A Physical Layer Security Edge Computation Unloading and Result Transfer Method for HPLC

Abstract

This invention relates to a physical layer security edge computing offloading and result transmission method for HPLC, comprising the following steps: Step A, calculating a statistically significant offloading channel based on the transmission topology and transmission distance of MIMO-HPLC; Step B, calculating the information leakage ratio and received signal-to-noise ratio based on the statistical information of the offloading channel and the eavesdropping channel; Step C, constructing the state space, action space, and reward function of deep reinforcement learning and their statistical characteristics; Step D, designing a two-stage stochastic iterative optimization strategy based on the statistical characteristics to obtain the optimal value of the reward function; Step E, determining the action space, and obtaining the state space based on the action space and the reward function, ultimately obtaining the transmission parameters of the system. This invention ultimately achieves an optimal transmission method that balances secure transmission rate and offloading latency, obtaining the optimal offloading ratio and CPU frequency allocation ratio.

Description

Technical Field

[0001] This invention relates to the field of power communication, and more particularly to the field of HPLC data acquisition and calculation. Background Technology

[0002] High-speed power line communication (HPLC) is a new generation of power metering data acquisition method that can collect electricity metering information from a wide range of users at high frequencies. However, with the continuous integration of new energy sources into the grid and the widespread application of electric vehicles, a large amount of electricity consumption information, load data, and dispatch instructions need to be transmitted via HPLC to achieve data acquisition, analysis, and control execution in new power systems based on microgrids. To reduce data latency, an energy controller, i.e., a multi-access edge computing server, is deployed at the central control node of each microgrid. Data is offloaded to the multi-access edge computing server via HPLC. However, the offloaded data needs to be secure, but HPLC carries the risk of data leakage during the offloading process. To address this issue, traditional application-layer encryption methods can be used to prevent data leakage; however, this method results in excessive computational load and significant computational latency, making it unsuitable for data acquisition, analysis, and control execution in new power systems. According to information theory, directional signal processing can be used to prevent information leakage and achieve secure information offloading, thereby realizing the secure transmission of information offloading and computation results. Summary of the Invention

[0003] To address the aforementioned issues of secure offloading and secure transmission of computation results, this invention proposes a method for secure physical layer edge computation offloading and result transmission. Specifically, it is based on a stochastic optimization deep reinforcement learning method. This method utilizes dynamically changing channel state information, combined with performance constraints such as information offloading tasks and latency, to achieve secure offloading of information and secure transmission of computation results.

[0004] The embodiments of the present invention provide the following technical solutions:

[0005] A method for physical layer security edge calculation offloading and result transmission of HPLC includes the following steps: Step A, calculate the statistically significant offloading channel based on the transmission topology and transmission distance of MIMO-HPLC;

[0006] Step B: Calculate the information leakage ratio and the received signal-to-noise ratio based on the statistics of the offloading channel and the eavesdropping channel.

[0007] Step C: Construct the state space, action space, and reward function of deep reinforcement learning, as well as its statistical properties;

[0008] Step D: Based on statistical characteristics, design a two-stage stochastic iterative optimization strategy to obtain the optimal value of the reward function;

[0009] Step E: Determine the action space, and obtain the state space based on the action space and the reward function, ultimately obtaining the system's transmission parameters.

[0010] Preferably, step A specifically includes:

[0011] A1, each electrical device that acquires and controls information uses multi-phase power lines to transmit information, and the channel from the k-th electrical device to the multi-access edge server is h. k ∈C M×M C is a complex number, where M is the number of equivalent channels, k = 1, ..., K; h k Each element in is h k,i,j Where i, j = 1, ..., M, k is the k-th user, and i and j are the values ​​of h. k The element in the i-th row and j-th column of the matrix, h k,i,j =exp(-(b0+b1f) m )d i,j )g i,j Where b0 and b1 are constants relating to the distance attenuation factor and the frequency attenuation factor, respectively, which can be obtained through a single measurement, d i,j It is h k,i,j The distance from electrical equipment to the multi-access edge server; g i,j is the random fading factor, which is a complex number whose real and imaginary parts follow a standard normal distribution; f is the center frequency of the HPLC subcarrier, and m∈[0,1] is the exponent of the fading factor.

[0012] Preferably, step A further includes:

[0013] A2, Based on measurements, the statistical characteristics of the channel are obtained using channel estimation methods, i.e. Where d i,j The mean is E(h k,i,j ) is for h k,i,j Take the average. Preferably, step A further includes:

[0014] A3. Establish an offloading and control model. In this model, K users offload through a multiple-input multiple-output (MIMO) channel, and the users are aware of the eavesdropper's channel state information, which can be represented as t. n ∈C M×M C is a complex number, where n = 1, ..., N; the signal received at the multi-access edge server is represented as Where f k It is the signal processing matrix for the k-th user, x kFor the data uninstalled by the k-th user, n k To reduce noise at multiple access edge servers, w k h is the received signal processing matrix for user k. k This is the transmission channel matrix from the k-th user to the MEC server; the signal received at the n-th eavesdropper can be represented as... t k It is the transmission channel matrix from the k-th user to the n-th eavesdropper.

[0015] Preferably, step B specifically includes:

[0016] B1, the signal-to-interference-plus-noise ratio (SIR) of user k on the multi-access edge server can be expressed as: i represents all users excluding user k, and the power leaked from each user to eavesdropper n is ||t. k f k x k || 2 Then the signal-to-noise ratio for user k unloading is j represents all N eavesdropping users;

[0017] B2, the statistical distribution of the signal-to-interference-plus-noise ratio (SIR) for user k, i.e., the probability density function of the numerator, is: The probability density function of the denominator is Where α is the corresponding random variable of the molecule, χ c The average power allocation is given for the transmitted power in the user's equivalent channel; Ga(·) is the gamma function, and Γ(·) is the gamma distribution; β is the corresponding random variable in the numerator.

[0018] K is the number of users, and α represents SINR. k The molecule is ||w k h k f k x k || 2 N is the number of eavesdropping users, χ c The power transmitted in the user's equivalent channel, i.e. Where P k is the maximum transmit power of user k, and M is the number of equivalent channels;

[0019] B3, the statistical distribution of the signal leakage-to-noise ratio for user k, i.e., the probability density function of the denominator, is: Its molecular distribution is the same as that of step B2.

[0020] Preferably, step C specifically includes:

[0021] C1, Construct the state space S k ={SINR k ,ηc ,θ k ,λ k}, where η c For the information unloading delay, θ k ∈[0,1] represents the proportion of computational data that needs to be unloaded, when θ k When θ = 0, it means all calculations are performed on the user side; when θ = 0, it means all calculations are performed on the user side. k When λ = 1, it means all data needs to be offloaded to the multi-access edge server; k λ is the allocation coefficient for the CPU clock speed of the multi-access edge server. k = (0,1]; Construct the action space Preferred,

[0022] C2, Construct the reward function. in These are weighting coefficients, and their values ​​are... C S For a safe transmission rate, its long-term reward function is: in, It is a weighting factor, thus obtaining the weighted value for different reward times.

[0023] Preferred,

[0024] C3, based on statistical properties, calculates the long-term statistical value of the reward function, which can be expressed as:

[0025] Preferably, step D specifically includes:

[0026] D1 calculates the gradient of the long-term reward function, i.e. Furthermore, based on the Bellman equation, the value of Q is calculated.

[0027] D2, Determine the execution strategy, i.e. The strategy employs the conjugate gradient method to update the reward value.

[0028] D3, will (s t a t L t s t+1 Stored in temporary memory, with the iteration count set to B. max Definition update s t Let a be the state space at time t. t It is the action space at time t, s t+1 Let be the state space at time t+1.

[0029] Preferably, step E specifically includes:

[0030] E1, initialize S k ={SINR k ,η c ,θ k ,λ k}, The identity matrix is ​​used as the matrix value, and non-matrix values ​​are calculated based on this value.

[0031] E2 clears the memory temporarily storing data and sets its capacity to B. max =1024, the sampled value is B s =256;

[0032] E3, Obtain from Actions Based on the strategy, determine the timely reward amount. And it is stored in the memory of the data;

[0033] E4, adopting the latest Where ω k It is a gradient descent strategy;

[0034] E5, when B < B max Loop through D1 to E4, when B = B max Read the set of actions in memory. and θ in the state set k ,λ k .

[0035] Compared with existing technologies, the above technical solution has the following advantages:

[0036] This invention utilizes deep reinforcement learning optimization tools to analyze the HPLC-based safe offloading method and result transmission method. Based on the random characteristics of the state space and action space, it uses random matrix theory to optimize SLNR and physical layer secure transmission rate, and finally obtains the transmit signal processing matrix, the multi-access edge computing receive matrix, and the optimal offloading ratio and CPU frequency allocation ratio. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This invention provides a method for physical layer security edge calculation unloading and result transmission of HPLC, which is an embodiment of the present invention. Detailed Implementation

[0039] As described in the background section, how to reduce HPLC unloading delay and ensure safe unloading are urgent problems to be solved.

[0040] The core idea of ​​this invention is to utilize a physical layer secure transmission strategy and employ deep reinforcement learning to achieve optimal transmission and reception across multiple time slots, thereby achieving secure offloading and secure transmission of results.

[0041] See Figure 1 This invention provides a method for physical layer security edge computing offloading and result transmission in HPLC, the method comprising:

[0042] Step A: Calculate the statistically significant offloading channel based on the transmission topology and transmission distance of MIMO-HPLC.

[0043] Step A specifically includes:

[0044] A1. Each electrical device that acquires and controls information uses multi-phase power lines to transmit information. The channel from the k-th (k=1,...,K) electrical device to the multi-access edge server is h. k ∈C M×M , where M is the number of equivalent channels. h k Each element in the array is h. k,i,j Where i, j = 1,...,M; h k,i,j =exp(-(b0+b1f) m )d i,j )g i,j Where b0 and b1 are constants relating to the distance attenuation factor and the frequency attenuation factor, respectively, which can be obtained through a single measurement, d i,j It is h k,i,j The distance from electrical equipment to the multi-access edge server; g i,j is the random fading factor, which is a complex number whose real and imaginary parts follow a standard normal distribution. f is the center frequency of the HPLC subcarrier.

[0045] A2, Based on measurements, the statistical characteristics of the channel are obtained using channel estimation methods, i.e. Where d i,j The mean is R; E(h) k,i,j ) is for h k,i,j Take the average.

[0046] A3. Establish an offloading and control model. In this model, K users offload their data through a Multiple Input Multiple Output (MIMO) channel, and the users are aware of the eavesdropper's channel state information. This can be represented as t n ∈C M×M, where n = 1, ..., N; its channel is similar to the channel model in step A2. The signal received at the multi-access edge server can be represented as Where f k It is the signal processing matrix for the k-th user, x k For the data uninstalled by the k-th user, n k To reduce noise at multiple access edge servers, w k This is the signal processing matrix for user k. The signal received at the nth eavesdropper can be represented as...

[0047] Step B: Calculate the information leakage ratio and the received signal-to-noise ratio based on the statistics of the offloading channel and the eavesdropping channel.

[0048] Step B specifically includes:

[0049] B1, the signal-to-interference-plus-noise ratio (SIR) of user k on multiple access edge servers can be expressed as: Meanwhile, the power leaked to the eavesdropper n is ||t k f k x k || 2 Then the signal-to-noise ratio for user k unloading is

[0050] B2, the statistical distribution of the signal-to-interference-plus-noise ratio (SIR) for user k, i.e., the probability density function of the numerator, is: The probability density function of the denominator is Where α is the corresponding random variable of the molecule, χ c The average power allocation is the transmit power in the user's equivalent channel; Ga(·) is the gamma function, and Γ(·) is the gamma distribution; β is the corresponding random variable in the numerator.

[0051] B3, the statistical distribution of the signal leakage-to-noise ratio for user k, i.e., the probability density function of the denominator, is: Its molecular distribution is the same as that of B2.

[0052] Step C: Construct the state space, action space, and reward function of deep reinforcement learning, as well as its statistical properties;

[0053] Step C specifically includes:

[0054] C1, Construct the state space S k ={SINR k ,η c ,θ k ,λ k}, where η c For the information unloading delay, θ k∈[0,1] represents the proportion of computational data that needs to be unloaded, when θ k When θ = 0, it means all calculations are performed on the user side; when θ = 0, it means all calculations are performed on the user side. k When λ = 1, it means all data needs to be offloaded to the multi-access edge server. k λ is the allocation coefficient for the CPU clock speed of the multi-access edge server. k = (0,1). Constructing the action space.

[0055] C2, Construct the reward function. in These are weighting coefficients, and their values ​​are... C S For a safe transmission rate, its long-term reward function is: in, It is a weighting factor, thus obtaining the weighted value for different reward times.

[0056] C3, based on statistical properties, calculates the long-term statistical value of the reward function, which can be expressed as:

[0057] Step D: Based on statistical characteristics, design a two-stage stochastic iterative optimization strategy to obtain the optimal value of the reward function;

[0058] Step D specifically includes:

[0059] D1 calculates the gradient of the long-term reward function, i.e. Furthermore, based on the Bellman equation, the value of Q is calculated.

[0060] D2, Determine the execution strategy, i.e. The strategy employs the conjugate gradient method to update the reward value.

[0061] D3, will (s t a t L t s t+1 Stored in temporary memory, with the iteration count set to B. max Definition update

[0062] Step E: Determine the action space, and obtain the state space based on the action space and the reward function, ultimately obtaining the system's transmission parameters.

[0063] Step E specifically includes:

[0064] E1, initialize S k ={SINR k ,η c ,θ k,λ k}, The identity matrix is ​​used as the matrix value, and non-matrix values ​​are calculated based on this value.

[0065] E2 clears the memory temporarily storing data and sets its capacity to B. max =1024, the sampled value is B s =256;

[0066] E3, Obtain from Actions Based on the strategy, determine the timely reward amount. And it is stored in the memory of the data;

[0067] E4, adopting the latest Where ω k It is a gradient descent strategy;

[0068] E5, when B < B max Loop through D1 to E4, when B = B max Read the set of actions in memory. and θ in the state set k ,λ k .

[0069] This invention utilizes stochastic matrix theory, taking advantage of the stochastic characteristics of HPLC channels and eavesdropping user channels, to analyze the distributions of SINR and SLNR. Based on these distributions, a reinforcement learning optimization method based on stochastic optimization is constructed. This method can effectively prevent information leakage during HPLC offloading, improving the system's information security performance. It can obtain the globally optimal solution while balancing offloading delay and information security. This method guarantees optimal system performance and robustness over a long time horizon.

[0070] The various sections in this manual are described in a progressive manner, with each section focusing on the differences from the others. Similar or identical sections can be referred to each other.

[0071] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for physical layer security edge calculation unloading and result transmission in HPLC, characterized in that, The steps include: Step A, calculating the statistically significant offloading channel based on the transmission topology and transmission distance of MIMO-HPLC; Step B: Calculate the information leakage ratio and the received signal-to-noise ratio based on the statistics of the offloading channel and the eavesdropping channel. Step B specifically includes: B1, the signal-to-interference-plus-noise ratio (SIR) to user k on the multi-access edge server is expressed as: i represents all users excluding user k, and the power leaked from each user to eavesdropper n is ||t. k f k x k || 2 Then the signal-to-noise ratio for user k unloading is j represents all N eavesdropping users; f k It is the signal processing matrix for the k-th user, x k For the data uninstalled by the k-th user, n k To reduce noise at multiple access edge servers, w k h is the received signal processing matrix for user k. k This is the transmission channel matrix from the k-th user to the MEC server; the signal received at the n-th eavesdropper can be represented as... t k It is the transmission channel matrix from the k-th user to the n-th eavesdropper; B2, the statistical distribution of the signal-to-interference-plus-noise ratio (SIR) for user k, i.e., the probability density function of the numerator, is: The probability density function of the denominator is χ c The average power transmitted in the user's equivalent channel; Ga(·) is the gamma function, and Γ(·) is the gamma distribution; β is the corresponding random variable in the numerator; K is the number of users, and α represents SINR. k The molecule is ||w k h k f k x k || 2 N is the number of eavesdropping users, χ c The power transmitted in the user's equivalent channel, i.e. Where P k is the maximum transmit power of user k, and M is the number of equivalent channels; B3, the statistical distribution of the signal leakage-to-noise ratio for user k, i.e., the probability density function of the denominator, is: Its molecular distribution is the same as in step B2; Step C: Construct the state space, action space, and reward function of deep reinforcement learning, as well as its statistical properties; Step D: Based on statistical characteristics, design a two-stage stochastic iterative optimization strategy to obtain the optimal value of the reward function; Step E: Determine the action space, and obtain the state space based on the action space and the reward function, and finally obtain the transmission parameters of the system.

2. The method for physical layer security edge calculation unloading and result transmission of HPLC according to claim 1, characterized in that, Step A specifically includes: A1, each electrical device that acquires and controls information uses multi-phase power lines to transmit information, and the channel from the k-th electrical device to the multi-access edge server is h. k ∈C M×M C is a complex number, where M is the number of equivalent channels, k = 1, ..., K; h k Each element in is h k，ij Where i, j = 1, ..., M, k is the k-th user, and i and j are the values ​​of h. k The element in the i-th row and j-th column of the matrix, h k，i，j =exp(-(b0+b1f) m )d i，j )g i，j Where b0 and b1 are constants relating to the distance attenuation factor and the frequency attenuation factor, respectively, which can be obtained through a single measurement, d i，j It is h k，i，j The distance from electrical equipment to the multi-access edge server; g i，j is the random fading factor, which is a complex number whose real and imaginary parts follow a standard normal distribution; f is the center frequency of the HPLC subcarrier, and m∈[0,1] is the exponent of the fading factor.

3. The method for physical layer security edge calculation unloading and result transmission of HPLC according to claim 2, characterized in that, Step A also includes: A2, Based on measurements, the statistical characteristics of the channel are obtained using channel estimation methods, i.e. Where d i，j The mean is E(h k，i，j ) is for h k，i，j Take the average.

4. The method for physical layer security edge calculation unloading and result transmission of HPLC according to claim 3, characterized in that, Step A also includes: A3. Establish an offloading and control model. In this model, K users offload through a multiple-input multiple-output (MIMO) channel, and the users are aware of the eavesdropper's channel state information, which can be represented as t. n ∈C M×M C is a complex number, where n = 1, ..., N, and n represents the nth eavesdropper; the signal received at the multi-access edge server is represented as Where f k It is the signal processing matrix for the k-th user, x k For the data uninstalled by the k-th user, n k To reduce noise at multiple access edge servers, w k h is the received signal processing matrix for user k. k This is the transmission channel matrix from the k-th user to the MEC server; the signal received at the n-th eavesdropper can be represented as... t k It is the transmission channel matrix from the k-th user to the n-th eavesdropper.

5. The method for physical layer security edge calculation unloading and result transmission of HPLC according to claim 4, characterized in that, Step C specifically includes: C1, construct the state space S k ={SINR k η c θ k , λ k }, where η c For the information unloading delay, θ k ∈[0,1] represents the proportion of computational data that needs to be unloaded, when θ k When θ = 0, it means all calculations are performed on the user side; when θ = 0, it means all calculations are performed on the user side. k When λ = 1, it means all data needs to be offloaded to the multi-access edge server; k λ is the allocation coefficient for the CPU clock speed of the multi-access edge server. k = (0, 1]; Construct the action space 6. The method for physical layer security edge calculation unloading and result transmission of HPLC according to claim 5, characterized in that, C2, Construct the reward function, R t (θ)=θSLNR k +(1-θ)C s Where θ is the weighting coefficient, and its value is θ = (0, 1], C s For a safe transmission rate, its long-term reward function is: in, It is a weighting factor, thus obtaining the weighted value for different reward times.

7. The method for physical layer security edge calculation unloading and result transmission of HPLC according to claim 6, characterized in that, C3, based on statistical characteristics, calculates the long-term statistical value of the reward function, which can be expressed as:

8. The method for physical layer security edge calculation unloading and result transmission of HPLC according to claim 7, characterized in that, Step D specifically includes: D1 calculates the gradient of the long-term reward function, i.e. Furthermore, based on the Bellman equation, the value of Q is calculated. D2, determine the execution strategy, i.e. The strategy employs the conjugate gradient method to update the reward value. D3, will (s t a t L t s t+1 Stored in temporary memory, with the iteration count set to B. max Definition update s t Let a be the state space at time t. t It is the action space at time t, s t+1 Let be the state space at time t+1.

9. The method for physical layer security edge calculation unloading and result transmission of HPLC according to claim 8, characterized in that, Step E specifically includes: E1, initialize S k ={SINR k η c θ k , λ k }, The identity matrix is ​​used as the matrix value, and non-matrix values ​​are calculated based on this value. E2 clears the memory temporarily storing data and sets its capacity to B. max =1024, the sampled value is B s =256; E3, Obtain from Actions Based on the strategy, determine the timely reward value R. t (θ)=θSLNR k +(1-θ)C s And it is stored in the memory of the data; E4, adopting the latest Where ω k It is a gradient descent strategy; E5, when B < B max Loop through D1 to E4, when B = B max Read the set of actions in memory. and θ in the state set k , λ k .

Images