Reconfigurable smart surface RIS-assisted secure communication method and system

Abstract

This application provides a reconfigurable smart surface (RIS)-assisted secure communication method and system, applied in the field of wireless communication technology. It includes: constructing an undirected weighted graph containing high-altitude platform (HAP) nodes, reconfigurable smart surface (RIS) nodes, ground gateway nodes, ground user nodes, and unauthorized aerial vehicle (AAV) nodes; during communication between the HAP nodes and the ground gateway nodes, constructing an optimization objective function based on the communication beamforming vector, the sensing beamforming vector, and the RIS phase shift matrix; determining the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix that satisfy preset constraints based on the target feature vectors of each node in the undirected weighted graph, and then substituting them into the optimization objective function to obtain the optimal effective concealment rate; and determining that the communication process is secure when the optimal effective concealment rate is greater than a preset rate threshold. This improves the concealment and security of communication between the HAP nodes and the ground gateway nodes.

Description

Technical Field

[0001] This application relates to the field of wireless communication technology, and in particular to a reconfigurable smart surface RIS-assisted secure communication method and system. Background Technology

[0002] As mobile communication technology evolves towards air-to-ground integration, building a three-dimensional wireless communication network with wide-area coverage has become an important development trend. High Altitude Platforms (HAPs), as key aerial nodes, possess characteristics such as long dwell time, wide coverage, and good line-of-sight propagation conditions. They can effectively compensate for the insufficient coverage of ground communication systems in complex environments and have significant application value in scenarios such as emergency communication, wide-area IoT, and low-altitude intelligent networks.

[0003] However, during air-to-ground wireless transmission, due to the long propagation distance and complex environment, signals are susceptible to path loss, scattering fading, and electromagnetic interference, leading to a decline in communication link performance. To address this, Integrated Sensing and Communication (ISAC) technology integrates communication and sensing functions, enabling the system to simultaneously transmit information and perceive its environment and targets, providing support for system optimization in complex scenarios. Due to the openness of wireless channels, communication information is easily detected or intercepted by external nodes. With the gradual opening of low-altitude airspace, autonomous aerial vehicles (AAVs) with maneuverability may pose a potential threat to communication systems, not only acquiring system information but also identifying communication behavior through signal detection, thereby reducing the security and stealth of the communication system.

[0004] Therefore, improving the security and concealment of air-to-ground wireless transmission has become an urgent problem to be solved. Summary of the Invention

[0005] This application provides a secure communication method and system assisted by a Reconfigurable Intelligent Surface (RIS). Through graph theory modeling, complex nodes such as HAPs, Reconfigurable Intelligent Surfaces (RIS), and AAVs are abstracted into an undirected weighted graph, accurately depicting the channel gain and interference correlation in an integrated air-to-ground scenario. By deeply coupling the communication beamforming vector and the sensing beamforming vector of the HAP nodes, and collaboratively controlling the phase shift matrix of the RIS nodes, refined reconstruction of the spatial signal propagation path is achieved. This method effectively enhances the transmission quality of legitimate links while utilizing sensing capabilities to monitor and suppress the illegal interception probability of AAV nodes in real time, significantly improving the system's effective concealment rate. It fundamentally solves the concealment and security problems of communication in high-altitude dynamic environments at the physical layer, exhibiting strong environmental adaptability and anti-interference capabilities.

[0006] This application provides a reconfigurable smart surface RIS-assisted secure communication method, including: Construct an undirected weighted graph containing high-altitude platform (HAP) nodes, reconfigurable intelligent surface (RIS) nodes, ground gateway nodes, ground user nodes, and illegal aerial vehicle (AAV) nodes; During the communication process between the HAP node and the ground gateway node, according to the communication beamforming vector Sensing beamforming vector and RIS phase shift matrix An optimization objective function is constructed, which is used to determine the effective covert rate of the communication process; Based on the target feature vectors of each node in the undirected weighted graph, determine the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix that satisfy the preset constraints. Substituting the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix into the optimization objective function yields the optimal effective concealment rate. The communication process is deemed secure if the optimal effective concealment rate is greater than a preset rate threshold.

[0007] According to an embodiment of this application, a reconfigurable smart surface RIS-assisted secure communication method is provided, wherein the method is based on a communication beamforming vector. Sensing beamforming vector and RIS phase shift matrix Construct an optimization objective function, including: based on the linear receive vector The communication beamforming vector The RIS phase shift matrix The channel vector between the RIS node and the ground gateway node The communication channel between the HAP node and the ground gateway node The channel vector between the RIS node and the ground gateway node. Construct the communication signal power term of the ground gateway node; based on the linear reception vector The sensing beamforming vector and the communication channel matrix between the HAP node and the ground gateway node. Construct the sensing interference power term of the ground gateway node; based on the RIS phase shift matrix... The transmit power of the ground user node, and the communication link between the RIS node and the AAV node. The uplink interference power term of the AAV node to the ground gateway node is constructed based on the channel vector between the RIS node and the ground user node; the optimization objective function is constructed based on the communication signal power term, the sensing interference power term and the uplink interference power term.

[0008] According to an embodiment of this application, a reconfigurable smart surface-assisted RIS-based secure communication method is provided. The step of determining the optimal communication beamforming vector, optimal sensing beamforming vector, and optimal RIS phase shift matrix satisfying preset constraints based on the target feature vectors of each node in the undirected weighted graph includes: extracting features from the initial feature vectors of each node in the undirected weighted graph to obtain a first node embedding vector for each node; and simultaneously inputting the first node embedding vectors of each node into a graph attention network (GAT), wherein the GAT includes... Layered graph attention layer It is an integer greater than 1; in the In the first layer of the graph attention layer, for all nodes, the first... The node is determined to be related to the first node. The node adjacent to the first The node; according to the first node; The first node embedding vector of the nth node and the nth node The first node embedding vector of the nth node determines the nth node in the first layer graph attention layer. The node for the first The attention weights of the nodes are determined, and the attention weights of the nodes in the first-level graph attention layer are updated. The first node embedding vector of the nth node is obtained to obtain the nth node. The second node embedding vector of each node; in the In the graph attention layers, excluding the first graph attention layer, the output of the first graph attention layer is based on the output of the adjacent previous graph attention layer. The node embedding vector and attention weight of the nth node, and the nth node's... The node embedding vector of the nth node is updated in each graph attention layer. The node embedding vector of the nth node is used to obtain the nth node. The embedding vector of the third node of each node; [The sentence is incomplete and requires further context to be translated accurately.] The third node embedding vector of each node output by the last graph attention layer in the layered graph attention layer is used as the target feature vector of each node; based on the target feature vector of each node, the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix that satisfy the preset constraints are determined.

[0009] According to an embodiment of this application, a reconfigurable smart surface RIS-assisted secure communication method is provided, wherein the preset constraints include at least: the total power constraint of the HAP node, and the power constraint on the RIS node. The constraints include the phase constraints of each reflective element, the concealment constraints of communication between the HAP node and the ground gateway node, the minimum sum rate constraints of the ground user node, and the perceived signal-to-noise ratio (SNR) constraints of the AAV node.

[0010] According to an embodiment of this application, a reconfigurable smart surface RIS-assisted secure communication method is provided, wherein the first-layer graph attention layer includes... One's attention, If it is an integer greater than 1, then according to the first... The first node embedding vector of the nth node and the nth node The first node embedding vector of the nth node determines the nth node in the first layer graph attention layer. The node for the first The attention weights of the nodes are determined, and the attention weights of the nodes in the first-level graph attention layer are updated. The second node embedding vector of each node includes: using the... The first attention head The attention head, according to the first attention head, Attention vectors corresponding to each attention head First preset mapping matrix Second preset mapping matrix , and the first The first node embedding vector of each node and the first The first node embedding vector of each node Determine the first attention layer in the first graph. The node for the first Attention coefficient of each node According to the aforementioned first The node for the first Attention coefficient of each node , and the first The attention coefficients of the nth node to its other neighboring nodes are used to determine the nth node. The node for the first Attention weights of each node The remaining adjacent nodes include the first Each node; based on the mapping matrix corresponding to the values ​​of each attention head. The attention heads mentioned above are the first ones. The node for the first Attention weights of each node, and the embedding vector of the first node and the first node embedding vector Update the first attention layer of the graph. The second node embedding vector of each node .

[0011] A reconfigurable smart surface RIS-assisted secure communication method according to an embodiment of this application further includes: constructing a characterization first mode. The first probability distribution of the AAV node received signal and the characterization of the second mode are described below. The second probability distribution of the signal received by the AAV node is described below; based on the channel vector between the HAP node and the AAV node. The sensing beamforming vector The RIS phase shift matrix The transmit power of the ground user node, and the communication link between the RIS node and the AAV node. The channel vector and communication beamforming vector between the RIS node and the ground user node. Determine from the first probability distribution To the second probability distribution KL divergence ; the KL divergence The condition being less than or equal to a preset condition serves as the concealment constraint.

[0012] A reconfigurable smart surface RIS-assisted secure communication method according to an embodiment of this application further includes: based on path gain according to radar ranging equations. Linear receive vector The channel matrix between the HAP node and the AAV node and the sensing beamforming vector Determine the sensing signal-to-noise ratio of the HAP node. The perceived signal-to-noise ratio Less than or equal to the perceived signal-to-noise ratio threshold This serves as the perceived signal-to-noise ratio (SNR) constraint.

[0013] According to an embodiment of this application, a reconfigurable smart surface RIS-assisted secure communication method is provided, the method further comprising: determining the transmit power of the ground user node and the receive vector of the ground gateway node. The channel vector between the RIS node and the ground gateway node The RIS phase shift matrix The uplink rate of the ground user node is determined by the channel vector between the RIS node and the ground user node. The uplink rate Greater than or equal to the minimum sum rate threshold , as the minimum sum rate constraint.

[0014] According to an embodiment of this application, a reconfigurable smart surface RIS-assisted secure communication method is provided, wherein determining the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix that satisfy preset constraints based on the target feature vectors of each node includes: Based on the target feature vectors of each node, construct a node embedding matrix. Embed the nodes into the matrix The input is fed into the beamforming vector decoder, which then processes the node embedding matrix. Feature extraction is performed on the target feature vectors of the HAP nodes and the ground gateway nodes to obtain the optimal communication beamforming vector and the optimal sensing beamforming vector that satisfy the preset constraints; the node embedding matrix is ​​then used to perform feature extraction. The input is fed into the RIS phase-shift decoder, which then processes the node embedding matrix. The target feature vector of the RIS node is used to extract features to obtain the optimal RIS phase shift matrix that satisfies the preset constraints.

[0015] This application also provides a reconfigurable smart surface RIS-assisted secure communication system, including: The model building module is used to construct an undirected weighted graph containing High Altitude Platform (HAP) nodes, Reconfigurable Smart Surface (RIS) nodes, ground gateway nodes, ground user nodes, and unauthorized aerial vehicle (AAV) nodes; and is used to perform beamforming based on communication beamforming vectors during communication between the HAP nodes and the ground gateway nodes. Sensing beamforming vector and RIS phase shift matrix An optimization objective function is constructed, which is used to determine the effective covert rate of the communication process; The parameter optimization module is used to determine the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix that satisfy preset constraints based on the target feature vectors of each node in the undirected weighted graph. The rate determination module is used to substitute the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix into the optimization objective function to obtain the optimal effective concealment rate. The security determination module is used to determine the security of the communication process when the optimal effective concealment rate is greater than a preset rate threshold.

[0016] This application also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the reconfigurable smart surface RIS-assisted secure communication method as described above.

[0017] This application also provides a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the reconfigurable smart surface RIS-assisted secure communication method as described above.

[0018] This application also provides a computer program product, including a computer program that, when executed by a processor, implements a RIS-assisted secure communication method for reconfigurable smart surfaces as described above.

[0019] The reconfigurable smart surface RIS-assisted secure communication method and system provided in this application embodiment utilizes an undirected weighted graph containing a high-altitude platform (HAP) node, a reconfigurable smart surface RIS node, a ground gateway node, a ground user node, and an unauthorized aerial vehicle (AAV) node. During communication between the HAP node and the ground gateway node, the communication beamforming vector is used... Sensing beamforming vector and RIS phase shift matrix An optimization objective function is constructed to determine the effective concealment rate of the communication process. Based on the target feature vectors of each node in the undirected weighted graph, the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix that satisfy preset constraints are determined. The optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix are substituted into the optimization objective function to obtain the optimal effective concealment rate. If the optimal effective concealment rate is greater than a preset rate threshold, the communication process is determined to be secure. By abstracting complex nodes such as HAP, RIS, and AAV into an undirected weighted graph through graph theory modeling, the correlation between channel gain and interference in the air-to-ground integrated scenario is accurately characterized. By deeply coupling the communication beamforming vector and sensing beamforming vector of the HAP node and coordinating the phase shift matrix of the RIS node, a refined reconstruction of the spatial signal propagation path is achieved. This method effectively enhances the transmission quality of legitimate links while using sensing capabilities to monitor and suppress the illegal interception probability of AAV nodes in real time, significantly improving the effective concealment rate of the system. It solves the concealment and security problems of communication in high-altitude dynamic environments from the physical layer, and has strong environmental adaptability and anti-interference capabilities. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in this application 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 this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a schematic diagram of the integrated sensing secure collaborative network model provided in the embodiments of this application; Figure 2 This is a flowchart illustrating the reconfigurable smart surface RIS-assisted secure communication method provided in an embodiment of this application. Figure 3 This is a schematic diagram showing the comparison curves of the effective concealment rate as a function of the number of iterations provided in the embodiments of this application; Figure 4 This is a schematic diagram of the convergence trajectory of the effective concealment rate provided in this application embodiment under the influence of different learning rates; Figure 5 This is a schematic diagram of the reconfigurable smart surface RIS-assisted secure communication system provided in the embodiments of this application; Figure 6 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] To better understand the embodiments of this application, the prior art will first be described in detail: Existing technologies primarily rely on physical layer security mechanisms, employing methods such as beamforming, artificial noise injection, and resource allocation optimization to enhance the channel difference between legitimate and eavesdropping links, thereby strengthening the secure transmission capabilities of communication systems. However, in complex air-to-ground collaborative scenarios, existing technologies still have significant shortcomings. First, for passive eavesdropping scenarios, existing technologies assume that AAV nodes only possess information reception capabilities, ignoring their ability to actively detect communication behavior, making it difficult to effectively reduce the risk of communication being detected. Second, AAV node modeling typically assumes that their positions are known or static, making it difficult to characterize the dynamic behavior of highly mobile AAVs, resulting in the inability to achieve continuous tracking and high-precision estimation of target states in real-world environments. Furthermore, due to the high coupling between communication beams, sensing beams, and RIS phase shift parameters, the resulting optimization problem is usually a high-dimensional non-convex problem. Traditional optimization methods have limitations in solution efficiency, convergence performance, and global optimality, making it difficult to meet the needs of real-time optimization in dynamic environments.

[0024] In other words, existing methods have certain limitations and still cannot improve the security and concealment of air-to-ground wireless transmission processes.

[0025] To address the aforementioned technical issues, the reconfigurable smart surface RIS-assisted secure communication method provided in this application abstracts complex nodes such as HAP, RIS, and AAV into an undirected weighted graph through graph theory modeling, accurately characterizing the channel gain and interference correlation in an integrated air-to-ground scenario. By deeply coupling the communication beamforming vector and sensing beamforming vector of the HAP node and coordinating the phase shift matrix of the RIS node, fine-grained reconstruction of the spatial signal propagation path is achieved. This method effectively enhances the transmission quality of legitimate links while using sensing capabilities to monitor and suppress the illegal interception probability of AAV nodes in real time, significantly improving the effective concealment rate of the system. It fundamentally solves the concealment and security problems of communication in high-altitude dynamic environments from the physical layer, exhibiting strong environmental adaptability and anti-interference capabilities.

[0026] It should be noted that the execution entity involved in the embodiments of this application can be a reconfigurable smart surface RIS-assisted secure communication system or an electronic device. Optionally, the electronic device may include: a computer / laptop, a mobile terminal, a server, an airborne processing unit or a control chip with computing processing functions, etc.

[0027] To better understand the embodiments of this application, an electronic device is used as an example. A sensor-integrated secure collaborative network model can be constructed first, serving as the foundation for implementing the reconfigurable smart surface RIS-assisted secure communication method provided in the embodiments of this application. The construction process of this sensor-integrated secure collaborative network model is as follows: Figure 1 This is a schematic diagram of the integrated sensing secure collaborative network model provided in an embodiment of this application. For example... Figure 1 As shown, the entity participants (also known as physical entities) in this integrated sensing and communication secure collaborative network model within the communication environment consist of HAP, RIS, ground gateway, and It consists of ground users and AAVs acting as threat nodes. Each entity participant is equipped with technical parameters, as detailed below: HAP equipped with One antenna, ground gateway equipped A RIS antenna is provided, equipped with M reflective elements, and these M reflective elements are deployed on the surface of a building. The phase shift matrix of the RIS is defined as follows. , , , Both M and are integers greater than 1. Subsequently, the electronic device acquires and establishes the three-dimensional spatial coordinates of each participant entity, where the coordinates of HAP are... The coordinates of the ground gateway are The coordinates of RIS are ; Among the ground users, the first The coordinates of the ground users are Since the three-dimensional position of the AAV changes over time, the coordinates of the AAV are defined as follows: , Indicates the current moment.

[0028] Based on physical layer transmission theory, electronic devices establish signal transmission models and performance evaluation indicators according to the physical attributes of each node in the integrated sensing and communication security collaborative network model.

[0029] Given that the AAV has monitoring capabilities and can detect communication between the HAP and the ground gateway, the HAP needs to balance sensing accuracy and communication stealth. Therefore, the signals transmitted by the HAP operate in two modes: one transmits only a sensing beam for real-time sensing of the AAV's spatial position, and the other simultaneously transmits signals to the ground gateway to issue commands. It can be represented as: .in, This indicates that the HAP only sends sensing signals to the ground gateway without any communication behavior, also known as the first mode; This indicates that the HAP is simultaneously sending sensing signals and communication signals to the ground gateway, also known as the second mode; Represents the sensing beamforming vector; Represents the communication beamforming vector; Indicates the perceived signal; This indicates the communication signal transmitted by the HAP to the ground gateway.

[0030] For ground network groups Each ground user uploads a signal to the ground gateway, which collects the ground signals. It can be represented as: .in, Indicates the first The transmit power of each ground user node, Indicates the relationship between RIS nodes and the first Channel vectors between ground user nodes; This represents the channel vector between the RIS node and the ground gateway node; Indicates the first Signals sent by individual users; This represents the noise matrix encountered by the ground gateway when receiving signals, and It follows a pattern with a mean of 0 and a covariance matrix of . The Gaussian distribution, i.e. ; Indicates noise power.

[0031] Based on this, the signals received by the ground gateway include signals from the airborne HAP and signals uploaded from the ground network, corresponding to two modes. It can be represented as: .in, This represents the communication channel matrix between the HAP node and the ground gateway node.

[0032] Under the constraints of covert communication, after receiving the signal, the ground gateway uses a linear combiner to perform matched filtering on the signal. Therefore, the covert communication rate between the HAP and the ground gateway can be expressed as: .in, , Indicates the receive vector. Represents the receive vector The conjugate matrix, and .

[0033] When ground users transmit uplink data to the ground gateway, AAV can easily capture the uplink radiated signal from the ground terminal. Based on this, the first... achievable speed for individual ground users It can be represented as: AAV eavesdropping rate It can be represented as: .in, This represents the communication link between the RIS node and the AAV node, and .

[0034] Understandably, by constructing a sensor-integrated secure collaborative network model, electronic devices have established a three-dimensional spatial network architecture that includes HAP, RIS, ground gateways, ground users, and AAV. This provides a unified mathematical modeling foundation and a quantifiable performance evaluation framework for achieving secure communication that balances sensing accuracy and communication concealment.

[0035] It should be noted that before constructing the integrated sensing and sensing security collaborative network model, the temporal correlation of continuous multi-slot sensing observation data is mined using the factor graph optimization method to overcome the nonlinearity and uncertainty of dynamic target tracking and obtain accurate state estimates of illegal aircraft. Specifically: First, based on the construction of a sensing-integrated secure collaborative network model, electronic devices, for HAPs with sensing capabilities, can receive echo signals from AAVs. The echo signal It can be represented as: .in, Represents the complex exponent term; Represents the array response vector between HAP and AAV, and , This indicates the horizontal azimuth angle of HAP relative to AAV. This indicates the pitch angle of the HAP relative to the AAV; This indicates the Doppler shift caused by AAV motion; Let represent the noise matrix, and the noise matrix is... It follows a pattern with a mean of 0 and a covariance matrix of . The Gaussian distribution, i.e. ; This represents the path gain based on the radar ranging equation. The calculation formula is: , This represents the radar cross section of the AAV. This represents the free space path loss between HAP nodes and AAV nodes; This represents a continuously observed variable, i.e., the current moment. This represents the round-trip time delay between the transmitted signal and the echo signal. The calculation formula is: ; This represents the straight-line distance between HAP and AAV; It represents the speed of light.

[0036] Next, the electronic device uses a matched filtering method to filter the echo signal. Processing is performed to extract the transmitted and echo signals. Round-trip delay This leads to the deduction of the three-dimensional distance between HAP and AAV. The electronic equipment uses the Multiple Signal Classification (MUSIC) algorithm to perform super-resolution spatial spectrum estimation on the array-received signal, and extracts the horizontal azimuth angle of the AAV relative to the HAP. and pitch angle Subsequently, the electronic device processed the echo signal. Perform a Fourier transform to extract the Doppler frequency shift caused by AAV motion. Therefore, the electronic device can obtain In the first time slot A four-dimensional observation vector at each time slot / moment This can be expressed as: .in, , It is an integer greater than 1.

[0037] Subsequently, the electronic device defines the AAV to be tracked in the first... The state vector at time n ,and Assuming the AAV's motion follows a constant velocity model, based on this state vector... The state evolution process of constructing AAV is as follows ,in, This represents the state transition matrix. It can be represented as: , Indicates the sampling interval between adjacent time points. Represents a 3rd order identity matrix. Indicates process noise, and It follows a pattern with a mean of 0 and a covariance matrix of . The Gaussian distribution, i.e. covariance matrix The covariance matrix is ​​obtained by mapping the power spectral density q of continuous-time acceleration white noise. It can be represented as: .

[0038] Next, the electronic device determines the location of the HAP itself. and speed, for the first At each time point, calculate the relative position vector between HAP and AAV. and horizontal distance relative position vector The calculation formula is: Horizontal distance The calculation formula is: .in, These represent the coordinate differences between HAP and AAV in the x, y, and z axes, respectively.

[0039] Subsequently, the electronic device follows the state vector. To predict the first The measurement vectors corresponding to the four-dimensional observation vectors at each time point, i.e., the four measurement vectors, are used to construct the measurement prediction function. The measurement prediction function It can be represented as: .in, This indicates that HAP and AAV are in the first... The three-dimensional distance at a given moment. The calculation formula is: ; Indicates that AAV is in the The velocity vector at each instant; This represents the velocity vector of the HAP.

[0040] Finally, the electronic equipment is based on four-dimensional observation vectors and measurement prediction functions. The state estimation problem of an AAV is modeled as a factor graph containing two types of factors: motion factors and measurement factors. Because there are errors between the predicted state values ​​and the actual motion model (i.e., the state transition model), and between the observed predicted values ​​and the actual observed values, the electronic device introduces two types of residual vectors to characterize these uncertainties. Specifically, the electronic device defines the measurement residual vector of the motion factors. This describes the discrepancy between the predicted state and the actual motion model, and measures the residual vector. It can be represented as: And assign the measurement residual vector A probability density function probability density function It can be represented as: Simultaneously, the electronic device defines the measurement residual vector of the measurement factor. To describe the deviation between actual and predicted observations, the residual vector is measured. It can be represented as: And assign the measurement residual vector A probability density function probability density function It can be represented as: ,in, This represents the pre-calibrated observation noise covariance matrix. It can be represented as: .in, This represents the variance of distance observation noise. This represents the variance of azimuth observation noise. This represents the variance of the observation noise at the pitch angle. This represents the variance of the Doppler frequency shift observation noise.

[0041] Based on this, the electronic device assumes that the motion factor and the measurement factor are statistically independent, and that the noise at different times is independent. Therefore, the joint posterior probability can be decomposed as follows: .in, This represents the prior knowledge of the AAV state at time 0, i.e., the initial prior factor. Subsequently, the electronic device transforms the AAV state estimation problem into a maximum likelihood estimation problem, which can be defined as: .

[0042] To facilitate numerical solutions, the electronic device takes the negative logarithm of the joint posterior probability and omits the constant term, transforming the maximum likelihood estimation problem into a weighted nonlinear least squares problem. Based on this, the state estimation problem of AAV can be redefined as: The first term is the initial prior residual. This represents the initial mean. The first term represents the initial covariance; the second term represents the motion factor residual; and the third term represents the measurement factor residual. Next, the electronic device, based on a Gauss-Newton iterative solution strategy, performs first-order linearization on the nonlinear measurement model in each iteration and constructs a globally sparse normal equation until the objective function converges. Finally, it solves the aforementioned nonlinear least squares problem and outputs the state estimate of the entire AAV trajectory as follows: State vector Indicates that AAV is in the The precise three-dimensional position and precise three-dimensional velocity at each moment.

[0043] Understandably, by using factor graph optimization algorithms to mine the temporal correlations of multi-slot observation data, the accuracy and robustness of AAV state estimation under complex dynamic environments are significantly improved, providing high-quality prior state support for secure communication and accurate perception decision-making in subsequent communication systems.

[0044] The following uses an electronic device as an example to illustrate in detail the reconfigurable smart surface RIS-assisted secure communication method provided in the embodiments of this application: Figure 2 This is a flowchart illustrating a RIS-assisted secure communication method for reconfigurable smart surfaces provided in an embodiment of this application. Figure 2 As shown, the method includes the following steps 201-205.

[0045] Step 201: Construct an undirected weighted graph containing High Altitude Platform (HAP) nodes, Reconfigurable Intelligent Surface (RIS) nodes, Ground Gateway nodes, Ground User nodes, and Unauthorized Aircraft (AAV) nodes.

[0046] Among them, the high-altitude platform (HAP) node refers to the airborne communication base station carrier deployed in the stratosphere (usually at an altitude between 17 and 55 kilometers).

[0047] A RIS node is a planar array composed of a large number of low-power, low-cost metamaterial units.

[0048] A ground gateway node is an infrastructure node that connects the space segment (such as HAP or satellite) to the ground core network.

[0049] Ground user nodes refer to legitimate terminal devices located within the ground communication coverage area.

[0050] Optionally, the ground user node can be a mobile handheld terminal, an IoT sensing node, or a vehicle-mounted communication unit, etc.

[0051] Optionally, the number of ground user nodes is indivual, It is an integer greater than 1.

[0052] An AAV node is a flight target that operates unauthorizedly in the airspace and may eavesdrop on or maliciously interfere with the communication system constructed by the high-altitude platform HAP node and the ground gateway node.

[0053] Optionally, the electronic device constructs an undirected weighted graph containing high-altitude platform (HAP) nodes, reconfigurable smart surface (RIS) nodes, ground gateway nodes, ground user nodes, and illegal aircraft (AAV) nodes, including: the electronic device abstracts the sensor-integrated security collaboration network model into an undirected weighted graph containing high-altitude platform (HAP) nodes, reconfigurable smart surface (RIS) nodes, ground gateway nodes, ground user nodes, and illegal aircraft (AAV) nodes.

[0054] In this embodiment of the application, the electronic device will integrate the HAP, RIS, ground gateway, AAV, and other components in the space. These physical entities, including ground users, are mapped to vertices in an undirected weighted graph, where each node carries its own spatial location information. Because it is an undirected weighted graph, electronic devices establish undirected edges between nodes and define the weights of each edge based on the physical properties of communication or interference, ultimately obtaining the undirected weighted graph. .in, This represents a set of nodes, including HAP nodes, RIS nodes, AAV nodes, ground gateways, and... There are one ground user node; and the number of graph nodes is set to [number]. ; Represents the set of edges.

[0055] Optionally, the electronic device uses a Channel-Aware Graph Neural Network (CA-GNN) to perform feature association on the undirected weighted graph and extract deep feature vectors that can characterize air-to-ground cooperative interference and communication gain.

[0056] Understandably, by using CA-GNN to aggregate channel state information and spatial geometric constraints between nodes, and mapping the physical attributes of heterogeneous nodes to a high-dimensional potential feature space, global perception and deep feature purification of complex air-to-ground interference environments can be achieved. This allows for the accurate quantification of nonlinear channel interaction relationships between heterogeneous nodes, providing highly robust and topologically adaptable feature support for subsequent joint optimization of beamforming and phase shift matrices.

[0057] Based on this, the feature vectors of each node in the undirected weighted graph can be obtained: in, Represents the feature vector of a HAP node. This represents the normalized maximum power threshold. This represents the number of antennas in the HAP node after normalization. The feature vector representing the ground gateway node. This indicates the number of antennas at the ground gateway node after normalization. The feature vector representing the RIS node. This indicates the number of RIS reflective elements after normalization; Represents the feature vector of an AAV node; Indicates the first Feature vectors of ground user nodes Indicates the normalized i-th The transmit power of each ground user node, This represents the channel norm statistics, and the last digit of each node's feature vector represents the node type identifier.

[0058] Therefore, the characteristic matrix can be obtained. The feature matrix It includes the physical attributes and channel statistics of each node in the undirected weighted graph.

[0059] It should be noted that this step maps heterogeneous physical entities across spatial domains, such as HAPs, RISs, ground gateways, ground users, and non-AAVs, into undirected weighted graphs. This transforms complex spatial relationships and electromagnetic propagation characteristics into vertex and edge weights in graph theory, effectively solving the problem of quantifying the communication and interference coupling relationships between multiple nodes. Simultaneously, it lays a solid mathematical foundation for subsequent use of graph neural networks to achieve optimal beamforming and phase shift control.

[0060] Step 202: During the communication process between the HAP node and the ground gateway node, based on the communication beamforming vector... Sensing beamforming vector and RIS phase shift matrix An optimization objective function is constructed, which is used to determine the effective covert rate of the communication process.

[0061] Among them, the communication beamforming vector It refers to the communication signal weight vector generated by the antenna array of the HAP node and used to carry service data.

[0062] Sensing beamforming vector It refers to the sensing signal weight vector generated by the antenna array of the HAP node for detecting the space environment.

[0063] RIS phase shift matrix It refers to the diagonal matrix formed by the phase shift coefficients of each reflection unit on the RIS node.

[0064] Effective covert rate refers to the maximum reliable transmission rate that a ground gateway node can achieve under the premise that the probability of AAV node detecting legitimate communication signals is lower than a preset threshold. This effective covert rate is used to characterize the information transmission capability between HAP node and ground gateway node in order to ensure that the communication behavior is not detected by a third party (i.e., AAV node).

[0065] In some embodiments, the electronic device is based on a communication beamforming vector. Sensing beamforming vector and sensing beamforming vector Construct an optimization objective function to determine the effective covert rate of the communication process. This objective function may include: electronic devices based on linear reception vectors... Communication beamforming vector RIS phase shift matrix Channel vector between RIS node and ground gateway node Communication channel between HAP node and ground gateway node The channel vector between the RIS node and the ground gateway node. Construct the communication signal power term of the ground gateway node; electronic equipment according to the linear reception vector Sensing beamforming vector and the communication channel matrix between HAP nodes and ground gateway nodes. Construct the sensing interference power term for the ground gateway node; electronic equipment based on the RIS phase shift matrix. Transmit power of ground user nodes, and communication links between RIS nodes and AAV nodes. The uplink interference power term of the AAV node to the ground gateway node is constructed by considering the channel vector between the RIS node and the ground user node. Finally, the electronic equipment constructs an optimization objective function based on the communication signal power term, the sensed interference power term, and the uplink interference power term.

[0066] Optionally, the expression for the communication signal power term of the ground gateway node is: .

[0067] Optionally, the expression for the sensed interference power term of the ground gateway node is: .

[0068] Optionally, the expression for the uplink interference power term of the AAV node to the ground gateway node is: .

[0069] Optionally, the expression for the objective function can be: .

[0070] It should be noted that this step, by constructing an optimization objective function that includes communication signal power, sensing interference power, and AAV node uplink interference power, integrates the communication and sensing beamforming of the HAP node and the phase shift control of the RIS node into the problem of maximizing the effective concealment rate, thereby improving the communication concealment of the ground gateway while suppressing AAV eavesdropping detection.

[0071] Step 203: Based on the target feature vectors of each node in the undirected weighted graph, determine the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix that satisfy the preset constraints.

[0072] The target feature vector refers to the high-dimensional representation vector of each node obtained after extracting features from the undirected weighted graph using CA-GNN.

[0073] Optionally, the preset constraints include at least: total power constraints of the HAP node, and constraints on the RIS node. The constraints include phase constraints of each reflective element, concealment constraints of communication between HAP nodes and ground gateway nodes, minimum sum rate constraints of ground user nodes, and perceived signal-to-noise ratio (SNR) constraints of AAV nodes.

[0074] The total power constraint of the HAP node refers to the upper limit of the total transmit power that the radio frequency link can provide when the HAP node is performing communication and sensing tasks simultaneously.

[0075] Optionally, the total power constraint of the HAP node can be expressed as: .in, This indicates the maximum transmission power.

[0076] Understandably, by limiting the total transmit power of the sensing beam and the communication beam to no more than the rated threshold, nonlinear distortion of the hardware circuits in the HAP node is effectively avoided, and while ensuring the system's energy efficiency, the probability of being detected by the AAV node due to excessive power is prevented.

[0077] On RIS nodes The phase constraint of a reflective element refers to the range of phase adjustment that each reflective unit of the RIS can achieve for the incident electromagnetic wave.

[0078] Optionally, on RIS nodes The phase constraint of a single reflecting element can be expressed as: .

[0079] Understandably, this phase constraint ensures that the reflecting element only adjusts the phase of the incident signal without changing the amplitude, thereby maximizing the use of electromagnetic wave energy and achieving directional enhancement or physical layer cancellation of air-to-ground signals through precise cooperative beamforming.

[0080] The concealment constraint of communication between HAP nodes and ground gateway nodes refers to the restrictions that ensure legitimate communication activities are not effectively detected by AAV nodes.

[0081] In some embodiments, the process of obtaining the concealment constraint is as follows: the electronic device constructs a representation of the first mode. First probability distribution of the received signal at the lower AAV node and characterizing the second mode Second probability distribution of the received signal at the lower AAV node The electronic device is based on the channel vector between the HAP node and the AAV node. Sensing beamforming vector RIS phase shift matrix Transmit power of ground user nodes, and communication links between RIS nodes and AAV nodes. Channel vector between RIS nodes and ground user nodes, and communication beamforming vector Determine from the first probability distribution To the second probability distribution KL divergence This electronic device will measure the KL divergence. Less than or equal to the preset condition, serving as a hidden constraint.

[0082] Alternatively, in the two different modes, the received signal of the AAV node can be represented as: .

[0083] The AAV node determines whether the HAP and the ground gateway are communicating based on the detection statistics of the received signal.

[0084] Optionally, based on the detection statistics of the AAV node's received signal, the detection decision of the AAV node can be expressed as: .

[0085] in, The sampling sequence of AAV, i.e. , This represents the detection statistics of the signal received by the AAV node, i.e., the average power of multiple independently received signal samples. It is the preset detection threshold of the AAV node. This indicates that the HAP node is sending communication signals to the ground gateway node. This indicates that the HAP node only sends sensing signals to the ground gateway node.

[0086] In actual signal detection, the AAV node's detection decisions may yield erroneous results, namely false alarms and missed detections. A false alarm indicates that there is no communication between the HAP node and the ground gateway node, but the AAV node detects that communication exists between them. A missed detection indicates that the HAP node sent a communication signal, but the AAV node failed to detect it.

[0087] Based on the two error scenarios, the detection error probability of the AAV node can be expressed as: .

[0088] in, It is the prior probability in the case of a false alarm. This represents the prior probability in the case of a missed detection. Indicates the probability of a false alarm. This indicates the probability of a missed detection.

[0089] To achieve concealment between HAP nodes and ground gateway nodes, the minimum detection error probability constraint of AAV nodes is transformed into one that follows the first probability distribution. To the second probability distribution KL divergence constraint.

[0090] Optionally, KL divergence The calculation formula is: .

[0091] in, Indicates the first mode The total power under the first mode The formula for calculating the total power is: . Indicates the second mode The total power under the second mode The formula for calculating the total power is: .

[0092] Alternatively, the concealment constraint can be expressed as: .in, This indicates the preset level of communication security concealment.

[0093] It should be noted that by transforming the complex AAV node detection error probability constraint into a quantitative index based on KL divergence, fine-grained control of the communication imperceptibility level is achieved, establishing a rigorous mathematical judgment criterion for covert communication.

[0094] The minimum sum rate constraint of ground user nodes refers to the requirement that, in order to meet the basic communication needs of multiple legitimate ground users, the system must ensure that the sum of the transmission rates of all ground user nodes is not lower than the minimum bandwidth threshold required by the service during the optimization process.

[0095] In some embodiments, the minimum sum rate constraint is obtained as follows: the electronic device determines the minimum sum rate constraint based on the transmit power of the ground user node and the receive vector of the ground gateway node. Channel vector between RIS node and ground gateway node RIS phase shift matrix And the channel vector between the RIS node and the ground user node, to determine the uplink rate of the ground user node. The electronic device will increase the uplink speed. Greater than or equal to the minimum sum rate threshold , as a minimum sum rate constraint.

[0096] Optionally, the uplink rate of the ground user node The calculation formula is: .

[0097] Alternatively, the minimum sum rate constraint can be expressed as: .

[0098] in, This indicates the preset minimum sum rate threshold.

[0099] It should be noted that by establishing a hard performance boundary based on minimum sum and rate threshold, while ensuring the quality of service for ground users, the excessive crowding of communication resources by sensing tasks is effectively prevented, and the fair and efficient allocation of sensing resources is achieved under the premise of ensuring user experience.

[0100] The signal-to-noise ratio (SNR) constraint of AAV nodes refers to the requirement that, in order to achieve accurate positioning and tracking of AAV nodes, the signal-to-noise ratio of the echo signal after the sensing signal transmitted by the HAP node is reflected by the AAV node and returns to the receiver must reach a certain strength threshold.

[0101] In some embodiments, the process of obtaining the perceived signal-to-noise ratio (SNR) constraint is as follows: the electronic device obtains the path gain based on the radar ranging equation. Linear receive vector Channel matrix between HAP nodes and AAV nodes and sensing beamforming vector Determine the perceived signal-to-noise ratio of the HAP node. The electronic device will sense the signal-to-noise ratio. Less than or equal to the perceived signal-to-noise ratio threshold As a constraint on the perceived signal-to-noise ratio (SNR).

[0102] Optionally, the perceived signal-to-noise ratio of the HAP node The calculation formula is: .

[0103] in, This represents the additive white Gaussian noise power at the HAP node's sensing receiver.

[0104] Optionally, the perceived signal-to-noise ratio (SNR) constraint can be expressed as: .

[0105] It should be noted that establishing a sensing signal-to-noise ratio constraint model based on the radar ranging equation ensures the detection quality of the echo signal from the HAP node to the AAV node, thereby guaranteeing the extraction accuracy of multi-dimensional observation vectors such as three-dimensional distance, horizontal azimuth angle, elevation angle, and Doppler frequency shift, and realizing reliable tracking and identification of the AAV node.

[0106] In some embodiments, the electronic device determines the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix that satisfy preset constraints based on the target feature vectors of each node in the undirected weighted graph. This may include: the electronic device extracting features from the initial feature vectors of each node in the undirected weighted graph to obtain the first node embedding vector of each node; and simultaneously inputting the first node embedding vectors of each node into a graph attention network (GAT), wherein the GAT includes... Layered graph attention layer Integers greater than 1; In the first layer of the graph attention layer, the electronic device targets the first node among all nodes. The node is determined to be related to the node. The node adjacent to the first The electronic device is based on the node; The first node embedding vector of the nth node and the nth node The first node embedding vector of the nth node determines the nth node in the attention layer of the first-level graph. For the node, the first node The attention weights of the nth node are determined, and the attention weights of the nth node in the first-level graph are updated. The first node embedding vector of the nth node is obtained to obtain the nth node. The second node embedding vector of each node; in In the graph attention layers other than the first graph attention layer, the electronic device, based on the output of the adjacent previous graph attention layer, [does something]. The node embedding vectors and attention weights of the nth node, and the nth node's attention weights. The node embedding vector of the nth node is used to update the nth node in the attention layer of each graph. The node embedding vector of the node is obtained. The third node embedding vector of each node; the electronic device will The third node embedding vector of each node output by the last layer of the graph attention layer is used as the target feature vector of each node. Based on the target feature vector of each node, the electronic device determines the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix that satisfy the preset constraints.

[0107] The initial feature vector refers to the feature vector of each node in the undirected weighted graph mentioned above.

[0108] In this embodiment of the application, the electronic device will use the feature vectors of each node. By mapping to a high-dimensional latent space through linear transformation, the first node embedding vector of each node is generated. ,and , This represents the number of neurons in the hidden layer of the GAT, and The number of neurons in the hidden layer It is positively correlated with the number of reflective elements in the RIS node and can be adaptively adjusted according to the number of reflective elements in the RIS node.

[0109] Next, the electronic device embeds the first node of each node into a vector. The input is fed into a GAT that includes an L-layer graph attention layer. In the first attention layer, the electronic device targets the first node among all nodes. Find the node with the first node. The node adjacent to the first The node, according to the first The node and the first The embedding vector of the first node corresponding to the nth node is calculated to obtain the nth node. For the node, the first node Attention weights of each node Subsequently, the electronic device performs weighted aggregation of features from adjacent nodes based on this attention weight, and updates the first attention layer. The first node embedding vector of the nth node is obtained to obtain the nth node. The second node embedding vector of each node.

[0110] exist In all graph attention layers except the first graph attention layer, the last layer (i.e., the first...) is the most important. Taking a layer as an example, the electronic device uses the attention layer (i.e., the adjacent upper layer) as a reference. The output of the first layer) Node embedding vectors of nodes and corresponding number Attention weights of each node , and passed Node embedding vectors of neighboring nodes By using weighted summation and combining it with an activation function for nonlinear mapping, the current graph attention layer is updated to obtain the th _ ... The node embedding vector of the nth node is obtained. The third node embedding vector of each node .

[0111] Finally, the electronic device embeds the third node corresponding to each node output from the last graph attention layer into an embedding vector. Feature fusion is performed to obtain the target feature vectors of each node.

[0112] It should be noted that, through The iterative processing of the layered graph attention layer can capture the high-order nonlinear correlations between nodes, transform the original physical features into target feature vectors that can characterize complex channel coupling, and provide feature support with global spatial topology information for the subsequent decision network. Thus, under the premise of ensuring perception accuracy and concealment constraints, it can achieve joint rapid optimization and accurate decision-making of communication, perception and RIS resources.

[0113] In some embodiments, the electronic device is based on the first The first node embedding vector of the nth node and the nth node The first node embedding vector of the nth node determines the nth node in the attention layer of the first-level graph. For the node, the first node The attention weights of the nth node are determined, and the attention weights of the nth node in the first-level graph are updated. The second node embedding vector of each node can include: the electronic device adopts... The first attention head The first attention head, according to the first Attention vectors corresponding to each attention head First preset mapping matrix Second preset mapping matrix , and passed The first node embedding vector of each node and the The first node embedding vector of each node Determine the first attention layer in the first-level graph. For the node, the first node Attention coefficient of each node The electronic device is based on the first For the node, the first node Attention coefficient of each node , and passed The attention coefficients of the nth node to its other neighboring nodes determine the nth node's attention coefficients. For the node, the first node Attention weights of each node The remaining adjacent nodes include the first Each node; the electronic device is based on a mapping matrix corresponding to the values ​​of each attention head. Each person's attention is focused on the next step. For the node, the first node Attention weights of each node, embedding vector of the first node and the first node embedding vector Update the attention layer of the first-level graph. The second node embedding vector of each node .

[0114] It should be noted that the process of updating the second node embedding vector in the first-layer graph attention layer is similar to that in other layers.

[0115] For example, with the first The first layer Taking the first attention point as an example, the electronic device determines the first... For the node, the first node The formula for calculating the attention coefficient of each node is: .

[0116] in, Represents a non-linear activation function. Indicates the first Layer The node embedding vector of each node. Indicates the first Layer The node embedding vector of each node. This indicates a vector concatenation operation.

[0117] Optionally, the electronic device determines the first For the node, the first node Attention weights of each node The calculation formula is: .

[0118] in, Indicates the first The first layer Under the first attention, the first Each node pairs with its neighboring nodes. Attention coefficient Indicates the first The set of neighboring nodes of a node.

[0119] Optionally, the electronic device is updated. The first in the layer Node embedding vectors of nodes The calculation formula is: .

[0120] in, Representation layer normalization operator, Indicates the first The first layer The mapping matrix corresponding to each attention head value is used to perform a linear transformation on the feature vectors of neighboring nodes.

[0121] Understandably, by transforming the node embedding vector using the attention vector and a pre-defined mapping matrix, it is possible to identify the differences in the influence weights of different neighboring nodes on the central node, and to use these methods in parallel. The single attention head can observe the network state from multiple independent subspaces simultaneously, greatly enhancing the completeness and robustness of feature representation. In addition, combined with layer normalization and feedforward network structure, it effectively accelerates the convergence speed of the model and avoids the gradient vanishing problem in deep network training, ensuring the accuracy of resource allocation decisions in complex interference environments.

[0122] In some embodiments, the electronic device determines the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix that satisfy preset constraints based on the target feature vectors of each node. This may include: the electronic device constructing a node embedding matrix based on the target feature vectors of each node. The electronic device embeds nodes into a matrix. The input is fed into the beamforming vector decoder, which then processes the node embedding matrix. Feature extraction is performed on the target feature vectors of the HAP nodes and the ground gateway nodes to obtain the optimal communication beamforming vector and the optimal sensing beamforming vector that satisfy preset constraints; the electronic device embeds the nodes into a matrix. The input is fed into the RIS phase-shift decoder, which then processes the node embedding matrix. Feature extraction is performed on the target feature vectors of the RIS nodes to obtain the optimal RIS phase shift matrix that satisfies the preset constraints.

[0123] Among them, the beamforming vector decoder is used to extract phase and amplitude correlations from the node features of HAP and ground gateway, and reconstruct complex domain beam vectors that meet transmit power constraints and sensing signal-to-noise ratio constraints.

[0124] The RIS phase-shift decoder is used to output a set of diagonal phase-shift matrix elements that satisfy constant modulus constraints and maximize channel gain or concealment based on the target feature vector of the RIS node.

[0125] In this embodiment, the electronic device fuses the target feature vectors of each node to obtain an embedding matrix of all nodes. All nodes are embedded in the matrix It can be represented as: .

[0126] Subsequently, the electronic device utilizes a multi-task learning architecture to configure the beamforming vector decoder and the RIS phase-shift decoder in parallel, embedding the node into the matrix respectively. The inputs are respectively fed into the beamforming vector decoder and the RIS phase shift decoder.

[0127] The electronic device extracts the node embedding matrix through a beamforming vector decoder. The embedding vectors of the HAP node and the ground gateway node are concatenated to obtain a joint feature vector. Then, the electronic device uses a linear layer combined with a Softmax activation function to predict the power allocation ratio between communication power and sensing power, i.e., the power allocation factor. Next, the electronic device performs nonlinear mapping through a two-layer perceptron containing a hidden layer to predict the joint feature vector separately, obtaining the original real and imaginary vectors of the communication beam and the sensing beam. Finally, the electronic device normalizes and scales the original real and imaginary vectors of the communication beam and the sensing beam using the power allocation factor, and reconstructs them into complex domain vectors, obtaining the optimal communication beamforming vector and the optimal sensing beamforming vector that satisfy preset constraints.

[0128] RIS phase-shift decoder only targets the node embedding matrix The electronic device processes the target feature vector belonging to the RIS node. It uses a RIS phase-shift decoder to perform deep feature extraction on the target feature vector of the RIS node, generating a real-valued vector of length M. Then, it uses the Tanh activation function to compress the value range of this real-valued vector to the (-1, 1) interval. Subsequently, the electronic device multiplies this real-valued vector by... A deterministic offset (such as a preset reference phase or a random perturbation term) is introduced to convert the abstract features into the initial phase value of each reflective element at the physical level. Finally, the electronic device maps the initial phase values ​​of all reflective elements onto the unit circle of the complex plane, constructs a diagonal phase shift matrix with diagonal elements having a modulus strictly equal to 1, and obtains the optimal RIS phase shift matrix that satisfies the preset constant modulus constraint.

[0129] It should be noted that this step generates the optimal communication beamforming vector and the optimal sensing beamforming vector, along with the optimal RIS phase shift matrix, through two independent but end-to-end jointly trained sub-decoders. This achieves collaborative decision-making between the communication sensing beam and the physical phase of the RIS node. While ensuring that preset constraints such as transmit power and constant modulus are met, the end-to-end joint training mechanism significantly reduces the solution latency and computational overhead of resource optimization problems in complex air-to-ground scenarios. Furthermore, the beamforming vector decoder effectively avoids the numerical instability problems commonly encountered by neural networks when directly processing complex number operations through independent prediction of real and imaginary parts and subsequent complex number reconstruction. The RIS phase shift decoder, by introducing Dropout operations and Tanh activation mapping, not only enhances the nonlinear fitting accuracy of the model in complex channel environments and significantly mitigates the overfitting risk in large-scale reflector scenarios, ensuring the algorithm's robustness to dynamic environments, but also effectively solves phase ambiguity and clustering problems through Tanh mapping combined with deterministic offset design. This ensures a uniform phase distribution on the complex plane unit circle, enabling fine-grained control of the air-to-ground reflection link and greatly improving the security of covert communication.

[0130] It is understandable that, in the process of solving for the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix in step 203, the electronic device also includes offline training and online inference of the CA-GNN model, as detailed below: To achieve unsupervised learning, the electronic device constructs a composite loss function that includes an objective function and a constraint penalty term. During training, this loss function ensures that the covert communication performance is maximized while satisfying preset constraints. This loss function can be expressed as: .in, This represents the objective function to be optimized. This represents the constraint weights that are dynamically adjusted over the training period. The perceptual signal-to-noise ratio (SNR) constraint penalty term is expressed as: ; The minimum sum rate constraint penalty term is expressed as: ; The hidden constraint penalty term is expressed as follows: ; This represents the smoothness regularization term; This represents the diversity regularization term. The loss function cleverly takes maximizing the effective concealment rate as the primary objective, transforming the constraints of each physical layer into squared penalty terms. By dynamically adjusting the constraint weights with the training cycle, it focuses on finding feasible solutions in the early stages of the model and on refining performance improvements in the later stages of training.

[0131] During the offline training phase, the electronic device employs mini-batch stochastic gradient descent for end-to-end optimization. In each iteration, a batch of channel instances are randomly sampled from the channel distribution, and a graph input is constructed based on the physical topology. The graph input undergoes iterative message passing and feature aggregation through a multi-layer graph attention mechanism to generate node embedding vectors. Subsequently, the perceptual decoder maps the node embedding vectors to communication beamforming vectors, perceptual beamforming vectors, and RIS phase shift matrices. Through an automatic differentiation mechanism, the gradient of the loss function is backpropagated to the network parameters, and finally updated through the Adaptive Moment Estimation (Adam) optimizer.

[0132] After convergence during the offline training phase, the parameters of the CA-GNN are fixed for online inference. The electronic device fixes the parameters of the CA-GNN and outputs the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix through one forward propagation for the instantaneous channel acquired in real time.

[0133] Understandably, through offline training, CA-GNN can learn a mapping relationship from the system channel state to the near-optimal resource allocation strategy. At the same time, through the online stage, it can efficiently approximate the solution of the original non-convex joint optimization problem with only a single forward propagation, achieving joint optimization with low complexity and high reliability.

[0134] Step 204: Substitute the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix into the optimization objective function to obtain the optimal effective concealment rate.

[0135] In this embodiment, the electronic device substitutes the optimal communication beamforming vector, optimal sensing beamforming vector, and optimal RIS phase shift matrix decoded in step 203 into the expression of the optimization objective function constructed in step 202. Under the premise of satisfying preset constraints such as total power constraint, phase constraint, concealment constraint, minimum sum rate constraint, and sensing signal-to-noise ratio (SNR) constraint, the optimal effective concealment rate under the current communication environment is calculated.

[0136] It should be noted that by feeding back the joint configuration parameters output by deep learning to the optimization target, a real-time quantitative evaluation of the covert communication performance under highly dynamic air-to-ground topology is achieved. While ensuring strict compliance of physical layer parameters, the global optimal solution of the non-convex joint optimization problem is approximated with extremely low computational complexity.

[0137] Step 205: Determine the security of the communication process when the optimal effective concealment rate is greater than the preset rate threshold.

[0138] Among them, the preset rate threshold refers to the preset minimum safe communication rate benchmark, which is used to measure the lower limit of the effective information transmission rate that can be achieved between the HAP node and the ground gateway node under the premise of meeting preset constraints.

[0139] In this embodiment, the electronic device compares the calculated optimal effective concealment rate with a preset rate threshold. If the optimal effective concealment rate is greater than the preset rate threshold, it is determined that the current system configuration can maintain an imperceptible state under the monitoring field of the AAV node with sufficient transmission efficiency, i.e., the communication process is determined to be safe and effective. If the optimal effective concealment rate is less than or equal to the preset rate threshold, it is determined that under the current channel environment or interference conditions, the system cannot meet the preset transmission efficiency requirements while ensuring concealment, i.e., the communication process has a security risk or transmission failure.

[0140] Understandably, this preset rate threshold can be dynamically configured based on the service quality requirements of the terrestrial gateway, the channel fading status of the real-time link, and the potential threat level of the AAV.

[0141] It should be noted that this step, by introducing a preset rate threshold as a criterion for judging secure communication, effectively achieves a dynamic balance between communication imperceptibility and service transmission reliability, providing a quantitative basis for real-time risk assessment and decision-making of covert communication systems in complex adversarial environments.

[0142] In the embodiments of this application, the technical solutions described in steps 201-205 above abstract complex nodes such as HAP, RIS, and AAV into an undirected weighted graph through graph theory modeling, accurately depicting the channel gain and interference correlation in the air-ground integrated scenario; by deeply coupling the communication beamforming vector and sensing beamforming vector of the HAP node, and coordinating the phase shift matrix of the RIS, a refined reconstruction of the spatial signal propagation path is achieved; while effectively enhancing the transmission quality of legitimate links, this method uses sensing capabilities to monitor and suppress the illegal interception probability of AAV nodes in real time, significantly improving the effective concealment rate of the system, and fundamentally solving the concealment and security problems of communication in high-altitude dynamic environments from the physical layer, exhibiting extremely strong environmental adaptability and anti-interference capabilities.

[0143] To better understand the embodiments of this application, simulation experiments are conducted on the reconfigurable smart surface RIS-assisted secure communication method provided in the embodiments of this application, and the simulation results are described in detail below: The specific simulation parameters for this experiment are as follows: the HAP is deployed at a fixed high-altitude location [0.0, 0.0, 15000.0], and the ground gateway and RIS are located at [490, 516, 0] and [280, 200, 30] respectively. To simulate mobile eavesdropping threats, the initial location of the AAV is set to [102.58, 25, 291.51]. Furthermore, the network includes... There are 10 ground users with spatially distributed coordinates of [200, 180, 0], [230, 120, 0], [300, 260, 0], and [360. 0, 210, 0]. The system carrier frequency is set to... The antenna gains of the HAP and the ground gateway are respectively and .

[0144] Figure 3 This is a schematic diagram showing the comparison curves of the effective concealment rate as a function of the number of iterations provided in the embodiments of this application. Figure 3 As shown, simulation results indicate that CA-GNN converges quickly and reaches the highest effective concealment rate in only about 100 iterations, improving performance by approximately 21.5% compared to the GA scheme. The baseline scheme uses simulated annealing for fine-grained search, but its performance is the worst due to its failure to fully utilize channel structure information. The GA scheme utilizes population diversity to improve the rate in the early stages of iteration, but as population homogenization intensifies, the search gets trapped in local optima and premature convergence occurs. The Alternating Iteration (AO) scheme, due to strong coupling of multiple variables, adopts a block coordinate descent strategy, resulting in a long period of saddle point stagnation during the optimization process, only escaping the local solution after 1600 iterations. The unsupervised deep neural network (DNN) scheme, because it treats the channel state as an unstructured vector, cannot capture complex variable features and struggles to find the globally optimal strategy under strict physical constraints.

[0145] Figure 4 This is a schematic diagram illustrating the convergence trajectory of the effective concealment rate provided in this application embodiment under the influence of different learning rates. For example... Figure 4 As shown, when the learning rate is small (LR=0.001), the limited step size of network weight updates leads to relatively slow convergence, requiring more training epochs to approach the optimal solution. While a larger learning rate (LR=0.007) accelerates initial feature extraction and gradient descent, it causes local oscillations in the later stages, resulting in fluctuations in steady-state performance. In contrast, the LR=0.004 curve maintains good stability while ensuring fast convergence, achieving the best balance between convergence efficiency and accuracy. The vertical dashed line represents the steady-state inflection point; the optimal learning rate configuration only requires approximately 800 iterations to reach performance saturation, significantly reducing training time compared to the 1800 iterations required for convergence with LR=0.001.

[0146] The following describes the reconfigurable smart surface RIS-assisted secure communication system provided in the embodiments of this application. The reconfigurable smart surface RIS-assisted secure communication system described below can be referred to in correspondence with the reconfigurable smart surface RIS-assisted secure communication method described above.

[0147] Figure 5 This is a schematic diagram of the structure of a reconfigurable smart surface RIS-assisted secure communication system provided in an embodiment of this application. For example... Figure 5 As shown, the system includes: a model building module 501, a parameter optimization module 502, a rate determination module 503, and a safety determination module 504.

[0148] Model building module 501 is used to construct an undirected weighted graph containing a High Altitude Platform (HAP) node, a Reconfigurable Smart Surface (RIS) node, a ground gateway node, a ground user node, and an unauthorized aerial vehicle (AAV) node; it is used to determine the communication beamforming vector during communication between the HAP node and the ground gateway node. Sensing beamforming vector and RIS phase shift matrix An optimization objective function is constructed to determine the effective covert rate of the communication process.

[0149] The parameter optimization module 502 is used to determine the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix that satisfy preset constraints based on the target feature vectors of each node in the undirected weighted graph.

[0150] The rate determination module 503 is used to substitute the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix into the optimization objective function to obtain the optimal effective concealment rate.

[0151] The security determination module 504 is used to determine the security of the communication process when the optimal effective concealment rate is greater than a preset rate threshold.

[0152] Optionally, the model building module 501 is specifically used to build the model based on the linear receiving vector. The communication beamforming vector The RIS phase shift matrix The channel vector between the RIS node and the ground gateway node The communication channel between the HAP node and the ground gateway node The channel vector between the RIS node and the ground gateway node. Construct the communication signal power term of the ground gateway node; based on the linear reception vector The sensing beamforming vector and the communication channel matrix between the HAP node and the ground gateway node. Construct the sensing interference power term for the ground gateway node; based on the RIS phase shift matrix... The transmit power of the ground user node, and the communication link between the RIS node and the AAV node. Given the channel vector between the RIS node and the ground user node, construct the uplink interference power term of the AAV node to the ground gateway node; based on the communication signal power term, the sensed interference power term, and the uplink interference power term, construct the optimization objective function.

[0153] Optionally, the parameter optimization module 502 is specifically used to extract features from the initial feature vectors of each node in the undirected weighted graph to obtain the first node embedding vector of each node; and simultaneously input the first node embedding vectors of each node into the graph attention network GAT, which includes Layered graph attention layer It is an integer greater than 1; in this In the first layer of the graph attention layer, for all nodes, the first... The node is determined to be related to the first node. The node adjacent to the first The node; according to the first node; The first node embedding vector of the nth node and the nth node's embedding vector The first node embedding vector of the nth node determines the nth node in the attention layer of the first-level graph. The node for the first... The attention weights of the nodes are determined, and the attention weights of the nodes in the first-level graph are updated. The first node embedding vector of the nth node is obtained to obtain the nth node. The second node embedding vector of each node; in this In the graph attention layers other than the first graph attention layer, the attention layer outputs the attention layer of the next higher graph attention layer. The node embedding vector and attention weights of the nth node, and the nth node's... The node embedding vector of the nth node is used to update the nth node in the attention layer of the graph. The node embedding vector of the nth node is obtained. The embedding vector of the third node of each node; The third node embedding vector of each node output by the last layer of the graph attention layer is used as the target feature vector of each node. Based on the target feature vector of each node, the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix that satisfy the preset constraints are determined.

[0154] Optionally, the parameter optimization module 502 is specifically used to employ the... The first attention head The first attention head, according to the first Attention vectors corresponding to each attention head First preset mapping matrix Second preset mapping matrix , and the first The first node embedding vector of each node and the first The first node embedding vector of each node Determine the first attention layer in the first-level graph. The node for the first... Attention coefficient of each node According to the first The node for the first... Attention coefficient of each node , and the first The attention coefficients of the nth node to its other neighboring nodes determine the nth node's attention coefficients. The node for the first... Attention weights of each node The remaining adjacent nodes include the first Each node; based on the mapping matrix corresponding to the values ​​of each attention head. The attention of each head is below the first The node for the first... Attention weights of each node, and the embedding vector of the first node. and the first node embedding vector Update the attention layer of the first layer graph. The second node embedding vector of each node .

[0155] Optionally, the preset constraints include at least: the total power constraint of the HAP node, and the constraints on the RIS node. The constraints include the phase constraints of the reflective elements, the concealment constraints of the communication between the HAP node and the ground gateway node, the minimum sum rate constraints of the ground user node, and the perceived signal-to-noise ratio (SNR) constraints of the AAV node.

[0156] Optionally, the parameter optimization module 502 is specifically used to construct a representation of the first mode. The first probability distribution of the signal received by the AAV node and the characterization of the second mode. The second probability distribution of the signal received by the AAV node; based on the channel vector between the HAP node and the AAV node. The sensing beamforming vector The RIS phase shift matrix The transmit power of the ground user node, and the communication link between the RIS node and the AAV node. The channel vector and communication beamforming vector between the RIS node and the ground user node. Determine from the first probability distribution To the second probability distribution KL divergence ; this KL divergence If the condition is less than or equal to the preset condition, it shall be regarded as the hidden constraint.

[0157] Optionally, the parameter optimization module 502 is specifically used to optimize the path gain based on the radar ranging equation. Linear receive vector The channel matrix between the HAP node and the AAV node and the sensing beamforming vector Determine the sensing signal-to-noise ratio of the HAP node. The perceived signal-to-noise ratio Less than or equal to the perceived signal-to-noise ratio threshold This serves as the signal-to-noise ratio (SNR) constraint for the perception.

[0158] Optionally, the parameter optimization module 502 is specifically used to optimize the parameters based on the transmit power of the ground user node and the receive vector of the ground gateway node. The channel vector between the RIS node and the ground gateway node The RIS phase shift matrix The uplink rate of the ground user node is determined by the channel vector between the RIS node and the ground user node. ; this uplink rate Greater than or equal to the minimum sum rate threshold This serves as the minimum sum rate constraint.

[0159] Optionally, the rate determination module 503 is specifically used to construct a node embedding matrix based on the target feature vector of each node. Embed the node into the matrix The input is fed into the beamforming vector decoder, which then embeds the node's matrix. Feature extraction is performed on the target feature vectors of the HAP node and the ground gateway node to obtain the optimal communication beamforming vector and the optimal sensing beamforming vector that satisfy the preset constraints; the node is then embedded into the matrix. The input is fed into the RIS phase-shift decoder, which then embeds the node's matrix. The target feature vector of the RIS node is used to extract features to obtain the optimal RIS phase shift matrix that satisfies the preset constraint.

[0160] Figure 6 This is a schematic diagram of the structure of the electronic device provided in an embodiment of this application. For example... Figure 6 As shown, the electronic device may include a processor 610, a communications interface 620, a memory 630, and a communication bus 640. The processor 610, communications interface 620, and memory 630 communicate with each other via the communication bus 640. The processor 610 can invoke logical instructions from the memory 630 to execute a reconfigurable smart surface RIS-assisted secure communication method.

[0161] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0162] On the other hand, embodiments of this application also provide a computer program product, which includes a computer program that can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer is able to execute the reconfigurable smart surface RIS-assisted secure communication method provided by the above methods.

[0163] In another aspect, embodiments of this application also provide a non-transitory computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the reconfigurable smart surface RIS-assisted secure communication method provided by the methods described above.

[0164] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0165] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0166] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A reconfigurable smart surface RIS-assisted secure communication method, characterized in that, include: Construct an undirected weighted graph containing high-altitude platform (HAP) nodes, reconfigurable intelligent surface (RIS) nodes, ground gateway nodes, ground user nodes, and illegal aerial vehicle (AAV) nodes; During the communication process between the HAP node and the ground gateway node, according to the communication beamforming vector Sensing beamforming vector and RIS phase shift matrix An optimization objective function is constructed, which is used to determine the effective covert rate of the communication process; Based on the target feature vectors of each node in the undirected weighted graph, determine the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix that satisfy the preset constraints. Substituting the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix into the optimization objective function yields the optimal effective concealment rate. The communication process is deemed secure if the optimal effective concealment rate is greater than a preset rate threshold.

2. The reconfigurable smart surface RIS-assisted secure communication method according to claim 1, characterized in that, The method based on the communication beamforming vector Sensing beamforming vector and RIS phase shift matrix Construct the optimization objective function, including: According to the linear reception vector The communication beamforming vector The RIS phase shift matrix The channel vector between the RIS node and the ground gateway node The communication channel between the HAP node and the ground gateway node The channel vector between the RIS node and the ground gateway node. Construct the communication signal power term of the ground gateway node; According to the linear receiving vector The sensing beamforming vector and the communication channel matrix between the HAP node and the ground gateway node. Construct the sensing interference power term of the ground gateway node; According to the RIS phase shift matrix The transmit power of the ground user node, and the communication link between the RIS node and the AAV node. And the channel vector between the RIS node and the ground user node, to construct the uplink interference power term of the AAV node to the ground gateway node; The optimization objective function is constructed based on the communication signal power term, the sensed interference power term, and the uplink interference power term.

3. The reconfigurable smart surface RIS-assisted secure communication method according to claim 1, characterized in that, The step of determining the optimal communication beamforming vector, optimal sensing beamforming vector, and optimal RIS phase shift matrix that satisfy preset constraints based on the target feature vectors of each node in the undirected weighted graph includes: Feature extraction is performed on the initial feature vectors of each node in the undirected weighted graph to obtain the first node embedding vector of each node; and the first node embedding vectors of each node are simultaneously input into the graph attention network GAT, wherein GAT includes Layered graph attention layer It is an integer greater than 1; In the In the first layer of the graph attention layer, for all nodes, the first... The node is determined to be related to the first node. The node adjacent to the first The node; according to the first node; The first node embedding vector of the nth node and the nth node The first node embedding vector of the nth node determines the nth node in the first layer graph attention layer. The node for the first The attention weights of the nodes are determined, and the attention weights of the nodes in the first-level graph attention layer are updated. The first node embedding vector of the nth node is obtained to obtain the nth node. The second node embedding vector of each node; In the In the graph attention layers, excluding the first graph attention layer, the output of the first graph attention layer is based on the output of the adjacent previous graph attention layer. The node embedding vector and attention weight of the nth node, and the nth node's... The node embedding vector of the nth node is updated in each graph attention layer. The node embedding vector of the nth node is used to obtain the nth node. The third node embedding vector of each node; The The third node embedding vector of each node output by the last graph attention layer in the layered graph attention layer is used as the target feature vector of each node. Based on the target feature vectors of each node, determine the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix that satisfy the preset constraints.

4. The reconfigurable smart surface RIS-assisted secure communication method according to any one of claims 1-3, characterized in that, The preset constraints include at least: the total power constraint of the HAP node, and the power constraint on the RIS node. The constraints include the phase constraints of each reflective element, the concealment constraints of communication between the HAP node and the ground gateway node, the minimum sum rate constraints of the ground user node, and the perceived signal-to-noise ratio (SNR) constraints of the AAV node.

5. The reconfigurable smart surface RIS-assisted secure communication method according to claim 3, characterized in that, The first-layer graph attention layer includes One's attention, If it is an integer greater than 1, then according to the first... The first node embedding vector of the nth node and the nth node The first node embedding vector of the nth node determines the nth node in the first layer graph attention layer. The node for the first The attention weights of the nodes are determined, and the attention weights of the nodes in the first-level graph attention layer are updated. The second node embedding vector of each node includes: Using the above The first attention head The attention head, according to the first attention head, Attention vectors corresponding to each attention head First preset mapping matrix Second preset mapping matrix , and the first The first node embedding vector of each node and the first The first node embedding vector of each node Determine the first attention layer in the first graph. The node for the first Attention coefficient of each node According to the aforementioned first The node for the first Attention coefficient of each node , and the first The attention coefficients of the nth node to its other neighboring nodes are used to determine the nth node. The node for the first Attention weights of each node The remaining adjacent nodes include the first One node; Based on the mapping matrix corresponding to each attention point The attention heads mentioned above are the first ones. The node for the first Attention weights of each node, and the embedding vector of the first node and the first node embedding vector Update the first attention layer of the graph. The second node embedding vector of each node .

6. The reconfigurable smart surface RIS-assisted secure communication method according to claim 4, characterized in that, The method further includes: Constructing the first representation pattern The first probability distribution of the AAV node received signal and the characterization of the second mode are described below. The second probability distribution of the AAV node received signal is described below; Based on the channel vector between the HAP node and the AAV node The sensing beamforming vector The RIS phase shift matrix The transmit power of the ground user node, and the communication link between the RIS node and the AAV node. The channel vector and communication beamforming vector between the RIS node and the ground user node. Determine from the first probability distribution To the second probability distribution KL divergence ; The KL divergence The condition being less than or equal to a preset condition serves as the concealment constraint.

7. The reconfigurable smart surface RIS-assisted secure communication method according to claim 4, characterized in that, The method further includes: Based on the path gain of the radar ranging equation Linear receive vector The channel matrix between the HAP node and the AAV node and the sensing beamforming vector Determine the sensing signal-to-noise ratio of the HAP node. ; The perceived signal-to-noise ratio Less than or equal to the perceived signal-to-noise ratio threshold This serves as the perceived signal-to-noise ratio (SNR) constraint.

8. The reconfigurable smart surface RIS-assisted secure communication method according to claim 4, characterized in that, The method further includes: Based on the transmit power of the ground user node and the receive vector of the ground gateway node The channel vector between the RIS node and the ground gateway node The RIS phase shift matrix The uplink rate of the ground user node is determined by the channel vector between the RIS node and the ground user node. ; The uplink rate Greater than or equal to the minimum sum rate threshold , as the minimum sum rate constraint.

9. The reconfigurable smart surface RIS-assisted secure communication method according to claim 1, 3, or 5, characterized in that, The step of determining the optimal communication beamforming vector, optimal sensing beamforming vector, and optimal RIS phase shift matrix that satisfy preset constraints based on the target feature vectors of each node includes: Based on the target feature vectors of each node, construct a node embedding matrix. ; Embed the nodes into the matrix The input is fed into the beamforming vector decoder, which then processes the node embedding matrix. The target feature vectors of the HAP node and the ground gateway node are used to extract features to obtain the optimal communication beamforming vector and the optimal sensing beamforming vector that satisfy the preset constraints. Embed the nodes into the matrix The input is fed into the RIS phase-shift decoder, which then processes the node embedding matrix. The target feature vector of the RIS node is used to extract features to obtain the optimal RIS phase shift matrix that satisfies the preset constraints.

10. A reconfigurable smart surface RIS-assisted secure communication system, characterized in that, include: The model building module is used to construct an undirected weighted graph containing High Altitude Platform (HAP) nodes, Reconfigurable Smart Surface (RIS) nodes, ground gateway nodes, ground user nodes, and unauthorized aerial vehicle (AAV) nodes; and is used to perform beamforming based on communication beamforming vectors during communication between the HAP nodes and the ground gateway nodes. Sensing beamforming vector and RIS phase shift matrix An optimization objective function is constructed, which is used to determine the effective covert rate of the communication process; The parameter optimization module is used to determine the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix that satisfy preset constraints based on the target feature vectors of each node in the undirected weighted graph. The rate determination module is used to substitute the optimal communication beamforming vector, the optimal sensing beamforming vector, and the optimal RIS phase shift matrix into the optimization objective function to obtain the optimal effective concealment rate. The security determination module is used to determine the security of the communication process when the optimal effective concealment rate is greater than a preset rate threshold.

Images