtain expressions of each cascade channel in the system by taking into account the Saleh-Valenzuela channel model with time-varying characteristics through the IRS-assisted high-speed rail physical layer safety system model;

[0013] The optimization problem construction module is used to construct a joint optimization problem of the base station beamforming matrix and the IRS phase shift with the goal of minimizing the probability of system confidentiality interruption;

[0014] The multivariable decoupling module is used to propose a two-stage secure beamforming algorithm to handle the coupling between variables in the proposed optimization problem through the joint optimization problem of the base station beamforming matrix and the IRS phase shift;

[0015] The optimal solution calculation module is used to use the coupling to obtain the optimal solution of the optimization problem through the CNN framework and the properties of the generalized Rayleigh quotient.

[0016] Compared with the prior art, the present invention has the following advantages and technical effects:

[0017] The present application provides an IRS-assisted physical layer security optimization method and system for high-speed rail communication systems, which solves the problem of high probability of system confidentiality interruption when there are illegal eavesdroppers in high-speed mobile scenarios, ensures the security of large-capacity and high-speed data transmission in high-speed rail scenarios, effectively improves the security of high-speed rail communication systems, and is low-cost and easy to deploy. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0019] Figure 1 This is a flow chart of a physical layer security optimization method for a high-speed railway communication system assisted by an IRS according to an embodiment of the present invention;

[0020] Figure 2 This is a schematic diagram of the structure of a high-speed rail millimeter wave secure communication model based on IRS-MIMO constructed in Example 2 of the present invention;

[0021] Figure 3 A relationship diagram of the confidentiality interruption probability when the method of the embodiment of the present invention changes the number of IRS reflection units and the Doppler frequency shift compensation error;

[0022] Figure 4 A relationship diagram of the confidentiality interruption probability between the method of the embodiment of the present invention and the change of the signal-to-noise ratio and whether the IRS is used and the random phase shift of the IRS;

[0023] Figure 5 The diagram is a relationship diagram between the method of the embodiment of the present invention and the change of the physical layer coding rate and whether IRS is used and the probability of confidentiality interruption under IRS random phase shift. DETAILED DESCRIPTION

[0024] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0025] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0026] Embodiment 1

[0027] This embodiment proposes an IRS-assisted high-speed rail communication system physical layer security optimization method, such as Figure 1 ,include:

[0028] Construct a high-speed rail millimeter-wave MIMO physical layer security system model based on IRS assistance, and combine it with the Saleh-Valenzuela channel model with time-varying characteristics to obtain the expressions of each cascade channel in the high-speed rail communication system;

[0029] Through the expressions of the cascade channels at each level, the PSO-LSTM neural network method is used to predict the train speed, obtain the Doppler frequency shift value, perform Doppler frequency shift compensation, and obtain the high-speed rail millimeter wave MIMO physical layer safety system model after Doppler frequency shift compensation;

[0030] Using the high-speed rail millimeter-wave MIMO physical layer security system model after Doppler frequency shift compensation, based on the IRS phase shift constraint and the base station maximum power constraint, with the goal of minimizing the probability of system confidentiality interruption, a joint optimization problem of the base station beamforming matrix and the IRS phase shift is constructed;

[0031] Based on the joint optimization problem of the base station beamforming matrix and the IRS phase shift, a two-stage beamforming algorithm is proposed to handle the coupling between variables in the joint optimization problem, and the optimal solution of the optimization problem is obtained through the coupling.

[0032] Furthermore, an intelligent reflective surface IRS is added to the high-speed rail communication model to construct the IRS-assisted high-speed rail millimeter wave MIMO physical layer security system model, including:

[0033] Assume that the trackside base station is N BS Large uniform planar array (UPA) E In the presence of Eve, an illegal eavesdropper, MR The mobile relay MR on the vehicle sends confidential signals. The IRS is deployed between the BS and the MR and consists of N reflection units. The phase shift matrix of the IRS is expressed as:

[0034]

[0035] In the formula, β 1,1 is the amplitude of the unit corresponding to the first row and first column of IRS, β m,n is the amplitude of the unit corresponding to the mth row and nth column of IRS, θ 1,1 is the phase of the unit corresponding to the first row and first column of the IRS, θ m,n is the phase of the cell corresponding to the mth row and nth column of the IRS, e is the base of the natural logarithm, and j is the imaginary unit, which represents the imaginary part;

[0036] In downlink transmission, the received signal at MR is expressed as:

[0037] y=(H BM +H IM ΘH BI )ws+n,

[0038] The received signal at Eve is expressed as:

[0039] y E =(H BE +HIE ΘH BI )ws+n E ,

[0040] In the formula, H BM represents the channel matrix between BS and MR, H IM represents the channel matrix between IRS-MR, H BI represents the channel matrix between BS-IRS, and w represents the channel matrix satisfying ||w(t)|| 2 =P T The base station beamforming matrix, P T is the maximum transmission power of the base station; s represents the single-stream confidential information sent by the BS to the MR, n and n E are the additive white Gaussian noise at MR and Eve, respectively, H BE is the channel matrix between BS-Eve, is the channel matrix between IRS-Eve, y is the received signal at MR, y E is the received signal at Eve.

[0041] Furthermore, based on the IRS high-speed rail physical layer safety system model, the Saleh-Valenzuela channel model with time-varying characteristics is considered, and the expressions of each cascade channel in the system are obtained, including:

[0042] Assume that in actual implementation, the IRS is deployed at an ideal position where the line-of-sight component exists, the positions of the BS, IRS, and Eve are fixed, and the MR moves at high speed with the train. Considering the Doppler effect, the BS-MR and IRS-MR channels are expressed as:

[0043]

[0044]

[0045] Where: L represents the number of multipaths, represents the departure angle of BS and the arrival angle of MR, denote the azimuth and elevation angles of departure from BS and arrival at MR, respectively. and They represent the antenna array response vector, f BM,t 、f IM,t represents the Doppler frequency shift in time slot t, T s Represents the sampling period of the system, L BM is the number of multipaths in the BS-MR link, L IM is the number of multipaths in the IRS-MR link, for The conjugate transpose of is the arrival angle of IRS-MR, is the large-scale fading coefficient, which obeys the complex Gaussian distribution Where PL(dB) = PL 0 +10κlg(d)+PL d , PL 0 represents the path loss from the transmission point to the reference point. In this embodiment, the reference distance is set to 1 meter. d represents the distance from the BS to the MR. κ is the Rice factor. It is a shadow fading with a mean of 0 and a complex Gaussian distribution.

[0046] The antennas at the BS and MR are linear arrays (ULA), and the steering vector of the ULA is:

[0047]

[0048] Where λ represents the carrier wavelength, d s Indicates the spacing between different antennas, usually set to d s =λ2. The UPA steering vector at IRS is:

[0049]

[0050] The antennas at the BS and MR are linear arrays ULA, where the steering vector of the ULA is:

[0051]

[0052] Where λ represents the carrier wavelength, d s represents the interval between different antennas, ζ is the arrival angle of the steering vector, is the steering vector of ULA;

[0053] The UPA steering vector at the IRS is:

[0054]

[0055] In the formula, is the UPA steering vector, is the azimuth steering vector, is the steering vector of the pitch angle.

[0056] Furthermore, based on the PSO-LSTM neural network prediction model of the received signal of the train speed, a more accurate Doppler frequency shift value is obtained, so as to make early compensation, including:

[0057] Construct an LSTM neural network speed prediction model. The storage unit of LSTM is controlled by three activation gates (forget gate, input gate and output gate). The update formula of the three gates and the corresponding unit information is:

[0058] Forget gate: controls which information of the memory unit in the previous time step is retained in the unit information update, and forgets the parameter f in time step t t The calculation formula is:

[0059] f t =σ(W f ·[h t-1 ,x t ]+b f ),

[0060] Where W f is the weight matrix of the forget gate, h t-1 is the hidden state of the previous time step, x t is the input of the current time step, b f is the bias term of the forget gate.

[0061] Input gate: After the memory unit completes the forgetting process of information, the input gate determines how much the new candidate vector value participates in the update of the memory unit information. Its output parameter i t Indicates the degree of information retention:

[0062] i t =σ(W i ·[h t-1 , x t ]+b i ),

[0063] Where W i is the weight matrix of the input gate, b i is the bias term of the input gate.

[0064] State information update: selectively update the candidate state value c through the forget gate and input gate t :

[0065]

[0066] In the formula, is the weight matrix of the candidate memory unit, is the bias term of the candidate memory unit, c t is a candidate memory unit.

[0067] Output gate: processes the state value of the current memory unit and outputs it.

[0068] o t =σ(W o ·[h t-1 ,x t ]+b o ),

[0069] h t =o t*tanh(c t ),

[0070] In the formula, o t is the activation value of the output gate at the current time step, W o is the weight matrix of the output gate, b o is the bias term of the output gate, h t is the hidden state of the current time step, tanh is the hyperbolic tangent activation function, σ is the sigmoid activation function, and the definition is as follows:

[0071]

[0072] Where z is the value input to the activation function and e is the base of the natural logarithm.

[0073] Use the PSO algorithm to optimize the parameter selection of the LSTM neural network, and use the mean square error MSE of the neural network training as the fitness function of the PSO algorithm:

[0074]

[0075] In the formula, m is the number of samples, y is i and Represent the true value and predicted value respectively;

[0076] The speed and position of the PSO algorithm particles are updated through the fitness function:

[0077] v id =v id +c 1 r 1 (p best -x id )+c 2 r 2 (g best -x id ),

[0078] x id =x id +v id ,

[0079] In the formula, c 1 and c 2 are acceleration coefficients, representing the learning factor between particles and particle groups; r 1 and r 2 are random numbers independently generated in the interval [0, 1]; v id and x id They represent the velocity and position of the particle at the i-th iteration respectively; p best represents the optimal position searched during the iteration of the i-th particle, that is, the individual extreme value; gbest It represents the optimal position searched during the entire particle swarm iteration process, that is, the global extreme value.

[0080] Furthermore, with the goal of minimizing the probability of system confidentiality outage, a joint optimization problem of the base station beamforming matrix and IRS phase shift is constructed, including:

[0081] The channel capacities at MR and Eve are expressed as:

[0082]

[0083] In the formula, represents the normalized beamforming matrix, and the confidentiality capacity of the system is: C s =[C m -C w ] + ,Considering that the channel state information of an illegal eavesdropper is difficult to obtain in a secure communication scenario, C s Unable to calculate, the probability of confidentiality interruption is expressed as:

[0084]

[0085] In the formula, R s is the physical layer coding rate, Since the illegal channel obeys Rayleigh fading, H BE , H IE They can be expressed as H BE =β d Z,H IE =β r C, where β d , β r are the large-scale fading factors, is the Kronecker product.

[0086] Introduce auxiliary variables:

[0087]

[0088] The probability of confidentiality interruption can be rewritten as: P out (R s )=1-F X (φ),

[0089] In the formula, F X (·) is the cumulative distribution function of X.

[0090] The approximate expression of the probability of confidentiality interruption can be derived from the Gamma distribution:

[0091]

[0092] In the formula, is the Gamma function, is the upper incomplete Gamma function. According to the properties of the upper incomplete Gamma function, the probability of confidentiality interruption P out along with decreases with the increase of .

[0093] The joint optimization problem of the base station beamforming matrix and the IRS phase shift is modeled as:

[0094]

[0095] Where C1 is the BS maximum transmit power constraint, C2 is the unit mode constraint of the IRS unit, and φ is the introduced auxiliary variable. is the additive Gaussian white noise at Eve, C m is the channel capacity of the legal channel, R S is the physical layer security coding rate, P T is the transmission power of the base station, is the large-scale fading factor of the BS-Eve link, is the large-scale fading factor of the IRS-Eve link, H BI represents the channel matrix between BS-IRS, b is the normalized beamforming matrix, where w is the beamforming vector sent by the base station, b H is the conjugate transpose of b, The phase of the reflected signal that can be adjusted for each element of the IRS.

[0096] Furthermore, a two-stage secure beamforming algorithm is proposed to handle the coupling between variables in the proposed optimization problem, including:

[0097] The joint optimization problem is decomposed into sub-problem 1 of finding the IRS phase shift based on CNN and sub-problem 2 of finding the base station transmit beam matrix based on the properties of the generalized Rayleigh quotient.

[0098] Specifically, the CNN framework and the properties of the generalized Rayleigh quotient are used to obtain the optimal solution to the optimization problem, including:

[0099] In the sub-problem 1 of finding the IRS phase shift based on CNN, the phase shift matrix Θ is obtained through the training channel data, and the legal channel H is selected. BM , H IM , H BI As input, preprocess the input data;

[0100] definition h IM,n Indicates H IM The nth column of Indicates HBI The nth row of n Corresponding to the cascade channel of the nth unit of IRS, the cascade channel is re-expressed as: Raw data fed into the neural network It can be viewed as an N with N+1 channels. MR ×N BS Image, that is, X=[H BM ; F 1 ; F 2 ;...;F N ]. The data input to the neural network is required to be real valued, while X is complex valued. In order to solve the problem of inconsistency between the original data and the input data, the real part, imaginary part and absolute value of each element of X are extracted to form a new three-dimensional matrix

[0101] Preprocessed data After two convolutional layers and three fully connected layers, the Lambda layer finally outputs the prediction result, and the optimal IRS phase shift matrix is: Θ * =diag(e j2πc ).

[0102] In subproblem 2 of finding the base station transmit beam matrix based on the properties of the generalized Rayleigh quotient, for any given phase shift matrix Θ, the objective function of the optimization problem can be reformulated as:

[0103]

[0104] In the formula, a is the introduced auxiliary variable, is the additive white Gaussian noise at Eve, R s is the physical layer security coding rate, σ 2 is the additive white Gaussian noise at MR, t is the introduced auxiliary variable, P T is the transmission power of the base station;

[0105] Q 1 =(H BM +H IM ΘH BI ) H (H BM +H IM ΘH BI ),

[0106] According to the properties of the generalized Rayleigh quotient, the optimal b is:

[0107]

[0108] In the formula, represents the eigenvector corresponding to the maximum eigenvalue of matrix X, λ max is the largest eigenvalue of X. Therefore, the optimal beamforming matrix is

[0109] Embodiment 2

[0110] like Figure 2 As shown, this embodiment provides a high-speed rail millimeter wave safety communication model based on IRS-MIMO, and the high-speed rail millimeter wave safety communication model based on IRS-MIMO includes: base station BS, IRS, MR, Eve and trackside buildings. A physical layer security optimization method for a high-speed rail communication system assisted by IRS includes the following steps:

[0111] S1. Communication parameter acquisition: path loss model, number of base station antennas, maximum base station transmit power, number of IRS units.

[0112] S2. Based on the obtained relevant parameters, refer to the Saleh-Valenzuela model of the millimeter wave channel to obtain the channel state information of each cascade channel, including: H BM , H BI , H IM , H BE , W IE .

[0113] S3. Predicting train speed based on PSO-LSTM and performing Doppler frequency shift compensation in advance are as follows:

[0114] S31. Construct an LSTM neural network speed prediction model. The storage unit of LSTM is controlled by three activation gates (forget gate, input gate and output gate). The update formula of the three gates and the corresponding unit information is:

[0115] Forget gate: controls which information of the memory unit at the previous time step is retained during the unit information update.

[0116] Forget parameter f at time step t t The calculation formula is:

[0117] f t =σ(W f ·[h t-1 ,x t ]+b f ),

[0118] Input gate: After the memory unit completes the forgetting process of information, the input gate determines how much the new candidate vector value participates in the update of the memory unit information. Its output parameter i t Indicates the degree of information retention:

[0119] it =σ(W i ·[h t-1 , x t ]+b i ),

[0120] State information update: selectively update candidate state values ​​through the forget gate and input gate

[0121] Output gate: processes the state value of the current memory unit and outputs it.

[0122] o t =σ(W o ·[h t-1 ,x t ]+b o ),

[0123] h t =o t *tanh(c t ),

[0124] Where W is the weight matrix, b is the bias term, which represents the product of matrix elements, and h t-1 is the hidden state of the encoder at the previous moment, x t is the sample value of the sequence to be predicted, tanh is the hyperbolic tangent activation function, σ is the sigmoid activation function, and the definition is as follows:

[0125]

[0126] S32. Use the PSO algorithm to optimize the parameter selection of the LSTM neural network. The process is as follows:

[0127] First, the MSE of the neural network training is used as the fitness function of the PSO algorithm:

[0128]

[0129] In the formula, m is the number of samples, y is i and represent the true value and the predicted value respectively.

[0130] The speed and position equations of the PSO algorithm particle update are:

[0131] v id =v id +c 1 r 1 (p best -x id )+c 2 r 2 (g best -xid ),

[0132] x id =x id +v id ,

[0133] In the formula, c 1 and c 2 is the acceleration coefficient, which represents the learning factor between particles and particle groups, r 1 and r 2 is a random number generated independently in the interval [0, 1]. id and x id They represent the velocity and position of the particle at the i-th iteration, respectively, best represents the optimal position searched during the iteration of the i-th particle, i.e., the individual extreme value, g best It indicates the optimal position searched during the entire particle swarm iteration process, that is, the global extreme value.

[0134] S4. Calculate the probability of confidentiality interruption to form the final problem that needs to be optimized, which is:

[0135] The channel capacities at MR and Eve are expressed as:

[0136]

[0137] In the formula, represents the normalized beamforming matrix, and the confidentiality capacity of the system is: C s =[C m -C w ] + ,Considering that the channel state information of an illegal eavesdropper is difficult to obtain in a secure communication scenario, C s Unable to calculate, the probability of confidentiality interruption is expressed as

[0138]

[0139] In the formula, R s is the physical layer coding rate, Since the illegal channel obeys Rayleigh fading, H BE , H IE They can be expressed as H BE =β d Z,H IE =β r C, where β d , β r are the large-scale fading factors, is the Kronecker product. Introduce auxiliary variables:

[0140]

[0141] The probability of confidentiality interruption can be rewritten as: P out (R s )=1-F X (φ),

[0142] In the formula, F X (·) is the cumulative distribution function of X. The approximate expression of the probability of confidentiality interruption can be derived from the Gamma distribution:

[0143]

[0144] In the formula, is the Gamma function, is the upper incomplete Gamma function. According to the properties of the upper incomplete Gamma function, the probability of confidentiality interruption P out along with decreases with the increase of .

[0145] The joint optimization problem of the base station beamforming matrix and the IRS phase shift is modeled as:

[0146]

[0147] Where: C1 is the BS maximum transmit power constraint, and C2 is the unit mode constraint of the IRS unit.

[0148] S5. Optimize the formed optimization problem: Use a two-stage algorithm to obtain the optimal solution based on the properties of CNN and generalized Rayleigh quotient, specifically:

[0149] S51, Initialization: H BM , H IM , H BI , P T , R s , σ 2 , β d , β r and κ;

[0150] S52. Using a two-stage algorithm, the optimization problem is decomposed into sub-problem 1 of finding the IRS phase shift based on CNN and sub-problem 2 of finding the base station transmit beam matrix based on the properties of the generalized Rayleigh quotient, and the optimal solution is obtained:

[0151] Subproblem 1: Θ * =diag(e j2πc );

[0152] Sub-question 2:

[0153] In this embodiment, the execution steps for solving the minimization of the probability of confidentiality interruption are shown in Table 1:

[0154] Table 1

[0155]

[0156] The specific conditions of this example are: carrier frequency 32GHz, system bandwidth 500MHz, base station transmit power 42dBm, number of BS antennas 4, number of MR antennas 2, number of Eve antennas 2, BS height 10m, MR height 2.5m, IRS height 2.5m, Eve height 2.5m, Rice factor 10dBm, time slot t 0.6ms.

[0157] Figure 3 The influence of different Doppler frequency offset compensation errors Δf on system performance when SNR=10dB and Rs=3.5bit / s / Hz is studied. Figure 3 It can be seen that the larger the frequency offset compensation error, the larger the Doppler frequency offset, and the higher the probability of confidentiality interruption of the system, indicating that the Doppler frequency offset will affect the security performance of the system. In addition, as the number of IRS reflection units increases, the confidentiality interruption probability under non-precise frequency offset compensation (Δf = 0.02 and Δf = 0.05) gradually approaches the performance curve of precise frequency offset compensation (Δf = 0), because the IRS itself also has the ability to reconstruct electromagnetic waves and mitigate Doppler frequency offset.

[0158] Figure 4 The relationship between signal-to-noise ratio and confidentiality interruption probability is studied when Rs=3.5bit / s / Hz. The simulation results show that the use of IRS assistance can significantly reduce the confidentiality interruption probability. The increase of signal-to-noise ratio will increase the confidentiality capacity, thereby reducing the confidentiality interruption probability.

[0159] Figure 5 The relationship between the physical layer coding rate and the probability of confidentiality interruption when the signal-to-noise ratio is 10dB is studied. Regardless of whether there is IRS assistance or not, the probability of confidentiality interruption will increase with R S The simulation results show that when the physical layer coding rate is low, the probability of confidentiality interruption with IRS assistance approaches 0, and the physical layer security optimization method and system of the IRS-assisted high-speed railway communication system of the present invention show good security performance.

[0160] Embodiment 3

[0161] This embodiment provides an IRS-assisted high-speed rail communication system physical layer security optimization system, including: a model construction module, an expression construction module, a speed prediction module, an optimization problem construction module, a multivariable decoupling module and an optimal solution calculation module;

[0162] The model building module is used to add IRS to the high-speed rail communication model and build a high-speed rail millimeter wave MIMO physical layer security system model based on IRS assistance;

[0163] The expression building module is used to obtain the expressions of each cascade channel in the system by taking into account the Saleh-Valenzuela channel model with time-varying characteristics through the IRS-assisted high-speed rail physical layer safety system model;

[0164] The optimization problem construction module is used to construct the joint optimization problem of the base station beamforming matrix and the IRS phase shift with the goal of minimizing the probability of system confidentiality interruption;

[0165] The multivariable decoupling module is used to propose a two-stage secure beamforming algorithm to handle the coupling between variables in the proposed optimization problem through the joint optimization problem of the base station beamforming matrix and the IRS phase shift;

[0166] The optimal solution calculation module is used to use the coupling to obtain the optimal solution of the optimization problem through the CNN framework and the properties of the generalized Rayleigh quotient.

[0167] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. An IRS-assisted high-speed railway communication system physical layer security optimization method, characterized in that: include: Construct a high-speed rail millimeter-wave MIMO physical layer security system model based on IRS assistance, and combine it with the Saleh-Valenzuela channel model with time-varying characteristics to obtain the expressions of each cascade channel in the high-speed rail communication system; Through the expressions of the cascade channels at each level, the PSO-LSTM neural network method is used to predict the train speed, obtain the Doppler frequency shift value, perform Doppler frequency shift compensation, and obtain the high-speed rail millimeter wave MIMO physical layer safety system model after Doppler frequency shift compensation; Using the high-speed rail millimeter-wave MIMO physical layer security system model after Doppler frequency shift compensation, based on IRS phase shift constraints and base station maximum power constraints, the base station beamforming matrix and IRS phase shift are jointly optimized with the goal of minimizing the probability of system confidentiality interruption.

2. The method according to claim 1, characterized in that An intelligent reflective surface IRS is added to the high-speed rail communication model to construct the IRS-assisted high-speed rail millimeter wave MIMO physical layer security system model, including: Assume that the trackside base station is N BS Large uniform planar array (UPA) E In the presence of Eve, an illegal eavesdropper, MR The mobile relay MR on the vehicle sends a confidential signal, and the IRS is deployed between the base station BS and MR to obtain the phase shift matrix of the IRS; Based on the phase shift matrix of the IRS, a received signal at MR and a received signal at Eve in downlink transmission are obtained.

3. The method according to claim 2, characterized in that The phase shift matrix of the IRS is: In the formula, β 1,1 is the amplitude of the unit corresponding to the first row and first column of IRS, β m,n is the amplitude of the unit corresponding to the mth row and nth column of IRS, θ 1,1 is the phase of the unit corresponding to the first row and first column of the IRS, θ m,n is the phase of the cell corresponding to the mth row and nth column of the IRS, e is the base of the natural logarithm, and j is the imaginary unit, which represents the imaginary part; The received signal at the MR is: y=(H BM +H IM ΘH BI )ws+n, The received signal at Eve is: y E =(H BE +H IE ΘH BI )ws+n E , In the formula, H BM represents the channel matrix between BS and MR, H IM represents the channel matrix between IRS-MR, H BI represents the channel matrix between BS-IRS, and w represents the channel matrix satisfying ||w(t)|| 2 =P T The base station beamforming matrix, P T is the maximum transmit power of the base station; s represents the single-stream confidential information sent by BS to MR, n and n E are the additive white Gaussian noise at MR and Eve, respectively, H BE is the channel matrix between BS-Eve, H IE is the channel matrix between IRS-Eve, y is the received signal at MR, y E is the received signal at Eve.

4. The method according to claim 3, characterized in that The expressions of the cascade channels are obtained, including: The IRS is deployed at an ideal position where the line-of-sight component exists. The positions of BS, IRS and Eve are fixed, and MR moves with the train. Considering the Doppler effect, the BS-MR and IRS-MR channels are expressed as H BM , H IM : Where: L represents the number of multipaths, denote the departure angle of BS and the arrival angle of MR respectively, denote the azimuth and elevation angles of departure from BS and arrival at MR, respectively. and They represent the antenna array response vector, f BM,t 、f IM,t represents the Doppler frequency shift in time slot t, T s represents the sampling period of the system, are the large-scale fading coefficient, L BM is the number of multipaths in the BS-MR link, L IM is the number of multipaths in the IRS-MR link, for The conjugate transpose of is the arrival angle of IRS-MR; The antennas at the BS and MR are linear arrays ULA, where the steering vector of the ULA is: Where λ represents the carrier wavelength, d s represents the interval between different antennas, ζ is the arrival angle of the steering vector, is the steering vector of ULA; The UPA steering vector at the IRS is: In the formula, is the UPA steering vector, is the azimuth steering vector, is the steering vector of the pitch angle.

5. The method according to claim 1, characterized in that Obtaining the Doppler frequency shift value includes: Use the PSO algorithm to optimize the parameter selection of the LSTM neural network, and use the mean square error MSE of the neural network training as the fitness function of the PSO algorithm: In the formula, m is the number of samples, y is i and Represent the true value and predicted value respectively; The speed and position of the PSO algorithm particles are updated through the fitness function: v id =v id +c1r1(p best -x id )+c2r2(g best -x id ), x id =x id +v id , Where c1 and c2 are acceleration coefficients, representing the learning factors between particles and particle groups; r1 and r2 are random numbers independently generated in the interval [0, 1]; v id and x id They represent the velocity and position of the particle at the i-th iteration respectively; p best represents the optimal position searched during the iteration of the i-th particle, that is, the individual extreme value; g best It represents the optimal position searched during the entire particle swarm iteration process, that is, the global extreme value.

6. The method according to claim 1, characterized in that The joint optimization of the base station beamforming matrix and the IRS phase shift includes: Based on the joint optimization problem of the base station beamforming matrix and the IRS phase shift, a two-stage beamforming algorithm is proposed to process the coupling between variables in the joint optimization problem, and the optimal solution of the optimization problem is obtained through the coupling; The joint optimization problem of the base station beamforming matrix and the IRS phase shift is: Where: C1 is the BS maximum transmit power constraint, C2 is the unit mode constraint of the IRS unit, φ is the introduced auxiliary variable, is the large-scale fading factor of the BS-Eve link, is the large-scale fading factor of the IRS-Eve link, H BI represents the channel matrix between BS-IRS, b is the normalized beamforming matrix, w is the beamforming vector sent by the base station, b H is the conjugate transpose of b, The phase of the reflected signal that can be adjusted for each element of the IRS.

7. The method according to claim 6, characterized in that Dealing with the coupling between variables in the joint optimization problem includes: The joint optimization problem is decomposed into sub-problem 1 of finding the IRS phase shift based on CNN and sub-problem 2 of finding the base station transmit beam matrix based on the properties of the generalized Rayleigh quotient.

8. The method according to claim 7, characterized in that The optimal solution to the optimization problem is obtained by using the CNN framework and the properties of the generalized Rayleigh quotient, including: In the sub-problem 1 of finding the IRS phase shift based on CNN, the phase shift matrix Θ is obtained by training channel data, and the legal channel H is selected. BM , H IM , H BI As input, preprocess the input data; definition h IM,n Indicates H IM The nth column of Indicates H BI The nth row of F n Corresponding to the cascade channel of the nth unit of IRS, the cascade channel is re-expressed as: Raw data fed into the neural network is considered as an N with N+1 channels MR ×N BS Image, that is, X=[H BM ;F1;F2;...;F N ], extract the real part, imaginary part and absolute value of each element in the original data X to form a new three-dimensional matrix Preprocessed data After two convolutional layers and three fully connected layers, the Lambda layer finally outputs the prediction result to obtain the optimal IRS phase shift matrix.

9. The method according to claim 8, characterized in that The optimal solution of the optimization problem is obtained by using the CNN framework and the properties of the generalized Rayleigh quotient, and further comprising: In the sub-problem 2 of finding the base station transmit beam matrix based on the properties of the generalized Rayleigh quotient, for any given phase shift matrix Θ, the objective function of the optimization problem is restated as: In the formula, a is the auxiliary variable introduced, σ 2 is the additive white Gaussian noise at MR, t is the introduced auxiliary variable, R S is the physical layer security coding rate, P T is the transmission power of the base station; Q1=(H BM +H IM ΘH BI ) H (H BM +H IM ΘH BI ), According to the properties of the generalized Rayleigh quotient, the optimal b is: In the formula, represents the eigenvector corresponding to the maximum eigenvalue of matrix X, λ max is the largest eigenvalue of X, and the optimal beamforming matrix is 10. An IRS-assisted high-speed railway communication system physical layer security optimization system, characterized in that: include: Model building module, expression building module, speed prediction module, optimization problem building module, multivariable decoupling module and optimal solution calculation module; The model building module is used to add IRS to the high-speed rail communication model to build a high-speed rail millimeter wave MIMO physical layer security system model based on IRS assistance; The expression building module is used to obtain expressions of each cascade channel in the system by taking into account the Saleh-Valenzuela channel model with time-varying characteristics through the IRS-assisted high-speed rail physical layer safety system model; The optimization problem construction module is used to construct a joint optimization problem of the base station beamforming matrix and the IRS phase shift with the goal of minimizing the probability of system confidentiality interruption; The multivariable decoupling module is used to propose a two-stage secure beamforming algorithm to handle the coupling between variables in the proposed optimization problem through the joint optimization problem of the base station beamforming matrix and the IRS phase shift; The optimal solution calculation module is used to use the coupling to obtain the optimal solution of the optimization problem through the CNN framework and the properties of the generalized Rayleigh quotient.

Images