Terahertz smart surface slice anti-quantum secure prediction orchestration method

Abstract

The application discloses a terahertz intelligent reflecting surface slice anti-quantum security prediction arrangement method, three-dimensional voxelized state data is formed by collecting and processing training sequences, echoes and network operation monitoring data; topological consistency features are obtained by three-dimensional blocking masking, encoding backfilling, local window correlation updating and introducing topological consistency constraints; a terahertz radio map is generated by solving a radio map parameter set; an obstacle distribution map is output by fusing multi-directional obstacle candidate results; a slice simulation copy is established, future state sequences are predicted, and service default risk metrics are calculated; a security control channel is established, expected signals are generated, and consistency verification is performed; when a fraud trigger disposal strategy is determined, joint arrangement instructions are generated according to the service default risk metrics and the fraud detection results and are issued. The application realizes a predictable, verifiable and anti-quantum security arrangement closed loop for a terahertz RIS network slice, and improves slice SLA guarantee capability and system security.

Description

Technical Field

[0001] This invention relates to the field of communication network orchestration and security control technology, specifically a terahertz intelligent reflective surface slice anti-quantum security prediction orchestration method. Background Technology

[0002] As terahertz communication is increasingly used in ultra-high bandwidth and ultra-low latency scenarios, terahertz links are highly sensitive to obstruction, scattering, pointing deviation, and environmental changes. Meanwhile, network slicing, as a key mechanism for providing differentiated services across multiple services, requires refined and dynamic orchestration in resource allocation, routing / scheduling, and air interface parameter configuration to ensure the Service Level Agreements (SLAs) and Quality of Service (QoS) of different slices.

[0003] In existing technologies, network slicing orchestration typically relies on historical statistics or short-term measurements, making it difficult to accurately predict occlusion and propagation changes in terahertz high-frequency environments. This leads to orchestration delays, resource waste, or increased risk of SLA defaults. Furthermore, as a programmable electromagnetic environment control component, the reflection coding state of the RIS (Reflection Coding System) is a critical parameter in the control plane. If subjected to deception, tampering, or signal injection attacks, it could cause a sharp drop in link performance, interruption of slice services, or even security incidents. Traditional cryptographic systems also face potential risks under the threat of quantum computing, and relying solely on encryption authentication is insufficient to verify the consistency of the "reflection coding-echo response" at the physical layer.

[0004] Therefore, a method is needed to construct radio maps and obstacle distributions that reflect changes in the propagation environment in terahertz and RIS scenarios; predict future states of network slices and assess SLA default risks; perform verifiable consistency checks on RIS reflection coding states to identify spoofing under quantum-safe control channel protection; link risk assessment with spoofing detection results, output joint orchestration instructions and execute them; thereby achieving a secure, predictable, and verifiable closed-loop network slice orchestration. Summary of the Invention

[0005] This invention proposes a quantum-safe prediction and orchestration method for terahertz smart reflective surface slices to solve the above-mentioned technical problems.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: A quantum-safe prediction and orchestration method for terahertz smart reflective surface slices is applied to a communication system including an orchestration control device, terahertz access nodes, reconfigurable smart reflective surfaces, and multiple network slices. The method includes: The training sequence signals and echo signals of the terahertz access node (AP) and the reconfigurable smart reflective surface (RIS), as well as network operation monitoring data, are collected and processed to generate three-dimensional voxelized state data to characterize the propagation environment and network status. The three-dimensional voxelized state data is subjected to three-dimensional block segmentation, masking, encoding, backfilling, and local window correlation update; structural continuity constraints and spatial adjacency consistency constraints are introduced to ensure that the output features meet the requirements of maintaining three-dimensional topological relationships, thus obtaining topologically consistent features; Using topological consistency characteristics as constraints, training sequence signals, echo signals, and network operation monitoring data are fused together. A pre-established conditional mapping relationship is invoked to solve for the terahertz radio map parameter set and generate a terahertz radio map. Based on the obstacle candidate results from multiple observation directions, a fusion rule of counting voting and weighted voting is adopted to form the obstacle occupancy, and the obstacle distribution map is output in binarized form. For each network slice, a slice simulation copy is established based on the radio map parameter set, obstacle distribution map and network operation monitoring data of the network slice, and the future state sequence is obtained by recursion according to the state transition equation within the future time window; The network slice's performance index sequence within a future time window is calculated based on the future state sequence, and compared with a preset service threshold to obtain a service default risk measure. The service default risk measure is a risk value obtained based on the proportion of time the performance index sequence exceeds the service threshold, the extent of the exceedance, or a weighted combination thereof. A quantum-resistant secure control channel is established, which includes: generating a session key using a quantum-resistant key encapsulation algorithm, and performing identity authentication and integrity verification on the orchestration control message using a quantum-resistant digital signature algorithm; Under the protection of the quantum-safe control channel, the orchestration control device sends the reflection coding state associated with the network slice to the RIS; based on the radio map parameter set and obstacle distribution map, the expected echo signal or expected received signal in the reflection coding state is calculated, and a likelihood ratio test statistic is constructed to perform a deception test on the consistency between the reflection coding state and the received signal; when the likelihood ratio test statistic is greater than the detection threshold, it is determined that there is deception and a security handling strategy is triggered. The detection threshold is determined according to the preset target detection probability constraint and false alarm probability constraint. Based on the service default risk measurement and deception detection results, a set of candidate joint orchestration actions is generated; the future state sequence and performance index sequence under each candidate joint orchestration action are deduced using slice simulation replicas, the predicted risk value corresponding to each candidate joint orchestration action is calculated, and the target joint orchestration action is determined by combining action cost and security constraints. The joint orchestration instructions for each network slice are output and executed via a quantum-secure control channel.

[0007] Furthermore, the specific implementation process for generating three-dimensional voxelized state data to characterize the propagation environment and network state includes: The training sequence signal and echo signal are collected by the terahertz access node and the reconfigurable smart reflective surface. Network operation monitoring data is collected from the slice management plane / data plane. The network operation monitoring data includes at least slice-level resource occupancy parameters, link quality parameters and service quality indicators. Time synchronization and spatial registration are performed on training sequence signals, echo signals and network operation monitoring data. The AP clock or control device clock is used as the reference, and timestamp alignment or interpolation alignment is used to make multi-source data correspond to the same time slot / same sampling period. The measurement point, observation azimuth, RIS attitude or array direction are mapped to a unified spatial coordinate system. The space is divided into a three-dimensional voxel grid. Each voxel records at least the spatial occupancy status, reflection / scattering intensity information, and link state markers associated with network slices, forming three-dimensional voxelized state data for characterizing the propagation environment and network state. .

[0008] Furthermore, the specific implementation process for obtaining the topology consistency feature includes: Three-dimensional voxelized state data Divided into several three-dimensional blocks The block is randomly selected as the masking block according to the masking ratio, and the masking process is performed; the unmasked block is input into the encoder to obtain local features; Each unmasked block The local features are backfilled to the corresponding positions on the complete 3D feature mesh; the masked region is represented by adjacent backfilled features; the 3D feature mesh is a mapping from voxel positions to feature vectors; To ensure that features are smooth and consistent in the local space and to preserve structural boundaries, correlation updates are performed within the local window. Within the local window, a correlation coefficient matrix is ​​calculated between the backfilled features and the local features of the unmasked blocks, and the backfilled features are then updated with a weighted value based on the correlation coefficient matrix. To ensure the preservation of three-dimensional topological relationships, structural continuity constraints and spatial adjacency consistency constraints are introduced. Based on the constraint criteria, output features that meet the requirements for preserving three-dimensional topological relationships are obtained, forming topological consistency features. The constraint formula for the structural continuity constraint is as follows: in, This is a penalty term for the gradient difference between adjacent voxels. For the set of adjacent voxel pairs, For the set of boundary voxel pairs, , For the voxel index in the 3D voxel mesh, ( () represents a pair of adjacent voxels. , Representing voxels , The corresponding feature vector, Represents the Euclidean norm; The boundary relaxation coefficient; The constraint formula for the spatial adjacency consistency constraint is as follows: in, This is a penalty term applied to the feature distance between adjacent blocks. For the set of adjacent partition pairs, For block With block Spatially adjacent and having boundary contact or pre-defined overlapping zones , Divided into blocks and partitioning The feature convergence vector on its boundary / overlapping band voxels.

[0009] Furthermore, the specific implementation process for generating the terahertz radio map includes: Using topology consistency features as constraints, link observations formed by training sequence signals and echo signals, as well as network operation monitoring data, are fused together. A pre-established conditional mapping relationship is invoked. The conditional mapping relationship is a regression model that is trained offline and can be incrementally updated online. The inputs are topology consistency features, link observations, and network operation monitoring data, and the output is a set of radio map parameters. Channel estimation / statistical extraction is performed on the training sequence signal and the echo signal to form an observation set that corresponds one-to-one with the spatial voxel position. The observations include at least measurable or estimable quantities such as received power / reference signal received power, signal-to-noise ratio or equivalent channel gain, and angle of arrival / delay spread. At the same time, link quality indicators and service-side statistical indicators in the network operation monitoring data are read. The observations are aligned with the voxel grid according to the timestamp, AP / RIS attitude and spatial coordinate index carried on them. Represent the radio map as a set of parameter fields on a three-dimensional voxel grid; By pre-establishing conditional mapping relationships, topological consistency features and geometric / configuration conditions are mapped to the prior parameter field or initial values ​​of the radio map; Based on a pre-defined forward propagation model, the parameter field is mapped to observable predicted values, and the deviation between the measured observations and the predicted observations is taken as the observation consistency error. At the same time, a priori regularity constraint obtained from topological consistency features is introduced to construct a fusion solution objective and solve the set of radio map parameters. The radio map parameter set is mapped to radio map raster values ​​on a 3D voxel grid. Voxels not directly observed are filled using spatial interpolation / extrapolation constrained by topological priors. Finally, the radio map parameter set and radio map raster are output, yielding a terahertz radio map. Furthermore, the process of obtaining the future state sequence includes: For each network slice, a slice simulation copy is established based on the radio map parameter set, obstacle distribution map, and network operation monitoring data of the network slice. The future state sequence is then recursively derived within a future time window according to the state transition equation. The slice simulation copy includes at least the slice state space and time interval. State vector, slice control action space, time The system includes the orchestration / control actions, the state transition equations used to deduce the state at the next moment from the current state and control actions, the observation functions used to map the simulation state to observable quantities, the set of model parameters related to the slice, and the obstacle distribution map used to constrain the line-of-sight / occlusion and propagation path effectiveness; the future time window is a sliding time window, and the sliding step size is determined by the service delay threshold or the channel coherence time.

[0010] Furthermore, the deception check for consistency between the reflection-coded state and the received signal includes: A quantum-safe control channel is established between the orchestration control device and the terahertz access node and / or the reconfigurable smart reflective surface. The quantum-safe control channel includes generating a session key using a quantum-resistant key encapsulation algorithm and performing identity authentication and integrity verification on the orchestration control messages using a quantum-resistant digital signature algorithm. Under quantum-safe control channel protection, the reflection coding state associated with the network slice is obtained through RIS, denoted as: in, This represents the number of RIS units. For the first Each RIS unit at time... Phase configuration; The imaginary unit; For RIS at the moment Reflection / phase control state; For the Each subcarrier transmits known training symbols. And receive the echo : in, To receive baseband observations; The training symbols are known. To receive noise; The equivalent channel matrix after RIS participation; Obstacle distribution map Incorporate the desired signal calculation for each candidate propagation path Define its set of voxel paths And define the occlusion attenuation coefficient: in, For three-dimensional voxel positions; The result of the obstacle occupancy determination; This is the occlusion attenuation coefficient; For path Effective attenuation under the influence of obstacles; The equivalent channel is corrected by the attenuation coefficient due to occlusion, thus improving the equivalent channel. Revised to To obtain the expected equivalent channel including obstacle constraints. The desired received signal is obtained: The residual vector is obtained based on the received baseband observations and the desired received signal. : Sample estimation of the sliding window length L based on the residual vector. : in, It is the conjugate transpose; It is the identity matrix; Here is the regularization constant; In order to be in The residual vector of the k-th subcarrier at time k; The consistency test statistic is calculated based on the sample estimation. : For continuous The consistency test statistics at each time point are summed to obtain the likelihood ratio test statistic. The cumulative formula is: in, This is the cumulative length; when If the value exceeds the detection threshold, deception is detected and a security handling strategy is triggered.

[0011] Furthermore, the specific implementation process of outputting the joint orchestration instructions for each network slice includes: The detection threshold is adaptively adjusted based on the service default risk metric, while satisfying the target detection probability constraint and the false alarm probability constraint; a baseline threshold independent of the slice is determined based on the false alarm constraint. To implement adaptive risk adjustment, the adjustment formula is: in, For slices At any moment Adaptive detection threshold; The baseline threshold is obtained from the false alarm constraint calibration; This is the adjustment coefficient; For use in measuring service default risk Mapped to A monotonically non-decreasing mapping function; when If the risk exceeds the preset risk threshold and the deception detection result is no deception, a performance-guaranteed orchestration will be executed. When the deception detection result indicates deception, the safety-first approach should be taken first. The safety-priority approach aims to reduce risk and control costs, and constructs a joint orchestration optimization function with the following optimization formula: in, In order to make joint decisions Lower slice Service default risk measurement; For the cost of the action; This is a coefficient representing the trade-off between risk and cost. When the deception detection result is deception, the RIS configuration search space is limited and the attack surface is reduced from the set of trusted RIS configurations. That is, the joint decision action that satisfies the optimization function of joint orchestration is selected from the set of candidate joint orchestration actions as the target joint orchestration action; a security handling strategy is triggered according to the target joint orchestration action instruction; the security handling strategy includes at least one of the following: freezing the current reflection configuration, falling back to the trusted reflection codebook, triggering a secondary challenge training sequence, or increasing the signature authentication frequency; The joint orchestration instruction includes a joint optimization solution for slice resource allocation and reflection configuration while meeting the objective of reducing service default risk measurement.

[0012] A communication system for quantum-safe predictive orchestration of terahertz smart reflective surface slices, the communication system comprising an orchestration control device, a terahertz access node (AP), a reconfigurable smart reflective surface (RIS), and a network slice set; The orchestration and control unit is used to collect network operation monitoring data, construct radio maps and obstacle maps, create slice simulation copies, calculate risks, and generate joint orchestration instructions; it is also responsible for establishing a quantum-safe control channel and performing identity authentication and integrity verification on control messages. Terahertz access nodes are used to send training sequences, receive echo / receive signals, collect link quality information, and report it. A reconfigurable intelligent reflective surface is used to adjust the reflection / scattering response according to the reflection code status issued by the orchestration control device, and to cooperate with echo acquisition or status reporting. A collection of network slices, each corresponding to a specific service SLA, with network operation monitoring data including slice-level resource usage parameters, link quality parameters, and service quality indicators.

[0013] Compared with existing technologies, the beneficial effects achieved by this invention are as follows: The terahertz smart reflective surface slice quantum-safe prediction and orchestration method provided by this invention is applied to a communication system including an orchestration control device, a terahertz access node, a reconfigurable smart reflective surface, and multiple network slices. The method includes: collecting training sequences, echoes, and slice operation monitoring data; performing time synchronization and spatial registration to form three-dimensional voxelized state data; obtaining topological consistency features through three-dimensional block masking, encoding backfilling, local window correlation updates, and the introduction of topological consistency constraints; solving for the radio map parameter set based on the features and generating a terahertz radio map, while simultaneously fusing multi-directional obstacle candidate results to output an obstacle distribution map; establishing a simulation copy for each slice and predicting future state sequences, calculating a service default risk metric; establishing a secure control channel using quantum-resistant key encapsulation and quantum-resistant digital signatures, issuing reflection-encoded states under this protection, generating desired signals using the radio map and obstacle distribution, constructing a likelihood ratio test statistic for consistency verification, determining deception if a threshold is exceeded, and triggering a handling strategy; finally, generating joint orchestration instructions based on the risk metric and deception detection results and issuing them for execution. By extracting voxelized state and topological consistency features, robust characterization of changes in occlusion / scattering environment is enhanced, improving the accuracy of terahertz radio map construction. Through multi-directional obstacle fusion and slice simulation copy prediction, SLA default risk can be quantified in advance within future time windows, enabling forward-looking orchestration. Control messages are protected by a quantum-safe control channel, and a likelihood ratio test based on echo and reflection coding consistency is further introduced to achieve combined protection of cryptographic security and physical layer verifiable consistency. Joint orchestration instructions are generated through the linkage of risk measurement and deception detection, enabling rapid and controllable resource and configuration adjustments in security incidents or high-risk scenarios, improving slice reliability and system security. This achieves a predictable, verifiable, and quantum-safe orchestration closed loop for terahertz RIS network slices, enhancing slice SLA assurance capabilities and system security. Attached Figure Description

[0014] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0015] Figure 1 This is a schematic diagram of the system structure in an embodiment of the present invention; Figure 2 This is a flowchart illustrating the overall process of the method in this embodiment of the invention; Figure 3 This is a schematic diagram of three-dimensional voxelization and topological consistency feature extraction in an embodiment of the present invention; Figure 4 This is a schematic diagram of radio map solving and obstacle fusion in an embodiment of the present invention; Figure 5 This is a schematic diagram of the slice simulation copy and risk measurement in an embodiment of the present invention; Figure 6 This is a schematic diagram of the quantum security control and likelihood ratio consistency test in an embodiment of the present invention; Figure 7 This is a schematic diagram of the joint orchestration output in an embodiment of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Please see Figure 1-7 In this first embodiment: a terahertz smart reflective surface slice quantum-safe prediction and orchestration method is applied to a communication system including an orchestration control device, a terahertz access node, a reconfigurable smart reflective surface, and multiple network slices. The communication system includes: The orchestration control unit is used to collect monitoring data, construct radio maps and obstacle maps, create slice simulation copies, calculate risks and generate joint orchestration instructions; it is also responsible for establishing a quantum-safe control channel and performing identity authentication and integrity verification on control messages.

[0018] Terahertz access node (AP): Used to send training sequences, receive echo / receive signals, collect and report link quality information; and perform some air interface configuration and slice bearer functions.

[0019] Reconfigurable Intelligent Reflective Surface (RIS): Adjusts the reflection / scattering response according to the reflection coding status issued by the orchestration control device, and cooperates with echo acquisition or status reporting.

[0020] Multiple network slices: Each slice, such as S1, S2, and S3, corresponds to a specific service SLA and has monitoring data such as slice-level resource usage parameters, link quality parameters, and service quality indicators. It exchanges service data with user equipment / service terminals.

[0021] Monitoring data may include, but is not limited to: CPU / memory / bandwidth usage, spectrum / timeslot usage, queue length, packet loss rate, end-to-end latency, jitter, throughput, reliability indicators, etc. Link quality parameters may include signal-to-noise ratio, received power, bit error rate, channel state information or its statistics, etc.

[0022] The method includes: Step 1: Collect and process the training sequence signals and echo signals of the terahertz access node AP and the reconfigurable smart reflector RIS, as well as the network operation monitoring data, to generate three-dimensional voxelized state data to characterize the propagation environment and network status. In one embodiment, the AP sends a training sequence signal (pilot) at the beginning of each scheduling cycle or each sliding time window. RIS according to the current reflection coding state During operation, the AP collects or receives echo signals. Simultaneously, network operation monitoring data is collected from the slice management plane / data plane. This data includes timestamps and device location or orientation information, providing a foundation for subsequent time synchronization and spatial registration.

[0023] In one embodiment, the following is performed on the training sequence signal, echo signal, and monitoring data: 1) Time synchronization: Based on the AP clock or control device clock, timestamp alignment or interpolation alignment is used to make multi-source data correspond to the same time slot / same sampling period; 2) Spatial registration: Mapping measurement points, observation azimuth, RIS attitude or array orientation, etc., to a unified spatial coordinate system; 3) Voxelization: The space is divided into a three-dimensional voxel grid. Each voxel records at least: spatial occupancy status, reflection / scattering intensity information, and link status markers associated with network slices. This forms three-dimensional voxelized state data.

[0024] The voxel size can be determined based on the scene scale and computing power. For example, 0.1m to 0.5m can be used indoors, and 0.5m to 2m can be used outdoors.

[0025] Step 2: Perform 3D block segmentation, masking, encoding, backfilling, and local window correlation update on the 3D voxelized state data; introduce structural continuity constraints and spatial adjacency consistency constraints to ensure that the output features meet the requirements of maintaining 3D topological relationships, and obtain topologically consistent features; In one embodiment, voxelized state data Divided into several three-dimensional blocks ,in, For a state tensor on a 3D voxel mesh, it must contain at least channels for space occupancy, scattering intensity, and link markers. To divide into preset blocks from Sub-blocks obtained by cropping. Based on the masking ratio. (For example, 10% to 60%) Randomly select a portion of blocks as masking blocks; input the unmasked blocks into the encoder. Obtain local features The local features of each unmasked block. Backfill to complete 3D feature mesh The corresponding position; the masked area is represented by the adjacent backfill feature.

[0026] To ensure smooth and consistent features within the local space and preserve structural boundaries, correlation updates are performed within the local window. Considered from voxel position The mapping to the eigenvector, denoted as voxel position The feature vector at position (dimension 1) The updated features are For example, a weighted aggregation form can be used: in: For the position index in the 3D voxel mesh; For The set of neighborhood voxels centered on the center (e.g. window, (A spherical neighborhood can also be used). For central voxels With neighboring voxels The normalized correlation weights, the above formula is for... Intra-similarity is normalized using softmax to satisfy... ; For the similarity function, cosine similarity can be used. Or correlation coefficient; This is for exponential operations.

[0027] To ensure the preservation of 3D topological relationships, the following two types of constraints are introduced (implemented in the form of regularization terms): 1) Structural continuity constraint: Suppresses non-physical abrupt changes in the characteristics of adjacent voxels within the same continuous region, while allowing variations at obstacle boundaries. Let It is a set of adjacent voxel pairs (e.g., 6 / 18 / 26 adjacent). This is the set of boundary voxel pairs; boundary voxel pairs can be determined by changes in obstacle distribution / occupancy labels between adjacent voxels (e.g., adjacent voxels with different occupancy labels are included). ).but: in, For boundary relaxation coefficients (e.g.) This is used to implement strong constraints in continuous regions and weak constraints in boundary regions. This is a penalty term for the gradient difference between adjacent voxels. For the set of adjacent voxel pairs, This is a set of boundary voxel pairs used to characterize voxel pairs that allow feature transitions at obstacle boundaries or segmentation boundaries. , For the voxel index in the 3D voxel mesh, ( () represents a pair of adjacent voxels. , Representing voxels , The corresponding feature vector, This represents the Euclidean norm.

[0028] Boundary voxel pairs can be represented by obstacle distribution / occupancy markers. The changes are determined, for example, when adjacent voxels occupy different labels, they are identified as boundary pairs: symbol( The ) represents the set difference (difference operation). Therefore: Indicates voxel pairs ( (belongs to the set of adjacent voxel pairs) However, it does not belong to the set of boundary voxel pairs. That is, for voxels located within the same continuous region (not at the boundary), a stronger continuity constraint should be imposed; Indicates voxel pairs ( Located in the boundary region, where certain characteristic mutations are allowed, the coefficients are used to determine the specific characteristics. Weaken the constraints.

[0029] Voxel representation The corresponding feature vector (e.g., 3D encoded / backfilled voxel features). Similarly. Symbols This represents the 2-norm (Euclidean norm). It represents the square, used to measure the sum of squares of the differences in features between adjacent voxels; the greater the difference, the greater the penalty.

[0030] 2) Spatial Adjacency Consistency Constraint: This constraint ensures that adjacent blocks have consistent features in overlapping or splicing boundary regions, avoiding block artifacts. Let... For the set of adjacent partition pairs, express and Spatially adjacent and having boundary contact or pre-defined overlapping zones, For block Feature convergence vectors on its boundary / overlapping band voxels (e.g., for this region) (Obtained by average pooling / weighted average), then: in, This is a penalty term applied to the feature distance between adjacent blocks. For the set of adjacent partition pairs, For block With block Spatially adjacent and having boundary contact or pre-defined overlapping zones , Divided into blocks and partitioning Feature convergence vectors on its boundary / overlapping band voxels; Constraint criterion: The above constraints are incorporated into the objective function in the form of regularization terms. For example: in, These are the weight coefficients. They are updated / trained iteratively to... Convergence and reaching a preset threshold are considered to satisfy topological consistency, for example, satisfying: in The features can be selected empirically through offline calibration or validation sets to ensure smooth features within continuous regions and consistent features at the boundaries of adjacent blocks, while allowing variations at obstacle boundaries. The resulting output features satisfy the requirement of preserving 3D topological relationships, forming topologically consistent features. .

[0031] Step 3: Using topological consistency features as constraints, integrate training sequence signals, echo signals, and network operation monitoring data, call the pre-established conditional mapping relationship, solve for the terahertz radio map parameter set, and generate the terahertz radio map. In one embodiment, topology consistency features As a constraint, the pre-established condition mapping relationship is invoked. Solving for the terahertz radio map parameter set yields parameters such as spatial path loss distribution. Or received power distribution The conditional mapping relationship can be an offline trained regression model, a graphical model, or a physical-data fusion model, and can support online incremental updates to adapt to environmental changes.

[0032] Radio maps can be represented as pairs of points on a spatial grid. or Visualization or numerical representation of the data is used for subsequent prediction and control.

[0033] This is performed based on the results of time synchronization and spatial registration. Inputs include: topology consistency features ( The calculation process involves the following mechanisms: link observations formed by training sequence signals and echo signals, and network operation monitoring data; the output is a set of radio map parameters and a rasterized radio map result. (3-1) Construction and unified representation of observations Channel estimation / statistical extraction is performed on the training sequence signal and the echo signal to form an observation set that corresponds one-to-one with the spatial voxel position. The observations include at least measurable or estimable quantities such as received power / reference signal received power, signal-to-noise ratio or equivalent channel gain, angle of arrival / delay spread, etc. Simultaneously, link quality indicators and service-side statistical indicators (such as packet loss rate, latency, throughput, etc.) from network operation monitoring data are read as auxiliary information for observation reliability and service constraint strength. All the above observations carry timestamps, AP / RIS attitude, and spatial coordinate indices for alignment with the voxel grid.

[0034] (3-2) Radio map parameterization Represent the radio map as a set of parameter fields on a three-dimensional voxel grid. For example, the following can be taken: , indicating the voxel position ( ) path loss distribution; or , indicating the voxel position ( The received power distribution; It can also be expanded to include parameter fields such as the dielectric absorption attenuation coefficient and additional losses from reflection / scattering.

[0035] (3-3) Implementation of conditional mapping relationship Pre-establish conditional mapping relationships ( This is used to map topological consistency features and geometric / configuration conditions to a priori parameter field or initial values ​​for the radio map. For example: ( This can be an offline-trained regression model (such as 3D convolutional regression, graphical regression, or kernel regression), with the input being... Including conditional quantities such as location / frequency / RIS coding status, the output is the initial value of the radio map parameters. ; or,( The approach employs a hybrid method combining physical models and data fitting: using terahertz propagation mechanisms (free-space diffusion, molecular absorption, reflection / scattering) as the framework, historical observations are used to calibrate the model coefficients, thereby obtaining an initial field that can be used for rapid extrapolation. .

[0036] The prior initial value does not directly replace the actual measurement, but provides interpretable constraints and convergence direction for subsequent fusion solution.

[0037] (3-4) Multi-source data fusion and parameter solving Based on forward propagation model parameter field The data is mapped to observable predicted values, and the deviation between measured and predicted observations is taken as the observation consistency error. Simultaneously, prior regularization constraints derived from topological consistency features are introduced to construct the fusion solution objective. Observation consistency error term: The errors of each observation frame / position / frequency point are weighted and summarized. The weights are determined by the observation signal-to-noise ratio, echo quality, time freshness, and service-side stability (the higher the quality, the greater the weight), thereby achieving unified fusion of training sequences, echoes and monitoring data; Topological prior regularization: The parameter field obtained from constraint solution. with prior initial value Maintain consistency in spatial structure to avoid non-physical jumps in continuous areas, while allowing parameter abrupt changes at occlusion boundaries; Smoothing and boundary preservation terms: Apply smoothing regularization to the voxel neighborhood and reduce the smoothing intensity for suspected obstacle boundary regions to balance stability and boundary clarity.

[0038] In engineering implementation, the above objectives can be solved using weighted least squares, maximum a posteriori (MAP) estimation, or recursive filtering: within each sliding time window... Iterative update of initial value When the convergence condition is met (e.g., the objective function decreases below a threshold or the number of iterations reaches the upper limit), the set of radio map parameters is output. The solution process naturally supports online incremental updates: only the affected voxel regions are updated after a new observation arrives, thus ensuring real-time performance.

[0039] (3-5) Radio map generation and output formats The obtained set of radio map parameters Mapped to radio map raster values ​​(e.g., path loss map or received power map) on a 3D voxel grid, voxels not directly observed can be completed using spatial interpolation / extrapolation constrained by topological priors. The final output is: 1) Radio map parameter set ; 2) Radio map grid.

[0040] Step 4: Based on the obstacle candidate results from multiple observation locations, the obstacle occupancy is formed by a fusion rule of counting voting and weighted voting, and the obstacle distribution map is binarized and output. In one embodiment, obstacle candidate results obtained from multiple observation locations (e.g., different AP directions, different RIS codes, different times / different measurement frames) are fused. Let the observation index be... },in, Let be the number of observations / azimuths involved in the fusion; For position indexing in a 3D voxel mesh (e.g.) ), for each position Formed from the first Candidate determination of each observation .

[0041] For the first Each observation pair of voxels The obstacle candidate results.

[0042] When the output is a binary decision , where 1 indicates that the voxel is an obstacle, and 0 indicates that it is not an obstacle or is empty.

[0043] When the output is confidence level / probability At this point, candidate thresholds can be used first. Convert to poll: .

[0044] The counting vote yielded the first fusion result: in This is an indicator function; it takes the value 1 if the condition is true, and 0 otherwise. When the confidence level is given, the above equation is equivalent to: .

[0045] The weighted vote yielded the second fusion result: in For the first One observation in voxels The reliability weight at each point is used to reflect the impact of different observation conditions on the reliability of the results; And can be normalized to make Thus ensuring It is numerically scale-stable. For example, It can be determined by at least one or more measurable / statistical factors: The signal-to-noise ratio / measurement quality of this observation (The higher the signal-to-noise ratio, the greater the weight); Obstruction level / Multipath interference intensity (The stronger the occlusion, the smaller the weight). Historical statistical reliability (For example, the lower the false alarm rate (FAR) for that location, the better.) The higher (the higher).

[0046] For example: in It is a non-negative adjustment coefficient.

[0047] The two types of fusion results are combined to obtain the occupation score: in These are combination coefficients used to weigh majority vote consistency. ) and quality-weighted credibility ( ).

[0048] Binarization standard: Will occupy the score With occupancy threshold After binarization, the obstacle distribution map is output. : in, The binarization threshold can be determined in any of the following ways: 1. Fixed threshold method: When the threshold is set to a fixed threshold, the threshold is set to a fixed threshold. Return to One Then take the commonly used threshold (e.g.) ); 2. Minimum Unanimous Votes Method: Stipulating at least... Only when an observation is determined to be an obstacle will an output be generated. At this time, it can be made (Corresponding counting field) or (Corresponding to the normalized field); 3. Constrained False Alarm Method: Select from the validation set / historical data This aims to ensure that the false alarm rate does not exceed the preset upper limit, while maximizing the detection rate.

[0049] Step 5: For each network slice, establish a slice simulation copy based on the radio map parameter set, obstacle distribution map and network operation monitoring data of the network slice, and recursively obtain the future state sequence according to the state transition equation within the future time window; In one embodiment, for each service slice Create a corresponding slice simulation copy (digital twin). It is used to simulate the evolution of slice link state and resource state over time under radio map and obstacle constraints, and to predict future state sequences within future time windows.

[0050] (5-1) For example, slice simulation copy It can be represented as the following tuple: in: Slice the state space; For a moment The state vector; To control the motion space for slicing; For a moment The arrangement / control of actions; This is the state transition function (dynamic equation), used to deduce the state at the next moment from the current state and control actions; The observation function is used to map the simulation state to observables (e.g., measured values). (throughput, latency, etc.) This is a set of model parameters related to the slice (e.g., noise spectral density, bandwidth, modulation and coding scheme, RIS reflection efficiency, service arrival statistics, etc.). This is an obstacle distribution map / occupancy map, used to constrain line-of-sight / occlusion and propagation path effectiveness.

[0051] Given a slice simulation copy ( From the structured representation of ) and the tuple definitions of its input and output, we can see that ( The computability of ) depends on a progressive refinement at three levels: state space ( ) by state vector ( The component system spanned by ) has specific components and their range of values ​​in ( Define a state vector within the state space. The components and physical meanings of ) are given in the text; secondly, the state transition function ( The update relationships of sub-states such as link quality, queues, and resources need to be expressed in a computable dynamic form. The relevant transition equations are given in the state transition equations, which make the state components have verifiable update rules under the action, environment, and disturbance effects, and are constrained by the radio map parameter set and obstacle distribution map. Third, after determining the initial state and the action sequence within the future window, the single-step transition of the state transition equation can be recursively expanded within the future time window. Thus, given the initial state and action sequence, the single-step transition of the state transition equation can be recursively expanded within the future time window to obtain the future state sequence. This sequence is used as the input source for the performance index sequence and the default risk measurement to obtain the future state sequence and output it.

[0052] Slice Simulation Copy ( The network operation monitoring data is not constructed in isolation. Its environmental constraints include at least a set of radio map parameters (which determines the values ​​of spatial path loss, received power, and equivalent channel propagation) and an obstacle distribution map (which determines line-of-sight conditions, occlusion attenuation, and the set of available reflection paths). Its state initialization and parameter calibration are based on the collected and spatiotemporally aligned slice monitoring data. Thus, the radio map parameter set, the obstacle distribution map, and the network slice monitoring data jointly constrain the effectiveness of the state transition function, thereby ensuring that subsequent recursive predictions have consistent data dependencies and causal chains.

[0053] (5-2) For example, slicing The state vector can be written as: in: For slices At any moment Signal-to-noise ratio; For received power; This is a resource occupancy vector or scalar, containing at least one resource dimension's occupancy rate, such as compute / bandwidth / spectrum / timeslot occupancy, often taking the form of... ; This is used to cache queue status (e.g., queue length / number of bytes in queue / queue quantity corresponding to queuing latency). ; For a Quality of Service (QoS) metric vector, such as throughput End-to-end delay Packet loss rate etc.; slices The state vector also includes interference power Available bandwidth Scheduling duty cycle, RIS configuration index, etc.; (5-3) The transition of the slice state at discrete time can be written as: in: Update function for slice dynamics; For a moment The set of environmental and system parameters includes at least noise and bandwidth parameters, transmit power limit, AP pointing / beam parameters, RIS coding / phase parameters, and interference statistics parameters; For the obstacle distribution map, determine occlusion / accessibility (e.g., whether the line-of-sight (LoS) holds true, and which reflection paths are valid). Slicing orchestration actions include at least one or more of the following: resource allocation (bandwidth / time slot / computing quota), routing / service chain selection, AP beam pointing, RIS phase / codebook selection, etc. This is a disturbance term used to characterize uncertainties such as measurement noise, rapid fading, and randomness of service arrival. For example, it can be a zero-mean random vector (such as Gaussian noise or bounded noise).

[0054] To ensure the model is feasible and verifiable, an unrestricted implementation example can be provided: The link quality sub-model, constrained by the obstacle distribution map, is as follows: in: This refers to the transmission power. Antenna / beam gain; For the equivalent channel, the distribution of obstacles The impact of the occlusion indication decision; Interference power; Noise spectral density; For bandwidth.

[0055] The queue and service sub-models are as follows: in: The state of the cache queue at the next time step t+1; For business arrival volume; For service volume / outbound volume, and It can be determined by the achievable rate, for example: in, This is a proportionality coefficient determined by protocol overhead / scheduling efficiency.

[0056] Resource occupancy sub-model, composed of actions The driver, specifically: in, This is a truncation function to ensure that the occupancy rate is within a certain range. Inside.

[0057] (5-3) gives a single-step transfer operator, which can transfer ( ) deduced to ( (5-4) Given an initial state and action sequence, the single-step transition of (5-3) is recursively expanded within the future time window to obtain a future state sequence. This sequence serves as the input source for the performance index sequence and default risk measurement. The operator is then iterated within the future time window according to the candidate action sequence to form a future state sequence output that can be directly used for risk calculation, thereby completing the closed-loop connection between model, prediction, and risk.

[0058] (5-4) Let the current time be . The future predicted step size is (Positive integer), the future time window is: The prediction process includes at least the following steps: The initial state is obtained from the current measurement and estimation. (For example, by (Obtained by inversion / filtering of observations).

[0059] Read the obstacle distribution map from the output. And determine the environmental parameters inside the window in the future. ( If the environmental parameters remain approximately constant within a short time window, then we can take... .

[0060] Determine future actions inside the window Actions can come from a predefined strategy, the output of a higher-level orchestrator, or a set of candidate actions (e.g., generated by resource budget or slice SLA constraints).

[0061] right Iterative execution: Obtain the predicted sequence .

[0062] Output the future state sequence It is used for subsequent slice resource orchestration, risk warning, or target optimization. Step 6: Calculate the performance index sequence of the network slice within the future time window based on the future state sequence, and compare it with the preset service threshold to obtain the service default risk measure. The service default risk measure is a risk value obtained based on the proportion of time when the performance index sequence exceeds the service threshold, the extent of the exceedance, or a weighted combination thereof. In one embodiment, based on the slice future state sequence Calculate the performance index sequence and with preset service thresholds Compare and obtain slices Service default risk measurement. The future time window is... },in To predict the number of steps, This is the window length.

[0063] (6-1) Performance index sequence The service quality metric for each slice at future time points can be a single metric or a vector of multiple metrics. For example, when taking the single metric form, it can be derived from the predicted state as follows: The mapping yields: in, This is a metric mapping function used to extract or calculate service metrics from the state vector. The following is an example implementation without limitations: Using latency as an indicator (the lower the better): When the predicted state includes information such as queue length / arrival rate / service rate, the predicted queuing latency can be obtained. And order: Using throughput as a metric (the higher the better): when the predicted state includes bandwidth With signal-to-noise ratio At that time, the rate can be estimated: Using packet loss rate as an indicator (the lower the better): When the predicted state includes queue overflow probability or packet loss statistics, let: When using multiple indicators Can be a vector: A single comprehensive indicator can be obtained through normalization and weighting (for subsequent unified comparison): in For the first Normalization results of each indicator Weighted and detrimental .

[0064] (6-2) Preset service threshold The minimum / maximum acceptable level specified in the Service Level Agreement (SLA) or business strategy is used to determine whether there is a risk of default. Considering the differences between "the higher the better" and "the lower the better" for different metrics, the comparison criteria are divided into two categories: Upper limit indicators (the lower the better): such as latency, packet loss rate, jitter, etc. The criteria for determining default are... If the indicator exceeds the allowed limit, it is considered a breach of contract.

[0065] Lower limit metrics (the higher the better): such as throughput, availability, reliability, etc. Default criteria are... If the indicator falls below the minimum guaranteed value, it is considered a breach of contract.

[0066] For consistent expression, a default indication function can be defined: The criteria for determining default shall be selected according to the above-mentioned upper limit or lower limit rules.

[0067] (6-3) Future Time Window The risk of default can be measured using one or more of the following methods.

[0068] 1) Risk of default ratio (The percentage of defaults occurring within the window), specifically: in: The larger the value, the more frequently defaults will occur within the future window; The length of the window; Indexing future moments; 2) Risk of Default Scope (i.e., the average degree to which defaults exceed the threshold): To be compatible with upper limit / lower limit indicators, a default over-limit amount is defined. : Upper limit type (the smaller the better): Lower limit type (the larger the better): The default margin risk is defined as follows: in The larger the value, the more serious the breach of contract.

[0069] 3) Weighted portfolio risk in And can be taken To ensure dimensional stability; Used to emphasize frequency risk, Used to emphasize magnitude risk; For slice identification; Indexing future moments; For future forecasting; The length of the window; To predict the number of steps; For the future predicted state of the slice; For the reason from The calculated performance indicators; Set a preset service threshold (SLA / policy given, which can be an upper or lower limit); For default indication function; Risk of default percentage; Risk of default magnitude; For comprehensive risks; For combined weights.

[0070] Step 7: Establish a quantum-resistant secure control channel, which includes: generating a session key using a quantum-resistant key encapsulation algorithm, and performing identity authentication and integrity verification on the orchestration control messages using a quantum-resistant digital signature algorithm; In one embodiment, a quantum-secure control channel is established between the orchestration control device (e.g., a slice orchestrator / controller) and the terahertz access node (e.g., an AP) and / or the smart reflective surface (RIS) controller. This channel carries subsequent slice orchestration instructions, RIS configuration distribution, status feedback, and other control plane messages. The security of the control channel includes at least: quantum-secure key negotiation, authentication and integrity of control messages, confidentiality protection, and replay protection.

[0071] (7-1) Participating entities shall include at least: Control terminal entity: orchestration control device ; Controlled entity: Terahertz access node and / or RIS controller .

[0072] In one embodiment, the entity pre-configures or registers the following long-term key materials: Quantum-resistant digital signature key pairs: per entity Has a signing private key With verification public key And during the system initialization or registration phase Distribute / register to peers or trusted directories (for identity authentication and integrity verification).

[0073] Quantum-resistant key binding (KEM) key pairs: controlled entities (e.g., or ) has KEM private key With KEM public key The control terminal can obtain the corresponding KEM public key (used for session key negotiation) through the registration information.

[0074] (7-2) In one embodiment, the control terminal and the controlled terminal negotiate the time through a quantum key encapsulation algorithm. Session key The process can be performed as follows (via control terminal). With the controlled end For example, Similarly): Obtain the KEM public key: Control terminal Obtain the controlled end KEM public key .

[0075] Encapsulation generates key material: The control terminal performs KEM encapsulation operations: in, Encapsulate the key with ciphertext (encapsulation result). This refers to the shared key material obtained through encapsulation.

[0076] Key derivation: Control end pair Obtain the session key by performing key derivation. : in, This is a key-derived function; context information may include slice identifier, entity identifier, timestamp, etc., used to isolate key spaces for different sessions / different slices.

[0077] Sending encapsulated ciphertext: The control end will Send to the controlled end .

[0078] Decapsulation and recovery key material: The controlled end performs KEM decapsulation operation: Consistent session key derivation: The controlled end also performs the following: When encapsulation and decapsulation are performed correctly, both ends will yield consistent results. .

[0079] Session key Confidentiality protection for control plane messages (e.g., symmetric encryption) and / or message authentication code generation (e.g., based on...) (integrity verification), and can be rotated according to a preset cycle (e.g., every seconds or per (Updated once per message).

[0080] (7-3) To prevent forgery and tampering, control messages use quantum-resistant digital signatures for authentication and integrity verification. Exemplary steps are as follows: Message framing: The control terminal generates control messages to be sent. It includes at least: message type, target entity identifier, slice identifier, and parameter payload.

[0081] Add an anti-replay field: Add a sequence number and timestamp to the message to form: in, For incrementing sequence numbers, For timestamps.

[0082] Computational summary: For Computational Summary .

[0083] Signature: The control terminal generates a signature using its signature private key. Send: Send To the controlled end.

[0084] Signature verification: The controlled end uses the control end to verify the public key. If the signature verification fails, the message is discarded and a security alert is recorded.

[0085] Two-way authentication: The controlled end can also sign the receipt / status report message to achieve two-way identity authentication.

[0086] (7-4) To defend against replay attacks, the controlled terminal maintains a replay window. And judged according to the following rules: Timestamp window: If If the message is deemed expired or abnormal, it will be rejected. This is the upper limit for allowed clock skew.

[0087] Sequence Number Window: The controlled end records the highest sequence number that has been received. ,like If the issue has already occurred, it will be considered a replay and the package will be rejected; among which... This is the window length.

[0088] Handling strategy: For messages judged to be replayable / expired, discard them and trigger an alarm / count; if necessary, trigger session key rotation or re-handshake.

[0089] (7-5) In one embodiment, the quantum-resistant key encapsulation algorithm can be ML-KEM; the quantum-resistant digital signature algorithm can be Falcon. This invention can also be replaced with other key encapsulation and digital signature algorithms that meet quantum-resistant security requirements without affecting the technical structure and effectiveness of this step.

[0090] Step 8: Under the protection of the quantum-safe control channel, the orchestration control device sends the reflection coding state associated with the network slice to the RIS; based on the radio map parameter set and obstacle distribution map, the expected echo signal or expected received signal in the reflection coding state is calculated, and a likelihood ratio test statistic is constructed to perform a deception test on the consistency between the reflection coding state and the received signal; when the likelihood ratio test statistic is greater than the detection threshold, it is determined that there is deception and a security handling strategy is triggered. The detection threshold is determined according to the preset target detection probability constraint and false alarm probability constraint. This step is used to observe and verify the consistency of the reflection coding state and echo response from the physical layer, in addition to the control channel security authentication, in order to detect inconsistencies caused by RIS state tampering, channel model abnormalities, or environmental changes, thereby triggering security measures and policy convergence.

[0091] (8-1) In one embodiment, given the RIS reflection coding state (phase / amplitude configuration) Its equivalent reflection matrix is ​​denoted as: in: This represents the number of RIS units. For the first Each unit at time... Phase configuration; The imaginary unit; For RIS at the moment The reflection / phase control state.

[0092] For the The subcarrier (or the first) (each pilot resource unit) sends known training symbols (Complex scalar or vector) and receive the echo. It can be written as: in: For receiving baseband observations (which can be scalars or vectors, with dimensions consistent with the number of receiving antennas); Given pilot / training symbols; The received noise (which can be approximated as zero-mean complex Gaussian noise); The equivalent channel matrix after RIS participation can be expanded as follows: in, For direct link channels, For the channel from the transmitter to the RIS, This is the channel from the RIS to the receiver.

[0093] Furthermore, in order to map the obstacle distribution... Incorporate the desired signal calculation for each candidate propagation path Define its set of voxel paths And define the occlusion attenuation coefficient: in: For three-dimensional voxel positions; For obstacle occupancy determination (1 represents an obstacle); This is the occlusion attenuation coefficient (the larger the value, the stronger the occlusion effect). For path Effective attenuation under the influence of obstacles.

[0094] Accordingly, the amplitude of the multipath components (or equivalent channel) is modified, for example, by adjusting the path amplitude. Revised to Thus, the expected equivalent channel including obstacle constraints is obtained. The desired received signal is finally obtained: .

[0095] (8-2) Define the residual (error) vector: Under the assumptions of no tampering / consistency, the residuals mainly consist of noise and unmodeled disturbances, and can be approximated as a zero-mean complex Gaussian distribution: in, Let be the residual covariance matrix. For example, it can be estimated using samples of a sliding window length (L): in: It is the conjugate transpose; For unit array; This is a regularization constant used to ensure that the matrix is ​​invertible; In order to be in The residual vector of the k-th subcarrier at time k.

[0096] Define the consistency test statistic (equivalent to a monotonic transformation of the generalized likelihood ratio): when The larger the value, the more severe the deviation of the observed signal from the expected signal.

[0097] To improve stability, continuous The cumulative (or average) statistics of each time point / symbol: in This is the cumulative length (detection of accumulated length).

[0098] (8-3) Let the threshold be... The judgment rule is as follows: like If the results are inconsistent, an alarm / action will be triggered; otherwise, if the results are consistent, the verification will pass.

[0099] (8-4) Determine the threshold by satisfying both constraints: False alarm probability constraints: In the case of consistency The probability of a false alarm does not exceed the upper limit; Detection probability constraints: In the case of falsification / inconsistency assumptions The probability of detection is not lower than the lower limit.

[0100] in: This represents the probability of a false alarm. The probability of detection; , Preset constraints; ; ; , It can be given by security policy or business requirements.

[0101] threshold Usually press first Constraints are set to prevent excessive false alarms, and then adjustments are made. (Accumulated length) or model accuracy, making The requirements must be met; ultimately, both constraints should be satisfied simultaneously.

[0102] (8-5) The process for determining the threshold provides two possible implementation methods (either one can be chosen, or a combination thereof): Method 1: Citing Quantiles Data was collected during a time period when normal and consistent data was known (no attacks, correct configuration). sample set .

[0103] Given a false alarm limit Take it Quantiles as thresholds: Evaluation on simulated tampered / inconsistent data .like If so, the threshold will be adjusted.

[0104] Method 2: Theoretical Distribution Method exist And when the covariance estimate is sufficient, It approximately follows a chi-square distribution (degrees of freedom are related to the observation dimension). The accumulated values ​​are... It approximates a chi-square distribution with degrees of freedom increasing proportionally. Therefore, the threshold can be determined from the quantiles of the chi-square distribution: in, For equivalent degrees of freedom, and related to the receiving dimension and the accumulated length ( Related; (8-6) When using the threshold obtained by method one / method two Can meet But it cannot be satisfied. In this case, adjustments can be made using the following methods until both constraints are satisfied simultaneously: Adjust accumulation length Increase This will improve the statistical measure's ability to distinguish persistent anomalies, thereby enhancing... , usually for The effects can be mitigated by recalibration. offset.

[0105] Adjusting the number of training symbols / pilot density: Increasing the number of observation samples can reduce covariance estimation error and improve detection stability.

[0106] Adjust the occlusion correction factor With model parameters Reduce residual bias caused by model errors, thereby reducing false alarms and improving detection from the source.

[0107] Dual threshold strategy: Set alarm thresholds With disposal threshold ( First, issue an alarm, then escalate the response to reduce the risk of mishandling.

[0108] fixed →Calibration →Measure →If insufficient, increase. And recalibrate → Until the constraints are met.

[0109] For a specific moment; For subcarrier / pilot index; This represents the number of RIS units. These are the RIS phase configuration and equivalent reflection matrix, respectively. This refers to the actual received signal; To receive the desired signal; For training symbols; For noise; For direct links and RIS-related sub-channels; Obstacle occupancy map; This is the occlusion attenuation coefficient; For path decay; For path voxels; For residuals; For residual covariance; For covariance estimation window; It is a regularization term; , To test the statistic; This is the cumulative length; (8-7) When the detection statistic satisfies If an inconsistency is found between the reflection encoding state and the echo response, a safety handling procedure is triggered. This safety handling includes at least one or more of the following, and can be executed according to risk level: 1) Risk Classification and Triggering Conditions To avoid mishandling due to occasional noise, tiered thresholds or continuous triggering conditions can be set, for example: Level 1 Alert (Minor Risk): If recently... Within a window, satisfy This will trigger a Level 1 alarm (logging only and enhanced encryption).

[0110] Level 2 Response (Severe Risk): If the following conditions are met or continuous If the threshold is exceeded once, a secondary action (freezing configuration, renegotiation, switching strategy, etc.) will be triggered.

[0111] in: For alarm statistics window length, As a threshold for the number of triggers, To upgrade the threshold.

[0112] 2) A set of actions to be taken (including at least one action) A. Freeze RIS configuration and rollback: Freeze the current RIS reflection code And roll back to the most recent trusted configuration that passed the consistency check. (This configuration can be saved by the control terminal with a version number / hash fingerprint).

[0113] B. Challenge—Response Requirements: The control unit sends a challenge sequence (e.g., random pilot sequence number / random RIS subcodebook index) within the security control channel. The controlled unit executes the challenge and sends back echo statistics; the control unit recalculates. and If it still exceeds the threshold, it is judged as strong inconsistency.

[0114] C. Key rotation and control channel hardening: When a secondary action occurs, session key rotation is triggered: KEM encapsulation / decapsulation is re-executed to generate a new session key. And improve the strength of control message authentication (e.g., increase the frequency of signatures or simultaneously superimpose based on...) (Message authentication code).

[0115] D. Slice arrangement downgrade or path switching: For the affected slices Perform conservative orchestration: limit high-risk paths, reduce modulation order, increase redundancy, switch to alternative APs / alternate RISs or direct links until consistency is restored.

[0116] E. Evidence Recording and Audit Playback: Record trigger time and statistics ,correspond Key observations digest (hash), threshold Together with the decision-making results, form an evidence package for auditing and tracing.

[0117] 3) Recovery conditions and exit strategy When continuous Each detection window meets (or below the alarm threshold) When consistency is restored, the process can be gradually exited: Unfreeze and allow minor adjustments to the RIS within the trusted codebook set; Restore the enhanced mode of the security control channel to the normal mode; Update the trusted configuration library (which will pass the verification) (and its fingerprint is written into a trusted set).

[0118] Through the above consistency verification and handling strategies, a closed loop of expected signal generation—statistical verification—threshold decision—safe handling—recovery convergence is achieved, thereby providing physical layer consistency protection in addition to the quantum-safe control channel, and improving the security and robustness of slice orchestration and RIS control. Step 9: Based on the service default risk measurement and deception detection results, generate a set of candidate joint orchestration actions; use slice simulation replicas to deduce the future state sequence and performance index sequence under each candidate joint orchestration action, calculate the predicted risk value corresponding to each candidate joint orchestration action, and determine the target joint orchestration action by combining action cost and security constraints, output the joint orchestration instructions for each network slice, and send them for execution through the quantum-resistant security control channel.

[0119] In one embodiment, the joint orchestration is based on service default risk measurement. Deception / tampering with detection results Driven by the joint efforts, among which For slice identification; For risk measurement; This is the result of the consistency check. For example, ,in This indicates that inconsistency / suspected tampering has been detected. This indicates that consistency has been achieved. The output of the joint orchestration includes at least: slice resource allocation adjustments, routing / scheduling parameter updates, and RIS reflection configuration updates.

[0120] (9-1) To enable high-risk slices to adopt a more sensitive detection strategy while ensuring controllable false alarms, the detection threshold is adaptively adjusted based on the false alarm constraint. Let the consistency test statistic be... The rule of judgment is as follows: If Then it is judged as inconsistent.

[0121] The false alarm probability is defined as: in This represents the assumption that there is no tampering / consistency. The false alarm constraint is: in The upper limit for false alarms can be preset, which can be given by business tolerance or security policy.

[0122] (9-2) In one embodiment, a baseline threshold independent of the slice is first determined based on the false alarm constraint. Then, risk adaptive adjustment is performed. The baseline threshold can be determined in any of the following ways (either one can be chosen): Method A: Quantile calibration of normal samples Collect a statistical sample set from a known, normal, and consistent data segment. Take it Quantiles are used as a baseline threshold: in Represents the quantile function. This is the upper limit for system-level false alarms.

[0123] Method B: Determining the theoretical distribution quantiles when Down When it approximately follows a chi-square distribution, we can take: in For degrees of freedom The chi-square distribution is in quantiles at the location; With observation dimension and cumulative length Related.

[0124] Each of the above methods explicitly guarantees that: when the threshold is taken as... At that time, the system's false alarm probability does not exceed .

[0125] (9-3) Based on satisfying the false alarm constraint, adjust the threshold according to the risk: in: For slices At any moment Adaptive detection threshold The baseline threshold is obtained from the false alarm constraint calibration; This is an adjustment coefficient; a larger value indicates greater sensitivity to risk. It is a monotonically non-decreasing mapping used to represent the risk quantity. Mapped to (For example or , ); It serves as a comprehensive risk measure (the higher the value, the higher the risk of default).

[0126] As can be seen from the above formula, when When it is large, The decrease makes it easier to trigger detection and more sensitive; when When it is small, the threshold is close to Maintain a low false alarm rate.

[0127] To ensure that the false alarm constraint is always met, adjustments can be made each time. Then, use the most recent normal sample pair Perform online estimation; if the limit is exceeded, then... Back to or improve Until satisfied .

[0128] (9-4) Set the risk threshold To preset risk thresholds (set by business SLAs or strategies), when This indicates a relatively high risk. An example of a joint orchestration rule is as follows: High risk but no tampering detected ( ): Perform performance-guaranteed orchestration: Increase slice resource quotas (bandwidth / time slots / computing power), update routing / scheduling parameters to improve link quality, and update RIS reflection configuration. To optimize (SNR) / throughput / latency.

[0129] Tampering or inconsistency detected ): Prioritize security-first handling: adopt the RIS configuration output by the security handling strategy (freeze / rollback to trusted configuration), and adopt conservative routing / scheduling strategies for slices, while enabling redundant resources to ensure critical SLAs; trigger control channel key rotation and challenge-response core when necessary.

[0130] (9-5) In another embodiment, joint orchestration is formulated as an optimization problem with security constraints. Let at time... Joint decision-making actions This includes variables such as resource allocation, routing / scheduling parameters, and RIS configuration. With the goal of risk reduction and cost control, for example: in: For in action Lower slice The predicted risk (which can be quickly assessed using a simulation copy); The cost of the action (e.g., additional bandwidth / computing power cost, switching cost, energy consumption cost, etc.); This is a risk-cost trade-off coefficient.

[0131] Simultaneously satisfying resource constraints and security constraints. For example, when the detection result... At that time, the RIS configuration must fall within the trusted set: in A trusted RIS configuration set can consist of a historically consistent configuration codebook and version number / fingerprint verification; when This allows for optimization within a larger search space.

[0132] (9-6) To limit the RIS configuration search space and reduce the attack surface when inconsistencies / suspected tampering are detected, a set of trusted RIS configurations is constructed. This set is used to constrain RIS configuration variables. The value can be set, and online updates and audit playback are supported.

[0133] (1) Configuration representation and fingerprint Representing RIS configurations as vectors: Where (N) is the number of RIS units, ( Let be the discrete or continuous phase control value of the (n)th unit. For ease of consistency verification and auditing, a configuration fingerprint (hash) is defined: in: For encoding functions (serializing the phase vector into fixed-byte sequences). Configure version number; Identify the RIS device; "(|)" is a hash function; "(|)" is a concatenation operator.

[0134] (2) Trusted entry data structure A trust set can consist of several trust entries, each denoted as: in: Configure for RIS; To configure fingerprint; Version number; For reliable scoring (used for ranking / elimination); This is the time of the last successful verification. For context labels (e.g., slice to which it belongs, AP pointing to, environment area number, obstacle distribution summary, etc.).

[0135] (3) Initial construction method (at least one) Initialization can be performed in one or a combination of the following ways: A. Codebook initialization: Use a subset of the system's pre-configured RIS codebook as the initial trusted set; B. Offline calibration and initialization: Under known normal conditions, perform consistency checks on several candidate configurations and add those that pass to the set; C. Historical Operation Initialization: Take the configurations that have been stable and passed the consistency test for a long time during the historical operation period and add them to the set.

[0136] (4) Adding / updating rules When a certain configuration Its credibility can be added or increased if one of the following conditions is met: Consistency pass condition: Within the most recent (J) detection windows, the pass count threshold is met. in, This is the minimum number of passes allowed.

[0137] Stability condition: The mean / variance of the statistics within the window is below a threshold. Context matching conditions: Configure the applicable context to be similar to the current context, such as the AP pointing to the same location or the obstacle distribution summary being similar, in order to avoid misuse across scenarios.

[0138] When the above conditions are met, add the configuration entry to the set or update its rating: in To update the coefficients; It can be positively correlated with the number of passes / stability; Compatible with risk measurement Or related to the number of abnormal occurrences.

[0139] (5) Elimination rules, the purpose of which is to prevent the set from becoming too large and the configuration from becoming outdated. To control the set size and remove invalid configurations, the following settings can be configured: Capacity limit: If it exceeds the limit, use the minimum. disuse; Expired and discarded: If More than the current time If so, it will be demoted or removed; Abnormal elimination: If the configuration triggers an inconsistency more than once in the most recent (J) windows. Then remove: (6) Usage rules, and The linkage ensures that safety constraints are executable. when In the event of inconsistencies, the RIS configuration must be selected from a trusted set: And prioritize the item with the highest rating and the best contextual match: in For context distance function, This is a weighting factor.

[0140] when At that time, optimization is allowed in a larger space, but if the optimization results pass the consistency test continuously, they can be added to the trust set according to (4) to realize the closed loop of exploration-verification-inclusion.

[0141] (7) Audit and Evidence Recording Each time from Record when selecting / updating / retiring configurations. The summary forms a replayable configuration evolution chain, facilitating security auditing and post-event traceability.

[0142] It needs to be explained that, For slices Default risk measurement; As a risk threshold; The judgment is determined by the detection criteria (0 for pass, 1 for inconsistency / suspected tampering). This is a consistency statistic; This is the cumulative length; An adaptive detection threshold; The baseline threshold; This is the risk adjustment coefficient; For risk mapping function; For joint choreography; Predicting risks for actions; For the cost of the action; For the weighting factor; Configure a set of trusted RIS configurations; Configure variables for RIS.

[0143] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0144] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A terahertz intelligent reflective surface slice quantum-safe prediction and orchestration method, characterized in that, The method includes: The training sequence signals and echo signals of the terahertz access node (AP) and the reconfigurable smart reflective surface (RIS), as well as network operation monitoring data, are collected and processed to generate three-dimensional voxelized state data to characterize the propagation environment and network status. The three-dimensional voxelized state data is subjected to three-dimensional block segmentation, masking, encoding, backfilling, and local window correlation update; structural continuity constraints and spatial adjacency consistency constraints are introduced to ensure that the output features meet the requirements of maintaining three-dimensional topological relationships, thus obtaining topologically consistent features; Using topological consistency characteristics as constraints, training sequence signals, echo signals, and network operation monitoring data are fused together. A pre-established conditional mapping relationship is invoked to solve for the terahertz radio map parameter set and generate a terahertz radio map. Based on the obstacle candidate results from multiple observation directions, a fusion rule of counting voting and weighted voting is adopted to form the obstacle occupancy, and the obstacle distribution map is output in binarized form. For each network slice, a slice simulation copy is established based on the radio map parameter set, obstacle distribution map and network operation monitoring data of the network slice, and the future state sequence is obtained by recursion according to the state transition equation within the future time window; The network slice's performance index sequence within a future time window is calculated based on the future state sequence, and compared with a preset service threshold to obtain a service default risk measure. The service default risk measure is a risk value obtained based on the proportion of time the performance index sequence exceeds the service threshold, the extent of the exceedance, or a weighted combination thereof. A quantum-resistant secure control channel is established, which includes: generating a session key using a quantum-resistant key encapsulation algorithm, and performing identity authentication and integrity verification on the orchestration control message using a quantum-resistant digital signature algorithm; Under the protection of the quantum-safe control channel, the orchestration control device sends the reflection coding state associated with the network slice to the RIS; based on the radio map parameter set and obstacle distribution map, the expected echo signal or expected received signal in the reflection coding state is calculated, and a likelihood ratio test statistic is constructed to perform a deception test on the consistency between the reflection coding state and the received signal; when the likelihood ratio test statistic is greater than the detection threshold, it is determined that there is deception and a security handling strategy is triggered. The detection threshold is determined according to the preset target detection probability constraint and false alarm probability constraint. Based on the service default risk measurement and deception detection results, a set of candidate joint orchestration actions is generated; the future state sequence and performance index sequence under each candidate joint orchestration action are deduced using slice simulation replicas, the predicted risk value corresponding to each candidate joint orchestration action is calculated, and the target joint orchestration action is determined by combining action cost and security constraints. The joint orchestration instructions for each network slice are output and executed via a quantum-secure control channel.

2. The terahertz smart reflective surface slice quantum-safe prediction and orchestration method according to claim 1, characterized in that, The specific implementation process for generating three-dimensional voxelized state data to characterize the propagation environment and network state includes: The training sequence signal and echo signal are collected by the terahertz access node and the reconfigurable smart reflective surface. Network operation monitoring data is collected from the slice management plane / data plane. The network operation monitoring data includes at least slice-level resource occupancy parameters, link quality parameters and service quality indicators. Time synchronization and spatial registration are performed on training sequence signals, echo signals and network operation monitoring data. The AP clock or control device clock is used as the reference, and timestamp alignment or interpolation alignment is used to make multi-source data correspond to the same time slot / same sampling period. The measurement point, observation azimuth, RIS attitude or array direction are mapped to a unified spatial coordinate system. The space is divided into a three-dimensional voxel grid. Each voxel records at least the spatial occupancy status, reflection / scattering intensity information, and link state markers associated with network slices, forming three-dimensional voxelized state data for characterizing the propagation environment and network state. .

3. The terahertz smart reflective surface slice quantum-safe prediction and orchestration method according to claim 2, characterized in that, The specific implementation process for obtaining the topology consistency feature includes: Three-dimensional voxelized state data Divided into several three-dimensional blocks The block is randomly selected as the masking block according to the masking ratio, and the masking process is performed; the unmasked block is input into the encoder to obtain local features; Each unmasked block The local features are backfilled to the corresponding positions on the complete 3D feature mesh; the masked region is represented by adjacent backfilled features; the 3D feature mesh is a mapping from voxel positions to feature vectors; To ensure that features are smooth and consistent in the local space and to preserve structural boundaries, correlation updates are performed within the local window. Within the local window, a correlation coefficient matrix is ​​calculated between the backfilled features and the local features of the unmasked blocks, and the backfilled features are then updated with a weighted value based on the correlation coefficient matrix. To ensure the preservation of three-dimensional topological relationships, structural continuity constraints and spatial adjacency consistency constraints are introduced. Based on the constraint criteria, output features that meet the requirements for preserving three-dimensional topological relationships are obtained, forming topological consistency features. The constraint formula for the structural continuity constraint is as follows: in, This is a penalty term for the gradient difference between adjacent voxels. For the set of adjacent voxel pairs, For the set of boundary voxel pairs, , For the voxel index in the 3D voxel mesh, ( () represents a pair of adjacent voxels. , Representing voxels , The corresponding feature vector, Represents the Euclidean norm; The boundary relaxation coefficient; The constraint formula for the spatial adjacency consistency constraint is as follows: in, This is a penalty term applied to the feature distance between adjacent blocks. For the set of adjacent partition pairs, For block With block Spatially adjacent and having boundary contact or pre-defined overlapping zones , Divided into blocks and partitioning The feature convergence vector on its boundary / overlapping band voxels.

4. The terahertz smart reflective surface slice quantum-safe prediction and orchestration method according to claim 3, characterized in that, The specific implementation process for generating the terahertz radio map includes: Using topology consistency features as constraints, link observations formed by training sequence signals and echo signals, as well as network operation monitoring data, are fused together. A pre-established conditional mapping relationship is invoked. The conditional mapping relationship is a regression model that is trained offline and can be incrementally updated online. The inputs are topology consistency features, link observations, and network operation monitoring data, and the output is a set of radio map parameters. Channel estimation / statistical extraction is performed on the training sequence signal and the echo signal to form an observation set that corresponds one-to-one with the spatial voxel position. The observations include at least measurable or estimable quantities such as received power / reference signal received power, signal-to-noise ratio or equivalent channel gain, and angle of arrival / delay spread. At the same time, link quality indicators and service-side statistical indicators in the network operation monitoring data are read. The observations are aligned with the voxel grid according to the timestamp, AP / RIS attitude and spatial coordinate index carried on them. Represent the radio map as a set of parameter fields on a three-dimensional voxel grid; By pre-establishing conditional mapping relationships, topological consistency features and geometric / configuration conditions are mapped to the prior parameter field or initial values ​​of the radio map; Based on a pre-defined forward propagation model, the parameter field is mapped to observable predicted values, and the deviation between the measured observations and the predicted observations is taken as the observation consistency error. At the same time, a priori regularity constraint obtained from topological consistency features is introduced to construct a fusion solution objective and solve the set of radio map parameters. The radio map parameter set is mapped to radio map raster values ​​on a three-dimensional voxel grid. For voxels not directly observed, spatial interpolation / extrapolation constrained by topological priors is used to complete the data. Finally, the radio map parameter set and radio map raster are output to obtain the terahertz radio map.

5. The terahertz smart reflective surface slice quantum-safe prediction and orchestration method according to claim 4, characterized in that, The process of obtaining the future state sequence includes: For each network slice, a slice simulation copy is established based on the radio map parameter set, obstacle distribution map, and network operation monitoring data of the network slice. The future state sequence is then recursively derived within a future time window according to the state transition equation. The slice simulation copy includes at least the slice state space and time interval. State vector, slice control action space, time The system includes the orchestration / control actions, the state transition equations used to deduce the state at the next moment from the current state and control actions, the observation functions used to map the simulation state to observable quantities, the set of model parameters related to the slice, and the obstacle distribution map used to constrain the line-of-sight / occlusion and propagation path effectiveness; the future time window is a sliding time window, and the sliding step size is determined by the service delay threshold or the channel coherence time.

6. The terahertz smart reflective surface slice quantum-safe prediction and orchestration method according to claim 5, characterized in that, The deception check for consistency between the reflection-coded state and the received signal includes: A quantum-safe control channel is established between the orchestration control device and the terahertz access node and / or the reconfigurable smart reflective surface. The quantum-safe control channel includes generating a session key using a quantum-resistant key encapsulation algorithm and performing identity authentication and integrity verification on the orchestration control messages using a quantum-resistant digital signature algorithm. Under quantum-safe control channel protection, the reflection coding state associated with the network slice is obtained through RIS, denoted as: in, This represents the number of RIS units. For the first Each RIS unit at time... Phase configuration; The imaginary unit; For RIS at the moment Reflection / phase control state; For the Each subcarrier transmits known training symbols. And receive the echo : in, To receive baseband observations; The training symbols are known. To receive noise; The equivalent channel matrix after RIS participation; Obstacle distribution map Incorporate the desired signal calculation for each candidate propagation path Define its set of voxel paths And define the occlusion attenuation coefficient: in, For three-dimensional voxel positions; The result of the obstacle occupancy determination; This is the occlusion attenuation coefficient; For path Effective attenuation under the influence of obstacles; The equivalent channel is corrected by the attenuation coefficient due to occlusion, thus improving the equivalent channel. Revised to To obtain the expected equivalent channel including obstacle constraints. The desired received signal is obtained: The residual vector is obtained based on the received baseband observations and the desired received signal. : Sample estimation of the sliding window length L based on the residual vector. : in, It is the conjugate transpose; It is the identity matrix; Here is the regularization constant; In order to be in The residual vector of the k-th subcarrier at time k; The consistency test statistic is calculated based on the sample estimation. : For continuous The consistency test statistics at each time point are summed to obtain the likelihood ratio test statistic. The cumulative formula is: in, This is the cumulative length; when If the value exceeds the detection threshold, deception is detected and a security handling strategy is triggered.

7. The terahertz smart reflective surface slice quantum-safe prediction and orchestration method according to claim 6, characterized in that, The specific implementation process of outputting the joint orchestration instructions for each network slice includes: The detection threshold is adaptively adjusted based on the service default risk metric, while satisfying the target detection probability constraint and the false alarm probability constraint; a baseline threshold independent of the slice is determined based on the false alarm constraint. To implement adaptive risk adjustment, the adjustment formula is: in, For slices At any moment Adaptive detection threshold; The baseline threshold is obtained from the false alarm constraint calibration; This is the adjustment coefficient; For use in measuring service default risk Mapped to A monotonically non-decreasing mapping function; when If the risk exceeds the preset risk threshold and the deception detection result is no deception, a performance-guaranteed orchestration will be executed. When the deception detection result indicates deception, the safety-first approach should be taken first. The safety-priority approach aims to reduce risk and control costs, and constructs a joint orchestration optimization function with the following optimization formula: in, In order to make joint decisions Lower slice Service default risk measurement; For the cost of the action; This is a coefficient representing the trade-off between risk and cost. When the deception detection result is deception, the RIS configuration search space is limited and the attack surface is reduced from the set of trusted RIS configurations. That is, the joint decision action that satisfies the optimization function of joint orchestration is selected from the set of candidate joint orchestration actions as the target joint orchestration action; a security handling strategy is triggered according to the target joint orchestration action instruction; the security handling strategy includes at least one of the following: freezing the current reflection configuration, falling back to the trusted reflection codebook, triggering a secondary challenge training sequence, or increasing the signature authentication frequency; The joint orchestration instruction includes a joint optimization solution for slice resource allocation and reflection configuration while meeting the objective of reducing service default risk measurement.

8. A communication system for quantum-safe predictive orchestration of terahertz smart reflective surface slices, using the quantum-safe predictive orchestration method for terahertz smart reflective surface slices as described in claims 1-7, characterized in that... The communication system includes an orchestration control device, a terahertz access node (AP), a reconfigurable smart reflective surface (RIS), and a network slice set. The orchestration and control unit is used to collect network operation monitoring data, construct radio maps and obstacle maps, create slice simulation copies, calculate risks, and generate joint orchestration instructions; it is also responsible for establishing a quantum-safe control channel and performing identity authentication and integrity verification on control messages. Terahertz access nodes are used to send training sequences, receive echo / receive signals, collect link quality information, and report it. A reconfigurable intelligent reflective surface is used to adjust the reflection / scattering response according to the reflection code status issued by the orchestration control device, and to cooperate with echo acquisition or status reporting. A collection of network slices, each corresponding to a specific service SLA, with network operation monitoring data including slice-level resource usage parameters, link quality parameters, and service quality indicators.

Images