A high-reliability low-latency covert communication system and method based on RSMA

Abstract

This invention provides a highly reliable, low-latency covert communication system and method based on RSMA, relating to the fields of wireless communication and information security technology. The method includes: constructing a robust channel model incorporating channel estimation errors; splitting, encoding, and modulating user messages at the base station to obtain a public data stream and a private data stream for legitimate user equipment; calculating the theoretical achievable rates of the public and private data streams; constructing a covert constraint model based on relative entropy; establishing a non-convex optimization problem of minimizing maxima and performing a deterministic transformation with the objective of maximizing the minimum average achievable rate; solving the problem using a continuous convex approximation and two-stage iterative algorithm; and, based on the obtained optimal transmit precoding matrix, weighted merging of the public data stream and the private data stream of legitimate user equipment to generate a baseband digital transmission signal and transmit it; and receiving and decoding the baseband digital transmission signal to recover the user message. This invention provides highly robust covert communication assurance.

Description

Technical Field

[0001] This invention relates to the fields of wireless communication and information security technology, and in particular to a highly reliable, low-latency covert communication system and method based on RSMA. Background Technology

[0002] With the evolution of 6G networks, Ultra-Reliable and Low-Latency Communications (URLLC) has become a core technology supporting critical tasks such as industrial automation, autonomous driving, and telemedicine. To meet extremely low latency constraints, URLLC typically uses finite block length (FBL) for data transmission. However, short packet transmission leads to a non-negligible block error rate (BLER), posing a significant challenge to system throughput and reliability. Furthermore, due to the openness of wireless channels, URLLC networks are highly vulnerable to malicious eavesdropping. Covert communication, as the highest standard of physical layer security, aims to transmit information to legitimate users while ensuring that eavesdroppers (willies) cannot detect the communication activity, which is crucial for highly secure wireless networks.

[0003] To improve spectrum efficiency in massive connectivity scenarios, multiple access (MA) technology is widely used. Traditional non-orthogonal multiple access (NOMA) and spatial division multiple access (SDMA) often face severe interference management bottlenecks in severely overloaded or strongly correlated channel scenarios. To address this issue, rate-splitting multiple access (RSMA) splits user messages into common and private streams and relies on successful interference cancellation (SIC) at the receiver. This allows for flexible soft handover between SDMA and NOMA, thus providing more robust interference management capabilities.

[0004] However, most existing research on RSMA and covert communication is based on the assumption of perfect Channel State Information (CSI). In real-world high-dynamic scenarios, due to quantization errors and feedback delays, the CSI obtained by the base station (BS) for legitimate users and eavesdroppers will inevitably contain errors. Ignoring these CSI uncertainties will not only lead to a sharp increase in the probability of legitimate user outages, compromising the reliability of URLLC, but also cause the covertness constraints to fail, exposing communication behavior.

[0005] Therefore, designing a robust RSMA covert communication resource allocation method under complex conditions of CSI error and FBL constraints is a technical problem that urgently needs to be solved. Summary of the Invention

[0006] To address the shortcomings of the existing technologies, this invention proposes a highly reliable, low-delay covert communication method based on RSMA by considering the actual bounded spherical channel uncertainty model and jointly optimizing the public and private flow precoding matrices of RSMA. The method aims to maximize the minimum average reachability of multiple users under the dual constraints of strict block error rate requirements of URLLC and high concealment requirements of eavesdroppers like Willie.

[0007] On the one hand, the present invention proposes a highly reliable, low-latency covert communication system based on RSMA, which includes: a base station, several legitimate user equipments and at least one illegal eavesdropping node; The base station is used to split the user message to be sent into public messages and private messages and encode them into code blocks of finite block length, modulate the code blocks into wireless signals, and transmit them to each legitimate user equipment through a wireless channel; The legitimate user equipment is used to receive and decode wireless signals from the base station; The illegal eavesdropping node is used to monitor the wireless signals transmitted by the base station to legitimate user devices.

[0008] Furthermore, the base station includes: The channel state information acquisition and evaluation module is used to calculate the nominal channel estimate of legitimate user equipment based on the received uplink signal; acquire the nominal channel estimate of illegal eavesdropping nodes; and calculate the channel uncertainty boundary parameters based on the finite feedback delay and quantization accuracy in the URLLC scenario. The baseband processing unit, with its built-in digital signal processor and convex optimization engine, is used to obtain system configuration parameters. It establishes a non-convex optimization problem with the objective of maximizing the minimum achievable rate among all legal user equipment. Based on the S-procedure lemma, and utilizing channel uncertainty boundary parameters and system configuration parameters, it transforms the channel error constraint in the non-convex optimization problem into a linear matrix inequality. A continuous convex approximation and two-stage iterative algorithm are employed to solve the transformed non-convex optimization problem, yielding RSMA scheduling parameters, the globally optimal transmit precoding matrix for the common data stream, and the optimal transmit precoding matrix for the private data stream. The RSMA signal processing module is used to split the user messages to be transmitted into streams according to RSMA scheduling parameters, and encode the split common messages and private messages of each legitimate user equipment into code blocks of finite block length. The obtained code blocks are modulated to obtain the common data stream and the private data stream of each legitimate user equipment. Based on the global optimal transmit precoding matrix for the common data stream and the optimal transmit precoding matrix for the private data stream, the module precodes the common data stream and the private data stream of each legitimate user equipment to generate a baseband digital transmission signal sequence. The radio frequency front end is used to convert the baseband digital transmission signal sequence into an analog signal, modulate it onto the radio frequency carrier band, and then amplify the modulated radio frequency signal to the set physical transmission power to obtain the transmission signal; Antenna arrays are used to transmit signals to authorized user equipment via wireless channels.

[0009] Furthermore, the channel uncertainty boundary parameters include: the error radius of legitimate users. The error radius with the eavesdropper ; The system configuration parameters include, but are not limited to: maximum transmit power. finite block length Concealment threshold Expected block error rate threshold and background noise variance; The power and block length constraints include: URLLC block error rate constraints, minimum block length constraints, and total base station transmit power constraints; The rate and decoding constraints include: common stream successful decoding constraint, total user rate constraint, and rate non-negativity constraint; The non-convex optimization problem is: to maximize the minimum achievable rate among all legal user devices, and to establish a minimization-maximization non-convex optimization problem under the constraints of power and block length constraints, rate and decoding constraints, and concealment constraints.

[0010] On the other hand, this invention proposes a highly reliable, low-latency covert communication method based on RSMA, which includes the following process: Obtain nominal channel estimates for legitimate user equipment and illegal eavesdropping nodes, and construct a robust channel model that includes channel estimation errors based on a bounded spherical uncertainty model; The base station splits the user messages to be transmitted, encodes the split public messages and the private messages of each legitimate user equipment into code blocks of finite block length, and modulates the obtained code blocks to generate a public data stream and a private data stream of each legitimate user equipment. Based on the robust channel model, the legitimate user equipment uses the RSMA theoretical achievable rate model under finite block length to calculate the theoretical achievable rate of the public data stream and the theoretical achievable rate of the private data stream. Based on the robust channel model, a concealment constraint model based on relative entropy is constructed to ensure that the relative entropy of the received signal by the illegal eavesdropping node satisfies the concealment constraint in both the base station transmitting signal and silent states. With the goal of maximizing the minimum reachable rate among all legal user equipment, a non-convex optimization problem of minimizing maxima is established under the conditions of power and block length constraints, rate and decoding constraints, and concealment constraints. Based on the semidefinite relaxation technique and the S-procedure lemma, the non-convex optimization problem of minimizing maxima is transformed into a deterministic optimization problem; A continuous convex approximation and two-stage iterative algorithm is used to solve the deterministic optimization problem, and the globally optimal transmission precoding matrix for the common data stream and the optimal transmission precoding matrix for the private data stream are extracted. Based on the globally optimal transmit precoding matrix for the public data stream and the optimal transmit precoding matrix for the private data stream, the base station performs weighted merging of the public data stream and the private data stream of each legitimate user equipment to generate a baseband digital transmission signal and transmit it. Legitimate user equipment receives baseband digital transmission signals and uses continuous interference cancellation technology to first decode the public data stream, then decode the private data stream, and recover the corresponding user messages.

[0011] Furthermore, the specific method for obtaining the nominal channel estimates of legitimate user equipment and illegal eavesdropping nodes, and constructing a robust channel model including channel estimation errors based on a bounded spherical uncertainty model, is as follows: Obtain the nominal channel estimates for legitimate user equipment and the nominal channel estimates for illegal eavesdropping nodes, respectively; Based on the nominal channel estimates of the legitimate user equipment and the nominal channel estimates of the illegal eavesdropping nodes, a robust channel model including channel estimation errors is constructed using the worst-case bounded spherical CSI uncertainty model, which restricts the real channel vectors of the legitimate user equipment and the illegal eavesdropping nodes to a continuous uncertainty region. The robust channel model that includes channel estimation error includes: Legitimate User Equipment The true channel vector For: legitimate user equipment nominal channel estimate With channel estimation error The sum of; the channel estimation error Satisfying bounded constraints, it is represented as: channel error vector The norm squared does not exceed the legal user equipment Maximum radius of the error region The square of; illegal eavesdropping nodes The true channel vector For: illegal eavesdropping nodes nominal channel estimate With channel estimation error The sum of; the channel estimation error Satisfying bounded constraints, it is represented as: channel error vector The norm squared does not exceed the illegal eavesdropping node Maximum radius of the error region The square of; legitimate user equipment The variance of the additive white Gaussian noise at the location is expressed as: Illegal eavesdropping nodes The variance of the additive white Gaussian noise at the location is expressed as: .

[0012] Furthermore, based on the robust channel model, the theoretical achievable rate of the public data stream and the theoretical achievable rate of the private data stream, calculated by the legitimate user equipment using the RSMA theoretical achievable rate model with finite block length, are expressed as follows: For any legitimate user equipment Perform the following procedure: Continuous Interference Cancellation (SIC) technology is used to decode the public data stream and identify legitimate user equipment. The signal-to-interference-plus-noise ratio (SIR) of decoding public data streams; The common data stream signal obtained from the decoding is reconstructed and subtracted from the transmitted signal to obtain the remaining signal; Decode the remaining signal for legitimate user equipment Private data streams and identification of legitimate user devices. Decoding the signal-to-interference-plus-noise ratio of private data streams; According to legitimate user equipment The signal-to-interference-plus-noise ratio (SIR) for decoding public data streams and private data streams is calculated using a theoretically achievable rate model based on RSMA, and then the legitimate user equipment (UE) is determined. Decode the reachable rates of public and private data streams.

[0013] Furthermore, the theoretically achievable rate model for RSMA under the finite block length is as follows: Assume that both the public data stream and the private data stream satisfy the maximum tolerable BLER constraint; the maximum tolerable BLER constraint is that the target block error rate corresponding to the public data stream is equal to the target block error rate corresponding to the private data stream. legitimate user equipment Achievable rate of decoding public data streams and legitimate user equipment Achievable rate of decoding private data streams Represented as: ; ; in, For legitimate user equipment The signal-to-interference-plus-noise ratio (SIR) of decoding public data streams; Indicates legitimate user equipment Decoding the signal-to-interference-plus-noise ratio of private data streams; The block length allocated for transmitting public data streams; The block length allocated for transmitting private data streams; The target block error rate for the public data stream; The target block error rate for the private data stream; This represents the inverse function of the Gaussian Q-function; Represents the channel dispersion function; and All of these are penalties.

[0014] Furthermore, the concealment constraint model based on relative entropy is as follows: Under both base station signal transmission and silent states, the KL divergence of the received signal by the illegal eavesdropping node is calculated, ensuring that the KL divergence satisfies the concealment constraint; wherein, the concealment constraint is: a concealment threshold. Twice the square and the maximum block length The ratio; Define auxiliary functions ; Represents variables; By finding the root of the auxiliary function, the concealment constraint model based on relative entropy is transformed into an additive white Gaussian noise variance at the illegal eavesdropping node. The absolute physical limitation is expressed as:

[0015] in, Auxiliary function The root; For illegal eavesdropping nodes The true channel vector The conjugate transpose of; This represents the precoding vector allocated by the base station for the public data stream; This indicates that the base station is a legitimate user equipment. The precoded vector allocated to the private data stream; This represents the set of legitimate user equipment.

[0016] Furthermore, the specific method for transforming the non-convex optimization problem of minimizing maxima into a deterministic optimization problem based on the semidefinite relaxation technique and the S-procedure lemma is as follows: Using a positive semidefinite relaxation technique, the precoding covariance matrix of the common data stream is defined as follows: Legitimate user equipment The precoding covariance matrix of the private data stream is ; The first-order Taylor expansion is used to process the theoretically achievable rate model of RSMA under finite block length, and the first-order Taylor expansion coefficients are obtained. Based on the S-procedure lemma, nonnegative auxiliary relaxation variables are introduced: For legitimate user equipment Signal-to-interference-plus-noise ratio (SIR) constraints for decoding public data streams; For legitimate user equipment Decoding the signal-to-interference-plus-noise ratio (SIR) of private data streams; This is used to constrain the concealment of illegally eavesdropping nodes; Based on the precoding covariance matrix of the public data stream, the precoding covariance matrix of the private data stream, the first-order Taylor expansion coefficients and non-negative auxiliary relaxation variables, the constraint condition of channel estimation error in the minimization-maximization non-convex optimization problem is transformed into a deterministic linear matrix inequality, thus obtaining a deterministic optimization problem. The minimization of the maxima nonconvex optimization problem includes constraints on channel estimation error, including: common stream successful decoding constraint, total user rate constraint, and concealment constraint.

[0017] Furthermore, the specific method for solving the deterministic optimization problem using continuous convex approximation and a two-stage iterative algorithm to extract the globally optimal transmission precoding matrix for the common data stream and the optimal transmission precoding matrix for the private data stream is as follows: The constant penalty factor in deterministic optimization problems Set the variables to zero and construct the first-stage optimization problem with the goal of maximizing the non-negative auxiliary slack variables; in, , To the maximum tolerable block error rate, ; The continuous convex approximation algorithm is used to iteratively solve the first-stage optimization problem, and a feasible solution for the initial precoding covariance matrix is ​​obtained that satisfies the constraints of the total base station transmit power and the concealment linear matrix inequality. Starting with a feasible solution of the initial precoded covariance matrix as the initial iteration point, the following iterative process is performed using a continuous convex approximation algorithm: For the RSMA theoretical achievable rate model with finite block length, a first-order Taylor expansion is performed at the current iteration point to obtain the convex lower bound expression; Substitute the penalty term into the convex lower bound expression to construct the second-stage optimization problem; The second-stage optimization problem is solved using a convex optimization solver to obtain a feasible solution of the precoded covariance matrix for the current iteration process, which is then used as the iteration point for the next iteration process. The feasible solution of the precoded covariance matrix includes: a precoded covariance matrix for the common data stream and a precoded covariance matrix for the private data stream. From the feasible solution of the precoding covariance matrix in the current iteration process, the precoding vectors allocated by the base station for the public data stream and the precoding vectors allocated by the base station for the private data stream of the legitimate user equipment are extracted by eigenvalue decomposition. Using the precoding vectors allocated by the base station for the public data stream and the precoding vectors allocated by the base station for the private data stream of the legitimate user equipment, the signal-to-interference-plus-noise ratio (SIRR) of the legitimate user equipment decoding the public data stream and the SIRR of the legitimate user equipment decoding the private data stream are calculated. Then, using the RSMA theoretical achievable rate model under finite block length, the achievable rate of the legitimate user equipment decoding the public data stream and the achievable rate of decoding the private data stream are calculated. Calculate the maximum minimum average rate for the current iteration round based on the achievable rates of decoding public data streams and private data streams for legitimate user equipment. When the maximum minimum average rate increment of two consecutive iterations is less than the set convergence tolerance, the iteration stops, and the feasible solution of the precoded covariance matrix obtained in the last iteration is taken as the global optimal solution of the precoded covariance matrix. By utilizing the global optimal solution of the precoding covariance matrix, we obtain the globally optimal transmit precoding matrix for the public data stream and the optimal transmit precoding matrix for the private data stream.

[0018] The beneficial effects of adopting the above technical solution are as follows: (1) Improved anti-interference capability and throughput under complex system load: The method of this invention introduces the Rate Split Multiple Access (RSMA) mechanism for physical layer signal splitting and precoding design. In various network environments where the system of this invention evolves from underload to severe overload, the base station can flexibly allocate the power of public and private flows, robustly manage interference between users, effectively overcome the problem of increased interruption probability caused by the limited spatial degrees of freedom in traditional multiple access technology, and improve the overall multi-user fair rate of the system.

[0019] (2) An effective balance between short packet transmission reliability and system concealment is achieved: The method of this invention delves into the core physical trade-off characteristics of transmission block length. Under the premise of strictly limiting the eavesdropper KL divergence (i.e., concealment accumulation constraint), it effectively reduces the inherent reliability penalty of short packet communication through joint resource allocation. The system of this invention ensures extremely low latency while meeting the strict block error rate requirements of URLLC, and achieves a net performance gain at the system level.

[0020] (3) Enhanced physical layer security robustness under imperfect CSI conditions: The method of this invention relies on the minimization-maximization (Max-Min) optimization design of the bounded spherical channel error model, and the radio frequency signal transmitted by the base station can accommodate channel estimation error fluctuations of both legitimate and eavesdropping purposes. Even under severe CSI error boundary conditions, the system of this invention can still ensure from the physical layer that the detection error rate of the eavesdropper is always maintained above the theoretical security lower bound, effectively preventing the exposure of communication behavior and providing highly robust covert communication protection. Attached Figure Description

[0021] Figure 1 This is a structural diagram of a highly reliable, low-latency covert communication system based on RSMA in this embodiment; Figure 2 This is a flowchart of a highly reliable, low-latency covert communication method based on RSMA in this embodiment; Figure 3 The diagram shows the relationship between maximizing the minimum average rate and the base station transmit signal-to-noise ratio (SNR) in this embodiment; where (a) is a comparison diagram of various multiple access schemes under different system loads; and (b) is a comparison diagram with the traditional linear precoding scheme. Figure 4 This is a graph showing the convergence performance of the continuous convex approximation and two-stage iterative algorithms in this embodiment under different concealment constraints and the number of legitimate user equipments. Figure 5 This is a graph showing the trend of maximizing the minimum average rate as a function of concealment requirements in this embodiment; Figure 6 This is a comparison diagram of the relationship between maximizing the minimum average rate and the finite block length in this embodiment; where (a) is the underload region. (a) Schematic diagram of maximizing minimum average rate; (b) Overload region A schematic diagram illustrating the maximization of the minimum average rate; Figure 7 This is a graph showing the relationship between maximizing the minimum average rate and the error boundary radius in this embodiment; Figure 8 This is a comparison chart showing the relationship between system outage probability and target rate requirement in this embodiment; Figure 9 This is a comparison diagram of the configuration relationship between maximizing the minimum average rate and system load (including the number of legitimate user equipment and the number of base station antennas) in this embodiment; wherein, (a) is a schematic diagram of the change of maximizing the minimum average rate with the number of legitimate user equipment; (b) is a schematic diagram of the change of maximizing the minimum average rate with the number of base station antennas. Figure 10 The diagram shows a comparison of the cumulative distribution function (CDF) of the illegal eavesdropping node under robust and non-robust designs in this embodiment. (a) is a schematic diagram of the cumulative distribution function (CDF) of the KL divergence, and (b) is a schematic diagram of the cumulative distribution function (CDF) of the optimal detection error probability of the illegal eavesdropping node. Detailed Implementation

[0022] To facilitate understanding of this application, specific embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and embodiments. The following embodiments are illustrative of the invention but are not intended to limit its scope. Rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of this application.

[0023] Example 1: This embodiment proposes a highly reliable, low-latency covert communication system based on RSMA, such as Figure 1 As shown, the system includes: a base station, several legitimate user devices, and at least one illegal eavesdropping node.

[0024] The base station is configured to split the user message to be sent into public messages and private messages and encode them into code blocks of finite block length, modulate the code blocks into wireless signals, and transmit them to each legitimate user equipment through a wireless channel; and receive uplink signals from legitimate user equipment.

[0025] The legitimate user equipment is used to receive and decode wireless signals from the base station; and to transmit uplink signals to the base station for feedback of channel status information.

[0026] The illegal eavesdropping node is used to monitor the wireless signals transmitted by the base station to legitimate user devices.

[0027] In this embodiment, the highly reliable, low-latency covert communication system based on RSMA includes: a base station (BS) at the downlink transmitting end, and multiple legitimate user equipments (UEs) and at least one illegal eavesdropping node (Willie) at the receiving end. The base station is mainly responsible for complex precoding calculations and signal transmission, while the legitimate user equipments need to have the decoding capability of continuous interference cancellation (SIC).

[0028] Specifically, the configuration in this embodiment is as follows: the number of base station antennas is... The finite block length is Set the Expected Block Error Rate (BLER) threshold to [value]. The concealment threshold is Distance between legitimate user equipment and base station Distributed between 3m and 15m, the distance between Willie and the base station is... The path loss index is The channel uncertainty boundary parameter is ,in The error radius for legitimate users; This represents the error radius of the illegally eavesdropping node. The finite block length is one of the parameters mentioned above. And expected block error rate threshold It is set strictly according to the typical requirements for short packet transmission in the 3GPP URLLC standard protocol; the number and distance distribution of antennas are configured according to the typical physical scenario of microcell indoor / semi-indoor covert communication.

[0029] To achieve the robust precoding and interference management proposed in this invention, the base station includes, but is not limited to, the following interactive and cooperative entity modules in its hardware architecture: The channel state information acquisition and evaluation module is used to calculate the nominal channel estimate of legitimate user equipment based on the received uplink signal; acquire the nominal channel estimate of illegal eavesdropping nodes; and calculate the channel uncertainty boundary parameters based on the finite feedback delay and quantization accuracy in the URLLC scenario.

[0030] The uplink signal includes: pilot signal or CSI feedback information.

[0031] The channel uncertainty boundary parameters include: the error radius of legitimate users. The error radius with the eavesdropper .

[0032] In this embodiment, the channel state information acquisition and evaluation module receives pilot signals or CSI feedback information sent by legitimate users through the base station's radio frequency receiving link to obtain their nominal channel estimates. Meanwhile, the base station typically can only obtain nominal channel estimates for unauthorized eavesdropping nodes (willies) based on the potential geographical boundaries of willies or long-term environmental sensing or direction-finding techniques. Furthermore, this module calculates and initializes channel uncertainty boundary parameters based on the limited feedback delay and quantization accuracy in URLLC scenarios, and outputs them to the baseband processing unit.

[0033] The baseband processing unit, with its built-in digital signal processor and convex optimization engine, is used to obtain system configuration parameters. It establishes a non-convex optimization problem with the objective of maximizing the minimum achievable rate among all legal user equipment. Based on the S-procedure lemma, and utilizing channel uncertainty boundary parameters and system configuration parameters, it transforms the channel error constraint in the non-convex optimization problem into a linear matrix inequality. A continuous convex approximation and two-stage iterative algorithm is employed to solve the transformed non-convex optimization problem, yielding RSMA scheduling parameters, the globally optimal transmit precoding matrix for the common data stream, and the optimal transmit precoding matrix for the private data stream.

[0034] The power and block length constraints include: URLLC block error rate constraints, minimum block length constraints, and total base station transmit power constraints.

[0035] The rate and decoding constraints include: public stream successful decoding constraint, total user rate constraint, and rate non-negativity constraint.

[0036] The non-convex optimization problem is: to maximize the minimum achievable rate among all legal user devices, and to establish a minimization-maximization non-convex optimization problem under the constraints of power and block length constraints, rate and decoding constraints, and concealment constraints.

[0037] The RSMA scheduling parameters include, but are not limited to: the rate allocation vector for the public data stream, the code block length allocation strategy, and the target rate for the private streams of each legitimate user equipment.

[0038] In this embodiment, the Base Band Unit (BBU) is the "central brain" that executes the two-stage iterative optimization algorithm of this invention, and it integrates a Digital Signal Processor (DSP) and a convex optimization solution engine. It acquires the channel uncertainty boundary parameters output by the channel state information acquisition and evaluation module, as well as the set system configuration parameters (such as maximum transmit power). finite block length Concealment threshold Expected block error rate threshold After considering factors such as background noise variance, channel error constraints are constructed. These constraints require that all real channels within the error range defined by the channel uncertainty boundary parameters satisfy the URLLC block error rate constraints for all legitimate user equipment, the successful decoding constraints for the common flow, and the concealment constraints based on relative entropy for illegal eavesdropping nodes. With the objective of maximizing the minimum average reachable rate among all legitimate users, a minimization-maximization (Max-Min) nonconvex optimization problem is established, satisfying the base station total transmit power constraints, the URLLC block error rate constraints for all users, the successful decoding constraints for the common flow, and the Willie concealment constraints based on relative entropy. The BBU is responsible for transforming the nonconvex channel error constraints into linear matrix inequalities (LMI) using the S-procedure lemma, and then calling the built-in optimization engine to perform a two-stage iterative calculation of Successive Convex Approximation (SCA) until the calculation converges, solving for the globally optimal transmit precoding matrix for the common flow. and the optimal transmit precoding matrix for private streams ;in Indicates the first A legitimate user device.

[0039] The RSMA signal processing module is used to split the user messages to be transmitted into streams according to RSMA scheduling parameters, and encode the split common messages and private messages of each legitimate user equipment into code blocks of finite block length. The obtained code blocks are modulated to obtain the common data stream and the private data streams of each legitimate user equipment. Based on the globally optimal transmit precoding matrix for the common data stream and the optimal transmit precoding matrix for the private data stream, the module precodes the common data stream and the private data streams of each legitimate user equipment to generate a baseband digital transmission signal sequence.

[0040] In this embodiment, the RSMA signal processing module and the digital precoder receive the optimal precoding matrix instruction issued by the BBU, including: RSMA scheduling parameters, a globally optimal transmission precoding matrix for the common data stream, and an optimal transmission precoding matrix for the private data stream. On the data plane, the user messages to be transmitted by the base station are subjected to rate-splitting, such as... Figure 1 As shown, user messages from legitimate user devices Each is split into common parts and private parts , This indicates the number of legitimate user devices; all the separated common parts are merged to obtain the separated common information. This process encodes and modulates public information and all the separated private parts into a public data stream. Private data streams for each user Subsequently, using a digital precoder, based on the aforementioned globally optimal transmission precoding matrix... and The corresponding data streams are linearly weighted and spatially multiplexed to achieve precoding and generate a baseband digital transmission signal sequence.

[0041] The radio frequency (RF) front end is used to convert the baseband digital transmission signal sequence into an analog signal, modulate it onto the RF carrier frequency band, and then amplify the modulated RF signal to the set physical transmission power to obtain the transmission signal.

[0042] Antenna arrays are used to transmit signals to authorized user equipment via wireless channels.

[0043] In this embodiment, the RF front-end sequentially converts the synthesized baseband digital transmission signal sequence into an analog signal via a digital-to-analog converter (DAC), then modulates it to the RF carrier frequency band via an up-converter, and finally amplifies it to the set physical transmit power via a power amplifier (PA) to obtain the transmit signal. Ultimately, this transmit signal is transmitted through a circuit containing... The physical antenna array of the antenna radiates electromagnetic waves outward, completing the highly reliable and covert transmission process of the entire downlink.

[0044] Example 2: This embodiment presents a highly reliable, low-latency covert communication method based on RSMA, implemented using the highly reliable, low-latency covert communication system based on RSMA described in Embodiment 1. Figure 2 As shown, the method includes the following steps: Obtain nominal channel estimates for legitimate user equipment and illegal eavesdropping nodes, and construct a robust channel model incorporating channel estimation errors based on a bounded spherical uncertainty model. In this embodiment, for a device Base station with root antenna The base station transmits data using a finite block length, consisting of one single-antenna legitimate user equipment (User, hereinafter also referred to as legitimate user) and one single-antenna illegal eavesdropping node (Willie, hereinafter also referred to as eavesdropper).

[0045] The specific method for obtaining the nominal channel estimates of legitimate user equipment and illegal eavesdropping nodes, and constructing a robust channel model including channel estimation errors based on a bounded spherical uncertainty model, is as follows: Obtain the nominal channel estimates for legitimate user equipment and the nominal channel estimates for illegal eavesdropping nodes, respectively.

[0046] In real-world dynamic communication environments, base stations typically struggle to obtain perfect Channel State Information (CSI). Without loss of generality, we assume that all channels experience independent Rayleigh fading and large-scale path loss. Specifically, from the base station to the node... Channel vector modeling is ,in, For node indexing; Index for legitimate user equipment; Index of nodes used for illegal eavesdropping; Indicates the distance from the base station to the node The channel vector; Indicates a large-scale decline. , Base station to node The distance; The path loss index; This indicates small-scale fading. ; This represents a matrix with a mean of 0 and a covariance equal to the identity matrix. The channel estimation error follows a complex Gaussian distribution. Furthermore, the sources of CSI uncertainty differ significantly between legitimate users and eavesdroppers. For legitimate users, the channel estimation error primarily stems from quantization errors arising under the strict delay constraints of URLLC and outdated channel feedback. Conversely, because the eavesdropper Willie maintains a strictly passive eavesdropping state to avoid detection, the base station cannot obtain its accurate CSI through traditional pilot training. Therefore, the base station typically can only obtain a nominal estimate of Willie's channel based on its potential geographical boundaries or long-term environmental awareness.

[0047] To provide deterministic performance guarantees under the aforementioned imperfect CSI conditions, this embodiment employs a worst-case bounded spherical CSI uncertainty model, restricting the true channel vector to a continuous uncertainty region, and constructing a robust channel model that includes channel estimation errors as follows: Legitimate User Equipment The true channel vector For: legitimate user equipment nominal channel estimate With channel estimation error The sum is expressed as: ; in, Indicates legitimate user equipment The true channel vector; Indicates legitimate user equipment The nominal channel estimate; Indicates legitimate user equipment The channel estimation error satisfies the bounded constraints. , Indicates channel estimation error The conjugate transpose of; For legitimate user equipment The maximum radius of the error region.

[0048] illegal eavesdropping nodes The true channel vector For: illegal eavesdropping nodes nominal channel estimate With channel estimation error The sum is expressed as: ; in, Indicates illegal eavesdropping nodes The true channel vector; Indicates illegal eavesdropping nodes The nominal channel estimate; Indicates illegal eavesdropping nodes The channel estimation error satisfies the bounded constraints. , Indicates channel estimation error The conjugate transpose of; For illegal eavesdropping nodes The maximum radius of the error region.

[0049] legitimate user equipment The variance of the additive white Gaussian noise (AWGN) at point is expressed as: Illegal eavesdropping nodes The variance of the additive white Gaussian noise at the location is expressed as: .

[0050] The base station splits the user messages to be transmitted, encodes the split public messages and the private messages of each legitimate user equipment into code blocks of finite block length, and modulates the obtained code blocks to generate a public data stream and a private data stream of each legitimate user equipment.

[0051] In this embodiment, the base station splits the user message to be sent into public messages and private messages for each legitimate user equipment, based on a finite block length constraint, i.e., not exceeding a preset maximum block length. The process involves encoding the separated public message and the private messages of each legitimate user equipment into code blocks of finite block lengths, and then modulating these code blocks to obtain the public data stream and the private data streams of each legitimate user equipment. Finally, by superimposing the public data stream and the private data streams of each legitimate user equipment, a transmission signal is generated and transmitted to each legitimate user equipment. It should be noted that the lengths of the encoded code blocks for the public message and the private messages of each legitimate user equipment can be the same or different, but they must all meet the requirement that their respective lengths do not exceed a preset maximum block length. Finite block length constraint.

[0052] The transmitted signal is represented as: ; in, Indicates the transmission of signals; The beamforming vector represents the common data stream, which is the precoding vector allocated by the base station to the common data stream; Common stream data symbols sent by the base station; Indicates legitimate user equipment The beamforming vector of the private data stream, i.e., the base station is a legitimate user equipment. The precoded vector allocated to the private data stream; For base station to send to legitimate user equipment Private data stream data symbols.

[0053] Based on the robust channel model, legitimate user equipment uses the RSMA theoretical achievable rate model with finite block length to calculate the theoretical achievable rate of the public data stream and the theoretical achievable rate of the private data stream.

[0054] In this embodiment, the base station utilizes the RSMA strategy to transmit public and private data streams simultaneously. At the legitimate user equipment end, SIC technology is employed to first decode the public stream, then eliminate public stream interference, and finally decode each individual private stream. Considering the FBL effect, the user's decoding rate is approximated by the Shannon capacity including a Q-function penalty term.

[0055] Based on the robust channel model, the legitimate user equipment uses the RSMA theoretical achievable rate model with finite block length to calculate the theoretical achievable rate of the public data stream and the theoretical achievable rate of the private data stream, expressed as follows: For any legitimate user equipment Perform the following procedure: Continuous Interference Cancellation (SIC) technology is used to decode the public data stream and identify legitimate user equipment. Decoding the signal-to-interference-plus-noise ratio of public data streams.

[0056] The common data stream signal obtained from the reconstruction and decoding is subtracted from the transmitted signal to obtain the remaining signal.

[0057] Decode the remaining signal for legitimate user equipment Private data streams and identification of legitimate user devices. Decoding the signal-to-interference-plus-noise ratio of private data streams.

[0058] Specifically, legitimate user equipment After receiving the signal, using Continuous Interference Cancellation (SIC) technology, the public data stream is first decoded, and then its own private data stream is decoded. This allows legitimate user equipment to... The signal-to-interference-plus-noise ratio (SINR) for decoding public and private data streams are expressed as follows: ; ; in, For legitimate user equipment The signal-to-interference-plus-noise ratio (SIR) of decoding public data streams; For legitimate user equipment The true channel vector The conjugate transpose of; This represents the precoding vector allocated by the base station for the public data stream; This indicates that the base station is a legitimate user equipment. The precoded vector allocated to the private data stream; Represents the set of legitimate user equipment. ; Indicates legitimate user equipment Decoding the signal-to-interference-plus-noise ratio of private data streams; This indicates that the base station is a legitimate user equipment. The precoded vectors allocated to the private data stream.

[0059] In the traditional Infinite Block Length (IBL) framework, the achievable rate is typically determined by the Shannon capacity. However, to meet the extremely low latency requirements of URLLC, this embodiment employs Finite Block Length (FBL) transmission.

[0060] Under the FBL framework, due to the introduction of a non-negligible Block Error Rate (BLER), the maximum achievable rate is lower than the classic Shannon capacity. Therefore, the achievable rates of the public and private data streams are respectively constructed using formulas that include a channel dispersion penalty term, i.e., considering the BLER penalty caused by the finite block length. Theoretical achievable rate models based on RSMA are then used to approximately calculate the achievable rates of legitimate user equipment decoding the public and private data streams.

[0061] The theoretically achievable rate model for RSMA under finite block length is as follows: ; ; in, For legitimate user equipment The achievable rate for decoding public data streams; The block length allocated for transmitting public data streams; For legitimate user equipment The achievable rate for decoding private data streams; The block length allocated for transmitting private data streams; The target block error rate for the public data stream; The target block error rate for the private data stream; This represents the inverse function of the Gaussian Q-function; This represents the channel dispersion function.

[0062] To simplify the design without loss of generality, it is assumed that all information flows share a unified maximum tolerable BLER constraint, i.e. Therefore, a constant penalty factor is defined. Furthermore, the channel dispersion function Represented as , Represents a variable.

[0063] Based on the robust channel model, a concealment constraint model based on relative entropy is constructed to ensure that the relative entropy of the received signal by the illegal eavesdropping node satisfies the concealment constraint in both the base station transmitting signal and silent states.

[0064] In covert communication, the illicit eavesdropping node Willie aims to intercept the received signals. Distinguish between the following binary assumptions: This is the base station silent assumption state, used to indicate that the base station remains silent. The base station transmission hypothesis state is used to indicate that the base station is transmitting a signal. The hypothesis testing model is expressed as follows: ; ; in, The vector of additive white Gaussian noise (AWGN) at the illegal eavesdropping node Willie; For illegal eavesdropping nodes The true channel vector The conjugate transpose of; This is the total composite radio frequency signal matrix transmitted by the base station.

[0065] The concealment constraint model based on relative entropy is as follows: Under both base station signal transmission and silent states, the KL divergence of the received signal by the illegal eavesdropping node is calculated, ensuring that the KL divergence satisfies the concealment constraint; wherein, the concealment constraint is: a concealment threshold. Twice the square and the maximum block length The ratio.

[0066] To ensure high concealment, the detection capability of the illegal eavesdropping node Willie must be limited. This requires the lower bound of Willie's detection error rate to approximate an ideal state, i.e., its relative entropy (KL divergence) must satisfy: ; in, Indicates KL divergence; Indicates in the assumption Below is the probability density function of the signal received by the illegal eavesdropping node; Indicates in the assumption Below is the probability density function of the signal received by the illegal eavesdropping node.

[0067] Assuming the signals in each time slot are independent Gaussian signals, then the received signal in and The following are respectively subject to variances of and The complex Gaussian distribution, in which , .

[0068] Define auxiliary functions ; Represents a variable.

[0069] By finding the root of the auxiliary function, the concealment constraint model based on relative entropy is transformed into an additive white Gaussian noise variance at the illegal eavesdropping node. The absolute physical limitation is expressed as: ; in, Auxiliary function The root.

[0070] In this embodiment, by limiting the relative entropy of the signal received by the illegal eavesdropping node under the base station silence assumption state and the base station transmission assumption state, the following condition is met: And using a monotonically decreasing function Finding the root transforms the implicit constraint model based on relative entropy into an equivalent absolute physical constraint on the signal variance.

[0071] With the goal of maximizing the minimum achievable rate among all legal user equipment, a non-convex optimization problem of minimizing maxima is established under the conditions of power and block length constraints, rate and decoding constraints, and concealment constraints.

[0072] In this embodiment, with the goal of maximizing the minimum reachable rate among all legitimate user equipment, a non-convex optimization problem of minimizing maximization (Max-Min) is established under the conditions of satisfying the constraints of successful decoding of common stream, total user rate, non-negativity of rate, URLLC block error rate, minimum block length, total base station transmit power, and the concealment constraint of illegal eavesdropping nodes based on relative entropy.

[0073] The nonconvex optimization problem that minimizes maxima is expressed as: ; ; ; ; ; ; ; ; in, This represents a non-convex optimization problem that minimizes maxima. Represents a set of precoded vectors; Indicates the code block length allocation strategy; A rate allocation vector representing a common data stream; , and All are optimization variables; This represents the set of real channel vectors for legitimate user equipment. Indicates the allocation of public data streams to legitimate user devices. The rate component; Indicates the allocation of public data streams to legitimate user devices. The rate component; Indicates legitimate user equipment The minimum achievable rate threshold; They represent the 1st, 2nd, and 3rd respectively. The block length allocated for the transmission of private data streams by a legitimate user equipment; Represents the trace operation of a matrix; This refers to the maximum transmit power of the base station; This represents the set of real channel vectors of the illegal eavesdropping nodes.

[0074] Based on the semidefinite relaxation technique and the S-procedure lemma, the non-convex optimization problem of minimizing maxima is transformed into a deterministic optimization problem.

[0075] In this embodiment, for an infinite number of non-convex channel uncertainty constraints, a semidefinite relaxation (SDR) technique is used to extract the precoding covariance matrix, and a first-order Taylor expansion is used to process the RSMA theoretical achievable rate model with a finite block length. For the concealment constraint and SINR constraint, auxiliary relaxation variables are introduced, and the S-procedure lemma is used to transform the non-convex robust constraint containing an infinite number of channel error variables in the non-convex optimization problem into a deterministic linear matrix inequality (LMI).

[0076] The specific method for transforming the minimization-maximization non-convex optimization problem into a deterministic optimization problem based on the semidefinite relaxation technique and the S-procedure lemma is as follows: Using a positive semidefinite relaxation technique, the precoding covariance matrix of the common data stream is defined as follows: Legitimate user equipment The precoding covariance matrix of the private data stream is .

[0077] The first-order Taylor expansion is used to process the theoretically achievable rate model of RSMA under finite block length, and the first-order Taylor expansion coefficients are obtained.

[0078] Based on the S-procedure lemma, nonnegative auxiliary relaxation variables are introduced: For legitimate user equipment Signal-to-interference-plus-noise ratio (SIR) constraints for decoding public data streams; For legitimate user equipment Decoding the signal-to-interference-plus-noise ratio (SIR) of private data streams; This is used to constrain the covertness of illegally eavesdropping nodes.

[0079] Based on the precoding covariance matrix of the public data stream, the precoding covariance matrix of the private data stream, the first-order Taylor expansion coefficients, and the non-negative auxiliary relaxation variables, the constraint condition of channel estimation error in the minimization-maximization non-convex optimization problem is transformed into a deterministic linear matrix inequality, thus obtaining a deterministic optimization problem.

[0080] The minimization of the maxima nonconvex optimization problem includes constraints on channel estimation error, including: common stream successful decoding constraint, total user rate constraint, and concealment constraint.

[0081] In this embodiment, auxiliary slack variables are introduced for the concealment constraint and the SINR constraint. , and Using the S-procedure lemma, robust constraints with infinite variables are transformed into deterministic linear matrix inequalities (LMIs). This is particularly relevant for constraints involving channel uncertainty (i.e., channel estimation error). and The quadratic inequalities include: the signal-to-interference-plus-noise ratio (SIR) constraint for legitimate user equipment decoding public data streams (i.e., the constraint for successful public stream decoding), the SIR constraint for legitimate user equipment decoding private data streams (implied in the total user rate constraint, i.e., the private data stream portion included in the total user rate constraint), and the concealment constraint for illegal eavesdropping nodes. Non-negative auxiliary slack variables are introduced. , and Applying the S-procedure lemma, the infinite constraint is rigorously transformed into a semi-definite positive LMI matrix of a definite dimension, i.e., a deterministic linear matrix inequality, as follows:

[0082] in, , and Both consist of the precoding covariance matrix of the public data stream, the precoding covariance matrix of the private data stream, and the nominal channel estimate. or It is composed of first-order Taylor expansion coefficients, and its specific form depends on whether the current data stream is a public data stream signal-to-interference-plus-noise ratio constraint, a private data stream signal-to-interference-plus-noise ratio constraint, or a hidden constraint. For non-negative auxiliary relaxation variables, the expression based on the constraints being processed is as follows: , and Any one of them; The radius of the error boundary is represented as the legal user equipment according to the constraints currently being processed. Maximum radius of the error region or illegal eavesdropping nodes Maximum radius of the error region .

[0083] A continuous convex approximation and two-stage iterative algorithm is used to solve the deterministic optimization problem, and the globally optimal transmit precoding matrix for the public data stream and the optimal transmit precoding matrix for the private data stream are extracted.

[0084] In this embodiment, a two-stage solution algorithm is designed, namely, a continuous convex approximation and two-stage iterative algorithm. Specifically, the first stage sets the FBL penalty factor to zero and quickly finds an initial feasible solution that satisfies the basic power and stealth constraints. The second stage restores the true FBL penalty factor and uses a standard convex optimization solver (such as CVX) combined with trust region constraints for iterative calculation until the convergence tolerance is met. Finally, the globally optimal transmit precoding matrix for the common data stream is output. and the optimal transmit precoding matrix for private data streams .

[0085] The specific method for solving the deterministic optimization problem using continuous convex approximation and a two-stage iterative algorithm to extract the globally optimal transmission precoding matrix for the public data stream and the optimal transmission precoding matrix for the private data stream is as follows: The constant penalty factor in deterministic optimization problems We set the variables to zero and construct the first-stage optimization problem with the goal of maximizing the non-negative auxiliary slack variables.

[0086] in, , To the maximum tolerable block error rate, .

[0087] A continuous convex approximation algorithm is used to iteratively solve the first-stage optimization problem, obtaining a feasible solution for the initial precoding covariance matrix that satisfies the constraints of the total base station transmit power and the concealment linear matrix inequality.

[0088] In this embodiment, the constant penalty factor in the deterministic optimization problem is... The initial solution is set to zero, with the goal of maximizing the non-negative auxiliary slack variables. The continuous convex approximation SCA algorithm is used to quickly find feasible solutions for the initial precoding covariance matrix that strictly satisfy the base station total transmit power constraint and the concealment LMI constraint (which can be obtained by transforming the concealment constraint into a deterministic linear matrix inequality). These solutions include the initial precoding covariance matrix for the public data stream and the initial precoding covariance matrix for the private data stream.

[0089] Starting with a feasible solution of the initial precoded covariance matrix as the initial iteration point, the following iterative process is performed using a continuous convex approximation algorithm: For the RSMA theoretical reachable rate model with finite block length, a first-order Taylor expansion is performed at the current iteration point to obtain the expression for the convex lower bound.

[0090] In this embodiment, the expression for the convex lower bound is: at the current iteration point, the linear approximation function obtained by performing a first-order Taylor expansion on the RSMA theoretical achievable rate model with a finite block length. This function is a lower bound of the original model and is a convex function.

[0091] Substitute the penalty term into the convex lower bound expression to construct the second-stage optimization problem.

[0092] The second-stage optimization problem is solved using a convex optimization solver to obtain a feasible solution for the precoded covariance matrix of the current iteration process, which is then used as the iteration point for the next iteration process. The feasible solution for the precoded covariance matrix includes: a precoded covariance matrix for the public data stream and a precoded covariance matrix for the private data stream.

[0093] From the feasible solution of the precoding covariance matrix in the current iteration process, the precoding vectors allocated by the base station for the public data stream and the precoding vectors allocated by the base station for the private data stream of the legitimate user equipment are extracted by eigenvalue decomposition.

[0094] By utilizing the precoding vectors allocated by the base station for the public data stream and the precoding vectors allocated by the base station for the private data stream of the legitimate user equipment, the signal-to-interference-plus-noise ratio (SIRR) of the legitimate user equipment decoding the public data stream and the SIRR of the legitimate user equipment decoding the private data stream are calculated. Then, using the RSMA theoretical achievable rate model under finite block length, the achievable rate of the legitimate user equipment decoding the public data stream and the achievable rate of decoding the private data stream are calculated.

[0095] Calculate the maximum minimum average rate for the current iteration round based on the achievable rates of decoding public data streams and private data streams for legitimate user equipment.

[0096] When the maximum minimum average rate increment of two consecutive iterations is less than the set convergence tolerance, the iteration stops, and the feasible solution of the precoded covariance matrix obtained in the last iteration is taken as the global optimal solution of the precoded covariance matrix.

[0097] By utilizing the global optimal solution of the precoding covariance matrix, we obtain the globally optimal transmit precoding matrix for the public data stream and the optimal transmit precoding matrix for the private data stream.

[0098] In this embodiment, the complete penalty item (i.e. or Substitute the solution into the convex lower bound expression for iterative correction. Using the feasible solution of the previous stage or the previous iteration as the starting point, use the continuous convex approximation technique to perform Taylor expansion at the current point to construct the convex lower bound. The pre-encoding matrix is ​​continuously updated through the convex optimization solver until the maximum minimum average rate increment of two consecutive iterations is less than the set convergence tolerance.

[0099] Based on the globally optimal transmit precoding matrix for the public data stream and the optimal transmit precoding matrix for the private data stream, the base station performs weighted merging of the public data stream and the private data stream of each legitimate user equipment to generate a baseband digital transmission signal and transmit it.

[0100] In this embodiment, based on the globally optimal transmit precoding matrix for the public data stream and the optimal transmit precoding matrix for the private data stream, the base station performs weighted and spatial multiplexing merging of the public data stream to be transmitted and the private data streams of each legitimate user equipment, generates a baseband digital transmission signal, and transmits it to all legitimate user equipment.

[0101] Legitimate user equipment receives baseband digital transmission signals and uses continuous interference cancellation technology to first decode the public data stream, then decode the private data stream, and recover the corresponding user messages.

[0102] In this embodiment, to verify the effectiveness of the proposed RSMA-based high-reliability, low-latency covert communication method (hereinafter referred to as the present invention), a target signal-to-noise ratio is set, and the performance differences between the present invention and traditional multiple access schemes are compared and analyzed through multiple sets of simulation experiments. The simulation results comprehensively evaluate the performance advantages of the present invention from multiple dimensions, including system rate, convergence, covertness constraints, finite block length transmission, channel estimation error, outage probability, system load, and robustness. The specific analysis is as follows: like Figure 3 As shown, within the set transmit signal-to-noise ratio range, the robust RSMA scheme provided by this invention enables the system to continuously increase its maximum minimum average rate. When the system is under overload, compared with traditional Space Division Multiple Access (SDMA), Non-Orthogonal Multiple Access (NOMA), typical Zero-Forcing (ZF), Regularized Zero-Forcing (RZF), and Maximum Ratio Transmission (MRT) schemes, this invention's scheme mitigates rate saturation by using a common flow for interference cancellation, verifying the effectiveness and stability of this precoding design in handling Channel State Information (CSI) errors and multi-user residual interference.

[0103] Based on this, such as Figure 4As shown, simulation results demonstrate that under different concealment constraint configurations, when the system executes the two-stage iterative optimization algorithm, its objective function (maximizing the minimum average rate) monotonically increases with the number of iterations and reaches stable convergence within a finite number of iterations (approximately 70 to 100). This result proves that the optimization solution framework constructed by the present invention has reliable convergence capability in practical system operations.

[0104] Further analysis of the impact of concealment requirements on system performance, such as... Figure 5 As shown, with the relaxation of concealment requirements (i.e., concealment threshold), the allowed transmit power of the system increases, and the average achievable rate for multiple users increases accordingly. Under this trend, the rate performance of the proposed solution consistently outperforms the benchmark solution, meeting the theoretical trade-off between system rate and concealment constraints.

[0105] Comparing the relationship between system speed and block length under finite block length (FBL) and infinite block length (IBL) frameworks, such as... Figure 6 As shown, under the FBL framework adopted in this invention, the system rate increases with the increase of the transmission block length, indicating that appropriately extending the block length can effectively reduce the reliability penalty of short packet communication. In contrast, the traditional IBL framework exhibits a decreasing rate trend due to the tightening of the covertness divergence threshold with increasing block length. This result verifies that the present invention has better performance adaptability in short packet covert communication scenarios.

[0106] like Figure 7 As shown, the data indicates that within the normal operating error range (such as...) The robust design provided by this invention can maintain a stable system rate. It was also observed that, due to the sensitivity of concealment constraints to global power, the performance degradation caused by the eavesdropper's CSI error is greater than that of the legitimate user. This degradation only occurs in the extremely high error region (…). When the system reaches a certain value, the rate decreases in order to meet the robust constraints, which is consistent with the theoretical characteristics of the bounded error model.

[0107] like Figure 8 As shown, under system overload configuration, the interruption probability of traditional SDMA and NOMA schemes increases rapidly with the increase in target rate requirements. However, the scheme of this invention (RSMA scheme) maintains a low interruption probability at the same target rate by flexibly adjusting the power allocation between the common and private flows, thus expanding the feasible rate range of the system under strict reliability constraints.

[0108] like Figure 9As shown, with the increase in the number of access users (i.e., the system evolves towards an overload state), the proposed solution exhibits a more gradual performance degradation compared to the comparative solution. Furthermore, in scenarios where the number of base station antennas increases, the proposed solution demonstrates a more significant rate increase, proving the effectiveness of this method in joint interference management utilizing spatial degrees of freedom and the power domain.

[0109] like Figure 10 As shown in the simulation results, under conditions of channel uncertainty, the communication actions generated by traditional non-robust designs risk leading to a detection error rate lower than the theoretical safety lower bound. After adopting the robust precoding design of this invention, the empirical cumulative distribution curve remains stably to the right of the theoretical lower bound, verifying that this scheme can effectively ensure that the set concealment index is not compromised under channel error fluctuations.

[0110] In summary, this invention provides a highly reliable, low-latency covert communication method and base station equipment based on RSMA. This technical solution addresses the common channel state information acquisition error problem in multi-user communication scenarios by constructing a joint precoding optimization framework that integrates a finite block length transmission mechanism and the KL divergence covertness criterion through the base station processing unit. At the physical execution level, the base station utilizes the S-procedure lemma and continuous convex approximation technique to transform the non-convex robust constraint containing infinite channel uncertainty into a finite-dimensional linear matrix inequality that can be precisely solved by the underlying processor, thereby generating the optimal signal sequence for controlling the RF front-end transmission.

[0111] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the claims of the present invention.

Claims

1. A highly reliable, low-latency covert communication system based on RSMA, characterized in that, The system includes: a base station, several legitimate user equipment devices, and at least one illegal eavesdropping node; The base station is used to split the user message to be sent into public messages and private messages and encode them into code blocks of finite block length, modulate the code blocks into wireless signals, and transmit them to each legitimate user equipment through a wireless channel; The legitimate user equipment is used to receive and decode wireless signals from the base station; The illegal eavesdropping node is used to monitor the wireless signals transmitted by the base station to legitimate user devices.

2. The highly reliable, low-latency covert communication system based on RSMA according to claim 1, characterized in that, The base station includes: The channel state information acquisition and evaluation module is used to calculate the nominal channel estimate of legitimate user equipment based on the received uplink signal; acquire the nominal channel estimate of illegal eavesdropping nodes; and calculate the channel uncertainty boundary parameters based on the finite feedback delay and quantization accuracy in the URLLC scenario. The baseband processing unit, with its built-in digital signal processor and convex optimization engine, is used to obtain system configuration parameters. It establishes a non-convex optimization problem with the objective of maximizing the minimum achievable rate among all legal user equipment. Based on the S-procedure lemma, and utilizing channel uncertainty boundary parameters and system configuration parameters, it transforms the channel error constraint in the non-convex optimization problem into a linear matrix inequality. A continuous convex approximation and two-stage iterative algorithm are employed to solve the transformed non-convex optimization problem, yielding RSMA scheduling parameters, the globally optimal transmit precoding matrix for the common data stream, and the optimal transmit precoding matrix for the private data stream. The RSMA signal processing module is used to split the user messages to be transmitted into streams according to RSMA scheduling parameters, and encode the split common messages and private messages of each legitimate user equipment into code blocks of finite block length. The obtained code blocks are modulated to obtain the common data stream and the private data stream of each legitimate user equipment. Based on the global optimal transmit precoding matrix for the common data stream and the optimal transmit precoding matrix for the private data stream, the module precodes the common data stream and the private data stream of each legitimate user equipment to generate a baseband digital transmission signal sequence. The radio frequency front end is used to convert the baseband digital transmission signal sequence into an analog signal, modulate it onto the radio frequency carrier band, and then amplify the modulated radio frequency signal to the set physical transmission power to obtain the transmission signal; Antenna arrays are used to transmit signals to authorized user equipment via wireless channels.

3. The highly reliable, low-latency covert communication system based on RSMA according to claim 2, characterized in that, The channel uncertainty boundary parameters include: the error radius of legitimate users. The error radius with the eavesdropper ; The system configuration parameters include, but are not limited to: maximum transmit power. finite block length Concealment threshold Expected block error rate threshold and background noise variance; The power and block length constraints include: URLLC block error rate constraints, minimum block length constraints, and total base station transmit power constraints; The rate and decoding constraints include: common stream successful decoding constraint, total user rate constraint, and rate non-negativity constraint; The non-convex optimization problem is: to maximize the minimum achievable rate among all legal user devices, and to establish a minimization-maximization non-convex optimization problem under the constraints of power and block length constraints, rate and decoding constraints, and concealment constraints.

4. A highly reliable, low-latency covert communication method based on RSMA, implemented using the highly reliable, low-latency covert communication system based on RSMA as described in any one of claims 1-3, characterized in that, This method includes the following steps: Obtain nominal channel estimates for legitimate user equipment and illegal eavesdropping nodes, and construct a robust channel model that includes channel estimation errors based on a bounded spherical uncertainty model; The base station splits the user messages to be transmitted, encodes the split public messages and the private messages of each legitimate user equipment into code blocks of finite block length, and modulates the obtained code blocks to generate a public data stream and a private data stream of each legitimate user equipment. Based on the robust channel model, the legitimate user equipment uses the RSMA theoretical achievable rate model under finite block length to calculate the theoretical achievable rate of the public data stream and the theoretical achievable rate of the private data stream. Based on the robust channel model, a concealment constraint model based on relative entropy is constructed to ensure that the relative entropy of the received signal by the illegal eavesdropping node satisfies the concealment constraint in both the base station transmitting signal and silent states. With the goal of maximizing the minimum reachable rate among all legal user equipment, a non-convex optimization problem of minimizing maxima is established under the conditions of power and block length constraints, rate and decoding constraints, and concealment constraints. Based on the semidefinite relaxation technique and the S-procedure lemma, the non-convex optimization problem of minimizing maxima is transformed into a deterministic optimization problem; A continuous convex approximation and two-stage iterative algorithm is used to solve the deterministic optimization problem, and the globally optimal transmission precoding matrix for the common data stream and the optimal transmission precoding matrix for the private data stream are extracted. Based on the globally optimal transmit precoding matrix for the public data stream and the optimal transmit precoding matrix for the private data stream, the base station performs weighted merging of the public data stream and the private data stream of each legitimate user equipment to generate a baseband digital transmission signal and transmit it. Legitimate user equipment receives baseband digital transmission signals and uses continuous interference cancellation technology to first decode the public data stream, then decode the private data stream, and recover the corresponding user messages.

5. The highly reliable, low-latency covert communication method based on RSMA according to claim 4, characterized in that, The specific method for obtaining the nominal channel estimates of legitimate user equipment and illegal eavesdropping nodes, and constructing a robust channel model including channel estimation errors based on a bounded spherical uncertainty model, is as follows: Obtain the nominal channel estimates for legitimate user equipment and the nominal channel estimates for illegal eavesdropping nodes, respectively; Based on the nominal channel estimates of the legitimate user equipment and the nominal channel estimates of the illegal eavesdropping nodes, a robust channel model including channel estimation errors is constructed using the worst-case bounded spherical CSI uncertainty model, which restricts the real channel vectors of the legitimate user equipment and the illegal eavesdropping nodes to a continuous uncertainty region. The robust channel model that includes channel estimation error includes: Legitimate User Equipment The true channel vector For: legitimate user equipment nominal channel estimate With channel estimation error The sum of; the channel estimation error Satisfying bounded constraints, it is represented as: channel error vector The norm squared does not exceed the legal user equipment Maximum radius of the error region The square of; illegal eavesdropping nodes The true channel vector For: illegal eavesdropping nodes nominal channel estimate With channel estimation error The sum of; the channel estimation error Satisfying bounded constraints, it is represented as: channel error vector The norm squared does not exceed the illegal eavesdropping node Maximum radius of the error region The square of; legitimate user equipment The variance of the additive white Gaussian noise at the location is expressed as: Illegal eavesdropping nodes The variance of the additive white Gaussian noise at the location is expressed as: .

6. The highly reliable, low-latency covert communication method based on RSMA according to claim 5, characterized in that, Based on the robust channel model, the legitimate user equipment uses the RSMA theoretical achievable rate model with finite block length to calculate the theoretical achievable rate of the public data stream and the theoretical achievable rate of the private data stream, expressed as follows: For any legitimate user equipment Perform the following procedure: Continuous Interference Cancellation (SIC) technology is used to decode the public data stream and identify legitimate user equipment. The signal-to-interference-plus-noise ratio (SIR) of decoding public data streams; The common data stream signal obtained from the decoding is reconstructed and subtracted from the transmitted signal to obtain the remaining signal; Decode the remaining signal for legitimate user equipment Private data streams and identification of legitimate user devices. Decoding the signal-to-interference-plus-noise ratio of private data streams; According to legitimate user equipment The signal-to-interference-plus-noise ratio (SIR) for decoding public data streams and private data streams is calculated using a theoretically achievable rate model based on RSMA, and then the legitimate user equipment (UE) is determined. Decode the reachable rates of public and private data streams.

7. The highly reliable, low-latency covert communication method based on RSMA according to claim 6, characterized in that, The theoretically achievable rate model for RSMA under finite block length is as follows: Assume that both the public data stream and the private data stream satisfy the maximum tolerable BLER constraint; the maximum tolerable BLER constraint is that the target block error rate corresponding to the public data stream is equal to the target block error rate corresponding to the private data stream. legitimate user equipment Achievable rate of decoding public data streams and legitimate user equipment Achievable rate of decoding private data streams Represented as: ； ； in, For legitimate user equipment The signal-to-interference-plus-noise ratio (SIR) of decoding public data streams; Indicates legitimate user equipment Decoding the signal-to-interference-plus-noise ratio of private data streams; The block length allocated for transmitting public data streams; The block length allocated for transmitting private data streams; The target block error rate for the public data stream; The target block error rate for the private data stream; This represents the inverse function of the Gaussian Q-function; Represents the channel dispersion function; and All of these are penalties.

8. The highly reliable, low-latency covert communication method based on RSMA according to claim 7, characterized in that, The concealment constraint model based on relative entropy is as follows: Under both base station signal transmission and silent states, the KL divergence of the received signal by the illegal eavesdropping node is calculated, ensuring that the KL divergence satisfies the concealment constraint; wherein, the concealment constraint is: a concealment threshold. Twice the square and the maximum block length The ratio; Define auxiliary functions ; Represents variables; By finding the root of the auxiliary function, the concealment constraint model based on relative entropy is transformed into an additive white Gaussian noise variance at the illegal eavesdropping node. The absolute physical limitation is expressed as: in, Auxiliary function The root; For illegal eavesdropping nodes The true channel vector The conjugate transpose of; This represents the precoding vector allocated by the base station for the public data stream; This indicates that the base station is a legitimate user equipment. The precoded vector allocated to the private data stream; This represents the set of legitimate user equipment.

9. The highly reliable, low-latency covert communication method based on RSMA according to claim 8, characterized in that, The specific method for transforming the minimization-maximization non-convex optimization problem into a deterministic optimization problem based on the semidefinite relaxation technique and the S-procedure lemma is as follows: Using a positive semidefinite relaxation technique, the precoding covariance matrix of the common data stream is defined as follows: Legitimate user equipment The precoding covariance matrix of the private data stream is ; The first-order Taylor expansion is used to process the theoretically achievable rate model of RSMA under finite block length, and the first-order Taylor expansion coefficients are obtained. Based on the S-procedure lemma, nonnegative auxiliary relaxation variables are introduced: For legitimate user equipment Signal-to-interference-plus-noise ratio (SIR) constraints for decoding public data streams; For legitimate user equipment Decoding the signal-to-interference-plus-noise ratio (SIR) of private data streams; This is used to constrain the concealment of illegally eavesdropping nodes; Based on the precoding covariance matrix of the public data stream, the precoding covariance matrix of the private data stream, the first-order Taylor expansion coefficients and non-negative auxiliary relaxation variables, the constraint condition of channel estimation error in the minimization-maximization non-convex optimization problem is transformed into a deterministic linear matrix inequality, thus obtaining a deterministic optimization problem. The minimization of the maxima nonconvex optimization problem includes constraints on channel estimation error, including: common stream successful decoding constraint, total user rate constraint, and concealment constraint.

10. The highly reliable, low-latency covert communication method based on RSMA according to claim 9, characterized in that, The specific method for solving the deterministic optimization problem using continuous convex approximation and a two-stage iterative algorithm to extract the globally optimal transmission precoding matrix for the public data stream and the optimal transmission precoding matrix for the private data stream is as follows: The constant penalty factor in deterministic optimization problems Set the variables to zero and construct the first-stage optimization problem with the goal of maximizing the non-negative auxiliary slack variables; in, , To the maximum tolerable block error rate, ; The continuous convex approximation algorithm is used to iteratively solve the first-stage optimization problem, and a feasible solution for the initial precoding covariance matrix is ​​obtained that satisfies the constraints of the total base station transmit power and the concealment linear matrix inequality. Starting with a feasible solution of the initial precoded covariance matrix as the initial iteration point, the following iterative process is performed using a continuous convex approximation algorithm: For the RSMA theoretical achievable rate model with finite block length, a first-order Taylor expansion is performed at the current iteration point to obtain the convex lower bound expression; Substitute the penalty term into the convex lower bound expression to construct the second-stage optimization problem; The second-stage optimization problem is solved using a convex optimization solver to obtain a feasible solution of the precoded covariance matrix for the current iteration process, which is then used as the iteration point for the next iteration process. The feasible solution of the precoded covariance matrix includes: a precoded covariance matrix for the common data stream and a precoded covariance matrix for the private data stream. From the feasible solution of the precoding covariance matrix in the current iteration process, the precoding vectors allocated by the base station for the public data stream and the precoding vectors allocated by the base station for the private data stream of the legitimate user equipment are extracted by eigenvalue decomposition. Using the precoding vectors allocated by the base station for the public data stream and the precoding vectors allocated by the base station for the private data stream of the legitimate user equipment, the signal-to-interference-plus-noise ratio (SIRR) of the legitimate user equipment decoding the public data stream and the SIRR of the legitimate user equipment decoding the private data stream are calculated. Then, using the RSMA theoretical achievable rate model under finite block length, the achievable rate of the legitimate user equipment decoding the public data stream and the achievable rate of decoding the private data stream are calculated. Calculate the maximum minimum average rate for the current iteration round based on the achievable rates of decoding public data streams and private data streams for legitimate user equipment. When the maximum minimum average rate increment of two consecutive iterations is less than the set convergence tolerance, the iteration stops, and the feasible solution of the precoded covariance matrix obtained in the last iteration is taken as the global optimal solution of the precoded covariance matrix. By utilizing the global optimal solution of the precoding covariance matrix, we obtain the globally optimal transmit precoding matrix for the public data stream and the optimal transmit precoding matrix for the private data stream.

Images