Computing task safe unloading method and device integrating communication and perception, equipment and medium

Abstract

The invention relates to the technical field of wireless communication and network security, and discloses a communication and sensing integrated computing task security unloading method, device, equipment and medium, the method comprises the following steps: emitting a composite signal, and receiving an echo signal corresponding to a sensing signal, the composite signal comprising a communication signal and the sensing signal; estimating an eavesdropping position according to the echo signal, wherein the eavesdropping position comprises an angle uncertainty interval of an eavesdropper; the system state information and the eavesdropping position are input into a decision model, a control strategy is output through the decision model, and the control strategy is used for adjusting the direction and energy of the composite signal, so that the communication signal received by the receiving end is enhanced, and other received signals are weakened; the sensing signal received by the eavesdropper is enhanced and the communication signal received by the eavesdropper is weakened; according to the method, the position of the eavesdropper is actively sensed, and the wave beam is regulated and controlled, so that the task is safely unloaded, and the safety and the resource utilization rate are improved.

Description

Technical Field

[0001] This invention relates to the fields of wireless communication and network security technology, specifically to a method, apparatus, device, and medium for securely offloading computing tasks that integrates communication and sensing. Background Technology

[0002] In current wireless communication systems, especially those designed for complex scenarios such as the Internet of Things (IoT) and vehicle-to-everything (V2X) networks, network nodes often need to possess environmental awareness capabilities while performing their primary communication tasks, enabling them to understand and respond to the surrounding situation. This is to prevent malicious eavesdropping nodes (such as malicious drones) from intercepting communication signals transmitted from the transmitter (such as IoT devices) to the receiver (such as satellites). However, in existing secure transmission schemes, the sensing function (for scanning for eavesdroppers) and the communication function (for transmitting useful communication signals) are often implemented using different devices, which is detrimental to the requirements of low-cost resources and high communication efficiency. Summary of the Invention

[0003] This invention provides a method, apparatus, device, and medium for securely offloading computing tasks that integrates communication and sensing, in order to solve the problem that existing wireless communication systems lack adaptive protection against mobile eavesdropping threats in dynamic and open environments.

[0004] In a first aspect, the present invention provides a method for securely offloading computing tasks by integrating communication and sensing. The method includes: transmitting a composite signal and receiving a corresponding echo signal, wherein the composite signal includes a communication signal and a sensing signal, and the echo signal is the echo signal corresponding to the sensing signal; estimating the eavesdropper's location based on the echo signal, wherein the eavesdropping location includes altitude and direction information; inputting system state information and the eavesdropping location into a decision model, and outputting a control strategy through the decision model, wherein the system state information is the state information of the communication network, and the communication network is the communication network of the transmitting end and the receiving end; the control strategy is used to adjust the direction and energy of the composite signal to enhance the communication signal received by the receiving end and weaken other received signals, and to enhance the sensing signal received by the eavesdropper and weaken the received communication signal; and adjusting the composite signal according to the control strategy to securely offload the computing task to the receiving end.

[0005] In one optional implementation, the system status information includes channel status information from the transmitter to the receiver, the signal-to-interference-plus-noise ratio (SIR) of the current communication link, the SIR of the current sensing link, and the instantaneous power consumption of the transmitter.

[0006] In one optional implementation, the decision model is a deep reinforcement learning model. The system state information and the eavesdropping location are input into the decision model, and the decision model outputs a control strategy. This includes: inputting the system state information and the eavesdropping location as environmental information for reinforcement learning into the decision model, and outputting a communication covariance matrix and a perception covariance matrix as the action space through the decision model; generating a first beamforming vector for controlling communication signals based on the communication covariance matrix; and generating a second beamforming vector for controlling perception signals based on the perception covariance matrix.

[0007] In one optional implementation, when there are multiple transmitters, the decision model outputs a communication covariance matrix and a perception covariance matrix based on system state information and eavesdropping locations. This includes: receiving system state information and eavesdropping locations transmitted by each transmitter; analyzing all received system state information and all eavesdropping locations as environmental information to obtain a set of covariance matrices as the action space; and feeding back the target communication covariance matrix and the target perception covariance matrix corresponding to the target transmitter from the covariance matrix set to the target transmitter.

[0008] In one optional implementation, estimating the eavesdropper's location based on the echo signal includes: estimating the eavesdropper's reference orientation angle and height information based on the echo signal; modeling an angle uncertainty interval using perception error and the reference orientation angle, and using the angle uncertainty interval as the eavesdropper's orientation information.

[0009] In one optional implementation, the reward function of the decision model includes at least one of a first scoring term, a second scoring term, a third scoring term, and a fourth scoring term; the first scoring term is determined based on the lowest data transmission rate among each transmitter; the second scoring term is determined based on the signal-to-interference-plus-noise ratio (SNR) of the sensing link at each transmitter; the third scoring term is determined based on the maximum eavesdropping SNR, which is the maximum predicted SNR within the angular uncertainty interval corresponding to each transmitter, and the eavesdropping SNR is the SNR of the eavesdropping link when the eavesdropper is eavesdropping; the fourth scoring term is determined based on the instantaneous power consumption of the transmitter.

[0010] In one optional implementation, the sum of the communication covariance matrix and the sensing covariance matrix corresponding to any transmitter satisfies a preset maximum transmit power constraint.

[0011] Secondly, the present invention provides a computing task security offloading device integrating communication and sensing. The device includes: a signal synthesis and transmission module for transmitting a composite signal and receiving a corresponding echo signal, wherein the composite signal includes a communication signal and a sensing signal, and the echo signal is the echo signal corresponding to the sensing signal; a positioning module for estimating the eavesdropper's location based on the echo signal, wherein the eavesdropping location includes height and direction information; a strategy decision module for inputting system state information and the eavesdropping location into a decision model and outputting a control strategy through the decision model, wherein the system state information is the state information of the communication network, and the communication network is the communication network of the transmitting end and the receiving end, and the control strategy is used to adjust the direction and energy of the composite signal to enhance the communication signal received by the receiving end and weaken other received signals, and to enhance the sensing signal received by the eavesdropper and weaken the received communication signal; and a composite signal adjustment and task offloading module for adjusting the composite signal according to the control strategy to securely offload the task to the receiving end.

[0012] Thirdly, the present invention provides an electronic device comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the computer instructions to perform the integrated communication and sensing computing task secure offloading method of the first aspect or any corresponding embodiment described above.

[0013] Fourthly, the present invention provides a computer-readable storage medium storing computer instructions for causing a computer to execute the integrated communication and sensing computing task secure offloading method of the first aspect or any corresponding embodiment described above.

[0014] Fifthly, the present invention provides a computer program product, including computer instructions for causing a computer to execute the integrated communication and sensing computing task secure offloading method of the first aspect or any corresponding embodiment described above.

[0015] The core advantage of this invention lies in integrating environmental perception, intelligent decision-making, and secure communication into an adaptive security offloading paradigm. By transmitting composite signals and analyzing the echoes of the perceived signals to obtain eavesdropper location information, it eliminates the dependence of traditional physical layer security schemes on prior information about the eavesdropping channel, addressing the problem of security strategies failing due to information lag in dynamic open environments. The system inputs the eavesdropping location and real-time network status into a decision model, which outputs a collaborative control strategy. This strategy has a dual guiding role: guiding the energy of the communication signal to concentrate on the legitimate receiving end, improving signal transmission efficiency and reliability; simultaneously guiding the energy of the perceived signal to the location of the eavesdropper, so that when it performs detection functions, it is transformed into directional artificial noise, interfering with the eavesdropper, thus achieving synergy between perception and security. The system generates composite signals according to the strategy, integrating communication and security protection at the physical layer. This ensures the confidentiality of computing task transmission, improves the system's robustness, adaptability, and overall performance against eavesdropping threats in complex wireless environments, and provides an inherently secure transmission solution for mobile communication systems. Attached Figure Description

[0016] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a schematic flowchart of a first method for securely offloading computing tasks that integrates communication and sensing according to an embodiment of the present invention.

[0018] Figure 2 This is a second flowchart illustrating a method for securely offloading computing tasks that integrates communication and sensing according to an embodiment of the present invention.

[0019] Figure 3 This is a schematic diagram of the deep reinforcement learning algorithm policy update process for a secure offloading method for integrated communication and perception computing tasks according to an embodiment of the present invention.

[0020] Figure 4 The PPO-based algorithm of the integrated communication and sensing computing task security offloading method according to an embodiment of the present invention is a schematic diagram of the maximum eavesdropping signal-to-noise ratio convergence graph.

[0021] Figure 5 This is a schematic diagram comparing the sensing signal-to-noise ratio (SINR) of a computing task security offloading method integrating communication and sensing according to an embodiment of the present invention.

[0022] Figure 6This is a schematic diagram comparing the minimum rates of a method for securely offloading computing tasks that integrates communication and sensing according to an embodiment of the present invention.

[0023] Figure 7 This is a structural block diagram of a computing task security offloading device integrating communication and sensing according to an embodiment of the present invention;

[0024] Figure 8 This is a schematic diagram of the hardware structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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, 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.

[0026] It is understood that before using the technical solutions disclosed in the various embodiments of the present invention, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in the present invention and their authorization should be obtained in accordance with relevant laws and regulations through appropriate means.

[0027] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0028] As the Internet of Things (IoT) expands into remote areas such as oceans and mountains, IoT devices, constrained by computing resources, need to offload computationally intensive tasks to receivers with greater computing resources to reduce latency and energy consumption. For example, traditional terrestrial communication networks struggle to provide continuous coverage. Integrated space-air-ground networks, by merging low-Earth orbit satellites, airborne platforms, and ground nodes, have become a key architecture for achieving seamless global connectivity. In this architecture, satellites allow resource-constrained IoT devices to offload computationally intensive tasks to them. The satellites then process the computations and return the results to the IoT devices, further reducing latency and energy consumption.

[0029] In the aforementioned schemes, some computationally intensive tasks involve financial or military secrets, requiring data security. However, the secure transmission mechanisms in related wireless communication systems have limitations when facing dynamic and open wireless environments. Traditional physical layer security schemes are typically based on idealized channel assumptions, and their core problems mainly lie in the following aspects:

[0030] 1. To prevent eavesdropping, it is necessary to detect eavesdroppers in the vicinity of the environment. However, the detection function (used to scan for eavesdroppers) and the communication function (transmitting communication signals as a task) of existing secure transmission schemes are often implemented by different devices, which is not conducive to the requirements of low cost resources and high communication efficiency.

[0031] 2. Related technologies rely excessively on channel state information. Most secure transmission mechanisms require accurate channel state information for eavesdropping, including parameters such as channel gain, phase, and the location of the eavesdropper. However, in real-world dynamic environments, due to the stealth and mobility of eavesdroppers, obtaining complete channel state information is quite difficult. This reliance on ideal channel conditions causes existing solutions to experience significant performance degradation in real-world testing.

[0032] 3. Most related technologies employ static or semi-static security strategies, which cannot effectively track the eavesdropper's movement trajectory and rapid changes in channel conditions. When the eavesdropper's location changes or channel conditions change, existing beamforming schemes and power allocation strategies are difficult to adjust in real time, resulting in reduced security protection effectiveness.

[0033] 4. Related technical issues are insufficient in dealing with uncertain threats. Due to the lack of real-time awareness of the location of eavesdroppers, existing security mechanisms are unable to provide continuous protection against mobile eavesdropping threats. When there are multiple eavesdroppers in the system or the location of eavesdroppers is unclear, traditional solutions often adopt conservative resource allocation strategies, which reduce communication efficiency and cannot ensure security performance.

[0034] The aforementioned issues collectively limit the secure transmission capabilities of relevant wireless communication systems in dynamic and open environments. In particular, in application scenarios with high requirements for real-time performance and reliability, such as task offloading, existing technologies struggle to provide reliable security while ensuring communication quality.

[0035] Therefore, this invention proposes a method for securely offloading computing tasks that integrates communication and sensing.

[0036] According to an embodiment of the present invention, an embodiment of a method for securely offloading computing tasks integrating communication and sensing is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0037] This embodiment provides a method for securely offloading computing tasks that integrates communication and sensing. Figure 1 This is a flowchart of a method for securely offloading computing tasks integrating communication and sensing according to an embodiment of the present invention, such as... Figure 1 As shown, the process includes the following steps:

[0038] Step S101: Transmit a composite signal and receive the corresponding echo signal. The composite signal includes a communication signal and a sensing signal, and the echo signal is the echo signal corresponding to the sensing signal.

[0039] Step S102: Estimate the eavesdropper's location based on the echo signal. The eavesdropping location includes height and direction information.

[0040] Step S103: Input the system status information and the eavesdropping location into the decision model, and output the control strategy through the decision model. The system status information is the status information of the communication network, which is the communication network of the transmitting end and the receiving end. The control strategy is used to adjust the direction and energy of the composite signal so that the communication signal received by the receiving end is enhanced and other received signals are weakened, and the perceived signal received by the eavesdropper is enhanced and the received communication signal is weakened.

[0041] Step S104: Adjust the composite signal according to the control strategy to safely offload the task to the receiving end.

[0042] The integrated communication and sensing computing task offloading method provided in this invention is applicable to various network architectures, including but not limited to transmitting devices in scenarios such as integrated air-space-ground networks, vehicle-to-everything (V2X) networks, industrial IoT, and terrestrial cellular networks. The transmitting end can be a drone, mobile terminal, roadside unit, satellite, or any communication node with signal transmission and sensing capabilities. Correspondingly, the receiving end can be a satellite gateway, edge server, vehicle-mounted terminal, or any network node with computing capabilities.

[0043] First, the transmitting end transmits a composite signal via a radio frequency front-end. This composite signal is formed by fusing communication signals for data transmission and sensing signals for environmental detection. The sensing signals employ pre-designed radar waveforms, such as linear frequency modulated signals or pseudo-random sequences. When the composite signal encounters an eavesdropper during propagation, the eavesdropper, acting as an unverified malicious node attempting to intercept communication, will scatter the signal. The scattered waves generated by the sensing signal components form an echo signal carrying information about the target's location and motion. The transmitting end captures this echo through a receiving link, completing the raw data acquisition of environmental information.

[0044] In one example, in an integrated air-space-ground network consisting of one or more terrestrial IoT devices with Integrated Sensing and Communication (ISAC) capabilities, a Low Earth Orbit (LEO) satellite (as an edge computing node), and a malicious drone of unknown location (as a mobile eavesdropper), the antenna arrays and maximum transmission power of each device are first assessed. Sensing accuracy threshold And the safe SINR (Signal to Interference plus Noise Ratio) threshold The configuration is then performed. Subsequently, each device actively transmits composite signals and receives radar echoes from potential targets in the environment.

[0045] Next, parameter estimation is performed on the echo signal, that is, target parameters are extracted from noisy observations using signal processing algorithms. Using direction-of-arrival (DOA) estimation algorithms, such as the MUSIC or Capon algorithms, the system can accurately calculate the three-dimensional spatial relationship between the eavesdropper and the transmitter, including horizontal azimuth, elevation, and relative distance information, thereby determining the eavesdropping location. This location information will serve as a key input for subsequent beamforming strategies, enabling the system to shift from passive defense to active threat detection. For example, the azimuth angle of an eavesdropping drone can be estimated based on the echo. .

[0046] After obtaining the eavesdropping location, the system state information and the eavesdropping location are jointly input into the decision model. The system state information refers to a set of multi-dimensional parameters including the channel matrix, signal-to-interference-plus-noise ratio (SNR), and device power consumption. The decision model generates control commands through forward propagation calculations, outputting a mathematical description of the transmitted beam shape and energy distribution—the control strategy. The core objective of this control strategy is to jointly optimize the transmission direction and energy distribution of the composite signal (i.e., the fusion of communication and sensing signals) in the spatial domain. Its specific function is manifested in two levels of coordinated regulation: firstly, phase modulation is used to create a main lobe gain in the receiving direction of the communication signal (i.e., beamforming makes the signal energy in this direction higher than in other directions), while simultaneously creating nulls in non-target directions (i.e., beamforming suppresses the signal energy in this direction); secondly, the sensing signal is concentrated in the direction of the eavesdropping location, achieving a unified function of detection and interference. Communication signals refer to useful signals that need to be offloaded to satellites for computation, while sensing signals, because they carry useless information, increase their energy and concentrate their transmission towards the eavesdropping location. This can be used as interference signals for eavesdroppers, causing them to capture a large amount of useless information and making it difficult for them to capture useful information.

[0047] Finally, a composite signal is generated according to the control strategy. This involves a digital signal processing process where multiple signal components are weighted and superimposed by the baseband processor, forming a set of communication and sensing waveforms superimposed in the time and frequency domains. This signal is emitted through an antenna array, specifically undergoing digital-to-analog conversion, up-conversion, and power amplification before being radiated into space. Through beamforming technology, the communication signal energy is focused towards the receiving end, while the sensing signal energy is focused on the eavesdropping area, creating spatial shielding. This allows computationally intensive tasks to be wirelessly transmitted to a remote server for processing.

[0048] In one example, each device Constructed composite signal for:

[0049]

[0050] in, This is the communication beamforming vector, used to transmit computational tasks to the satellite; For sensing / artificial noise beamforming vectors; For normalized communication symbols, It is a dedicated sensing reference signal (without carrying information).

[0051] Next, adaptive beamforming and security task offloading are performed. Based on the observed state, the system invokes the trained policy network and outputs the communication covariance matrix for each device. With the perception covariance matrix And based on this, generate the actual beam vector. and The equipment transmits signals according to this configuration, safely offloading computing tasks to the LEO satellite.

[0052] The integrated communication and sensing computing task offloading method provided in this invention actively detects and estimates the location of unknown eavesdroppers by transmitting composite signals and processing their echoes. This frees the system from reliance on prior knowledge of the eavesdropping channel, transforming traditional passive defense into proactive threat suppression based on environmental awareness. By constraining the composite signal, the sensing signal is simultaneously used as artificial noise against the eavesdropper, achieving a dual-purpose effect. The same time, spectrum, and power resources are used for both data transmission and security interference, improving the overall resource efficiency of the system. Spatial energy regulation is achieved through control strategies output by the decision model, which can control beamforming to concentrate communication signal energy towards the legitimate receiver while simultaneously concentrating sensing / interference signal energy towards the eavesdropper's location. This spatial dimension control improves the receiver's signal-to-noise ratio, ensures task offloading reliability, suppresses the signal quality of the eavesdropping link, and achieves synergistic optimization of communication performance and physical layer security. It possesses broad applicability and deployment flexibility. Furthermore, the sensing and communication functions can be realized simultaneously through a composite signal. Sending the composite signal through a single device also saves equipment costs. The method is applicable to various dynamic network architectures such as integrated air-space-ground networks, vehicle-to-everything (V2X) networks, and industrial IoT. It can be deployed on various nodes such as drones, satellites, and mobile terminals, demonstrating its scenario adaptability and practical value.

[0053] This embodiment provides a method for securely offloading computing tasks that integrates communication and sensing. Figure 2 This is a flowchart of a method for securely offloading computing tasks integrating communication and sensing according to an embodiment of the present invention, such as... Figure 2 As shown, the process includes the following steps:

[0054] Step S201: Transmit the sensing signal and receive the echo signal corresponding to the sensing signal. For details, please refer to [link to relevant documentation]. Figure 1 Step S101 of the illustrated embodiment will not be described again here.

[0055] Step S202: Estimate the eavesdropper's location based on the echo signal. The eavesdropping location includes altitude and direction information. See details below. Figure 1 Step S102 of the illustrated embodiment will not be described again here.

[0056] In an optional implementation, step S202 includes:

[0057] Step S2021: Estimate the reference direction angle of the eavesdropper based on the echo signal.

[0058] Step S2022: The true location of the eavesdropper is modeled as an angular uncertainty interval by using the perception error and the reference orientation angle, which serves as the eavesdropping location.

[0059] Specifically, when the sensing signal transmitted by the transmitter encounters the eavesdropping drone, it generates an echo. The receiver processes this echo signal and uses direction-of-arrival estimation algorithms (such as MUSIC, Capon, etc.) to calculate a preliminary, most probable azimuth angle of the eavesdropper. This estimate is used as a baseline or nominal azimuth for subsequent analysis, and a maximum perceived error is determined based on system characteristics or historical data. and based on the reference direction angle Construct an angularly uncertain interval centered on [the point]. This range represents the potential location of the eavesdropper. Subsequent security strategies, such as resource allocation and beamforming, will be optimized based on this uncertainty set to ensure the system remains secure even in the worst-case scenario.

[0060] This invention avoids decision-making failures caused by over-reliance on a single, potentially flawed, estimated point. It ensures that the system's security and confidentiality meet requirements throughout the entire angular uncertainty range, enhancing its robustness against interference and eavesdropping under conditions of inaccurate perception. It transforms idealized fixed-point thinking into a more practical range-based positioning approach. This makes the entire integrated sensing and communication system less vulnerable to perfect sensing results and more adaptable to a certain degree of sensing fluctuations.

[0061] Step S203: Input the system status information and the eavesdropping location into the decision model, and output the control strategy through the decision model. The system status information is the status information of the communication network, which is the communication network of the transmitting end and the receiving end. The control strategy is used to adjust the direction and energy of the composite signal so that the communication signal received by the receiving end is enhanced and other received signals are weakened, and the perceived signal received by the eavesdropper is enhanced and the received communication signal is weakened.

[0062] Specifically, the decision-making model is a deep reinforcement learning model, and step S203 above includes:

[0063] Step S2031: Input the system state information and the eavesdropping location as environmental information for reinforcement learning into the decision model, and output the communication covariance matrix and perception covariance matrix as the action space through the decision model.

[0064] Step S2032: Generate a first beamforming vector for controlling the communication signal based on the communication covariance matrix.

[0065] Step S2033: Generate a second beamforming vector for controlling the sensing signal based on the sensing covariance matrix.

[0066] The decision model in this invention uses a neural network trained through deep reinforcement learning. It learns the optimal policy mapping relationship through a large number of environmental interactions. It can be a deep reinforcement learning model such as Proximal Policy Optimization (PPO), Deep Deterministic Policy Gradient (DDPG), or Soft Actor-Critic (SAC).

[0067] In one example, deep reinforcement learning employs a proximal policy optimization algorithm. (Policy network) Modeled as a diagonal Gaussian distribution, It is a probabilistic policy represented by a neural network, with the input state s and output action a distribution, and θ being the parameters of the policy network (i.e., weights and biases). The goal of the neural network is to find the optimal θ through training, with its mean... with standard deviation Output from a shared multilayer perceptron (MLP) Let be the mean vector of the action distribution given state s. Given a state s, the standard deviation vector of the action distribution; simultaneously, an independent value network is trained. Used for Generalized Advantage Estimation (GAE). During training, the pruned probability ratio is used instead of the objective function to limit the policy update step size, ensuring the stability of the learning process and monotonic performance improvement. The policy update strategy of the PPO deep reinforcement learning algorithm is as follows: Figure 3 As shown.

[0068] The deep reinforcement learning solution framework based on near-end policy optimization transforms the complex non-convex dynamic optimization problem into a Markov decision process. This enables the system to autonomously learn the optimal beamforming strategy by interacting with the environment, provided it only acquires real-time system state information and angle estimates. It does not require prior knowledge of channel statistical characteristics or eavesdropper dynamic models. In particular, it can adapt to dynamic changes in the channel and the movement of UAVs online, making it especially suitable for complex and ever-changing integrated air-space-ground network environments. It has strong adaptability and online deployment capabilities.

[0069] Based on the aforementioned decision-making model, the transmitting end first fuses system state information, representing the real-time network status, with eavesdropping locations, representing security threats, to form environmental information within the reinforcement learning framework, which is then input into the decision-making model. This model, acting as a trained agent, performs forward reasoning through its complex internal neural network, ultimately outputting a specific decision action. The action space here is specifically defined as the communication covariance matrix and the perception covariance matrix. These two matrices are the mathematical core of the beamforming strategy. The communication covariance matrix determines the energy efficiency and directivity of the communication signal, while the perception covariance matrix simultaneously carries the perception accuracy and interference effectiveness.

[0070] In one example, based on the observed state, the system invokes a trained policy network and outputs the communication covariance matrix for each device. With the perception covariance matrix Based on this, the first beamforming vector is generated. That is, the aforementioned communication beamforming vector and the second beamforming vector. (i.e., the aforementioned sensing / artificial noise beamforming vector).

[0071] The first beamforming vector is a complex vector in which each element precisely controls the signal amplitude and phase of the corresponding antenna element. Its design goal is to enable the energy of the communication signal to be superimposed in phase in the direction of the receiver to form a high-gain main lobe.

[0072] The second beamforming vector optimization objective is not directed towards the legitimate receiver, but rather to concentrate the energy of the sensing signal onto the spatial region pointed to by the eavesdropping location, thereby integrating sensing and security jamming functions. Sensing and jamming share the same physical waveform and radiation direction at the signal level, serving two functions: First, as a sensing wave, the signal propagates in space and is scattered upon encountering objects such as eavesdroppers, forming echoes. The transmitting end can estimate the distance, speed, and direction of the eavesdropper by receiving and analyzing the known signal's delay, frequency shift, and angle of arrival (i.e., achieving the sensing function). Second, as a jamming wave, when the same signal reaches the eavesdropper, its energy overwhelms the weak communication signal the eavesdropper is trying to receive. Because this signal completely or partially overlaps with the communication signal in the frequency band, and its waveform is an unknown random or pseudo-random sequence for the eavesdropper, the eavesdropper cannot filter it out like a legitimate receiver, resulting in a sharp drop in the signal-to-interference-plus-noise ratio (SINR) and making it impossible to demodulate useful communication information (i.e., achieving the jamming function).

[0073] Step S204: Adjust the composite signal according to the control strategy to safely offload the task to the receiving end. For details, please refer to [link to relevant documentation]. Figure 1 Step S101 of the illustrated embodiment will not be described again here.

[0074] The integrated communication and sensing computational task offloading method provided in this embodiment transforms the complex beamforming strategy problem into an output covariance matrix, which is mathematically easier to process and optimize, thus improving the accuracy and efficiency of decision-making. Two independent beamforming vectors are used to independently control communication and security performance, avoiding mutual interference and enabling resource allocation. The second beamforming vector combines sensing and jamming functions, improving the utilization efficiency of spectrum and energy resources. A near-end policy optimization algorithm is used for policy learning. By modeling the policy network as a diagonal Gaussian distribution and having its parameters output by a shared multilayer perceptron, efficient exploration and precise control of the high-dimensional continuous action space are achieved. Simultaneously, combining independent value networks and generalized advantage estimation improves the accuracy and sample efficiency of policy gradient estimation. By introducing a pruning probability ratio to replace the objective function, the magnitude of policy updates is constrained, avoiding the risk of policy collapse during training. This provides a stable, reliable, and continuously self-optimizing intelligent decision-making scheme for complex and dynamic integrated air-space-ground network environments.

[0075] In one optional implementation, the system status information includes channel status information from the transmitter to the receiver, the signal-to-interference-plus-noise ratio (SIR) of the current communication link, the SIR of the current sensing link, and the instantaneous power consumption of the transmitter.

[0076] Specifically, the transmitting end performs channel estimation using pilot signals to obtain channel state information characterizing signal propagation properties; it measures the signal-to-interference-plus-noise power ratio of the communication link by demodulating the reference signal; it assesses the sensing link quality by analyzing the signal-to-noise ratio of the radar echo signal; and it monitors the power consumption of the transmission link in real time through a built-in power detection circuit. Parameters are synchronously acquired at a fixed sampling period, collectively forming a multi-dimensional observation vector reflecting the overall system state.

[0077] This invention provides an accurate environmental perception foundation for intelligent decision-making by establishing a complete state observation system that includes communication quality, sensing performance, and resource consumption. The collaborative monitoring of multi-dimensional states enables the system to comprehensively grasp the coupling relationship between channel conditions, communication quality, sensing accuracy, and energy consumption, providing sufficient decision-making basis for subsequent beamforming optimization. The integrated state perception mechanism enhances the system's adaptability to dynamic environments. By simultaneously considering communication, sensing, and energy consumption indicators in the state information, it achieves optimal resource allocation and ensures stable system operation in complex wireless environments.

[0078] In an optional implementation, when there are multiple transmitters, step S2031 includes:

[0079] Step a1: Receive the system status information and eavesdropping location sent by each sending end.

[0080] Step a2 involves analyzing all received system status information and all eavesdropping locations as environmental information to obtain a set of covariance matrices that form the action space.

[0081] Step a3: Feed back the target communication covariance matrix and the target perception covariance matrix corresponding to the target transmitter in the covariance matrix set to the target transmitter.

[0082] The decision model provided in this embodiment of the invention can be executed on any dedicated computing node such as a transmitting device with processing capabilities, a central processing unit in a system, an edge server, or a cloud platform.

[0083] First, the system receives system status information and eavesdropping location data collected and uploaded by various transmitters in the network. The system status information collected by each transmitter includes at least the local channel status information, communication link signal-to-interference-plus-noise ratio (SIR), sensing link SIR, and instantaneous power consumption of that device. The eavesdropping location information includes azimuth data estimated through sensing signals.

[0084] Next, all received system state information and all eavesdropping locations are fused to form a global environment state representation. This global environment state representation serves as input to the deep reinforcement learning decision model. After forward inference computation by the policy network, it outputs a set of covariance matrices. This set contains the communication covariance matrix and perception covariance matrix corresponding to each transmitter in the network, constituting the action space in the reinforcement learning framework.

[0085] Finally, the central processing unit extracts the target communication covariance matrix and target perception covariance matrix corresponding to each target transmitter from the output covariance matrix set, and feeds them back to the corresponding target transmitters via the communication link. After receiving their respective covariance matrices, each transmitter can then perform subsequent beamforming vector generation and signal transmission operations.

[0086] In one example, at each time slot t, the system controller (i.e., the agent built based on the decision model) acquires real-time environmental information s. t s t This is used to represent the system's state in time slot t. This state comprehensively reflects the communication, sensing, and security posture, specifically including: the uplink channel matrix from all terrestrial IoT devices to the low-Earth orbit satellite. , The uplink channel matrix from the k-th terrestrial IoT device to the low-Earth orbit satellite (with its real and imaginary parts in complex form input separately), and the azimuth angle of the eavesdropping drone estimated by each device based on integrated sensing and communication echoes. and the corresponding angular uncertainty , This represents the heading angle of the eavesdropping drone estimated by the k-th device based on the ISAC echo. This represents the estimated uncertainty of the orientation angle, as well as the current communication of each link. Perception and the instantaneous power consumption of each device, Let be the signal-to-interference-plus-noise ratio (SINR) of the k-th link. Let be the perceived signal-to-interference-plus-noise ratio (SINR) of the k-th link.

[0087] Based on observed environmental information, the agent outputs continuous actions. , The continuous action space of the agent's output in time slot t is the set of transmit covariance matrices of all devices. , Let be the transmit covariance matrix of the k-th device used to carry communication signals. Let be the emission covariance matrix of the k-th device used for sensing and detection, and artificial noise interference, where Used to carry communication signals Used for sensing and detection as well as artificial noise interference. In practical implementation, the covariance matrix can be decomposed using Cholesky decomposition (i.e., decomposed into the product of a lower triangular matrix and its transpose, thus ensuring the symmetric positive definiteness of the output matrix), and power normalization can be performed after action generation to naturally satisfy this constraint.

[0088] This invention employs a framework of centralized decision-making and distributed execution, integrating network-wide state information for collaborative decision-making. It intelligently balances communication rate, sensing accuracy, and security anti-interference capabilities. By utilizing a covariance matrix generation and power normalization mechanism based on Cholesky decomposition, it ensures the physical realizability of the beamforming matrix while satisfying transmit power constraints, thus improving the algorithm's practicality and reliability. Furthermore, for eavesdroppers not detected by the current transmitter but detected by other transmitters, information fusion allows each transmitter to exchange eavesdropper information. The resulting target communication covariance matrix and target sensing covariance matrix enable the current transmitter to accurately obtain the locations of all eavesdroppers in the environment, further improving the accuracy of eavesdropper detection.

[0089] In one optional implementation, the reward function of the decision model includes at least one of a first scoring term, a second scoring term, a third scoring term, and a fourth scoring term; the first scoring term is determined based on the lowest data transmission rate among each transmitter; the second scoring term is determined based on the signal-to-interference-plus-noise ratio (SNR) of the sensing link at each transmitter; the third scoring term is determined based on the maximum eavesdropping SNR, which is the maximum predicted SNR within the angular uncertainty interval corresponding to each transmitter, and the eavesdropping SNR is the SNR of the eavesdropping link when the eavesdropper is eavesdropping; the fourth scoring term is determined based on the instantaneous power consumption of the transmitter.

[0090] Specifically, the model is updated and converged online by continuously recording the state, action, reward, and next state of each interaction, forming an experience replay pool. Every certain time slot, the policy network is fine-tuned using experience to adapt to dynamic factors such as time-varying channels and UAV movement. This process is repeated until the policy performance is stable and the system reaches a safe, fair, and efficient operating state.

[0091] The reward function acts as a "judge" in reinforcement learning, guiding the accuracy of the action space generated by the decision model. The quality of the reward function directly impacts the capability of the decision model. This invention optimizes the reward function of the decision model using four scoring terms.

[0092] The first scoring criterion uses the communication rate (i.e., the lowest data transmission rate) of the worst-performing user in the system as the standard, aiming to ensure fairness and basic service quality across all links and avoid communication bottlenecks. In other words, it identifies the "data transmission speed" of all IoT devices and defines the score based on the slowest data transmission speed. The second scoring criterion is directly related to the signal-to-interference-plus-noise ratio (SIRR) of the sensing link, aiming to improve radar echo quality and thus ensure the accuracy of eavesdropper location estimation. The third scoring criterion focuses on system security by calculating the maximum possible SIRR for eavesdropping within the uncertain range of eavesdropper location and driving the agent to optimize beamforming to minimize the risk of information leakage in this worst-case scenario. The fourth scoring criterion constrains the instantaneous power consumption of the transmitter to promote system energy efficiency. By weighting and combining these competing objectives, the reward function can systematically guide the agent to achieve a dynamic and efficient optimal balance between communication performance, sensing accuracy, security strength, and energy consumption.

[0093] In one example, to collaboratively optimize communication fairness, perception reliability, physical layer security, and energy efficiency, the system calculates a composite reward. Among them, the main item It represents the current minimum data transmission rate, reflecting the maximum-minimum fairness criterion; the higher the minimum data transmission rate, the higher the score. This indicates a positive reward if the perceived SINR of all devices is not lower than the threshold. If so, a positive reward will be given. Otherwise, no reward will be given; As a penalty, it needs to be applied to each device within its angular uncertainty range. Search for the maximum SINR of eavesdropping within the internal search function, and construct a security penalty term using the global worst value. The higher the maximum eavesdropping SINR, the larger this value; in addition, penalties are imposed on power exceeding the limit. To encourage energy efficiency optimization, the higher the instantaneous power consumption at the transmitting end, the higher the score.

[0094] This invention, through the design of this composite reward function, transforms a complex, multi-objective engineering optimization problem into a single-objective reinforcement learning problem. The third scoring term of this reward function constrains the worst-case SINR for eavesdropping within the aforementioned angular uncertainty interval, meaning that even against an eavesdropper in the most advantageous position, the system can still effectively suppress the risk of information leakage. Figure 4 As shown, the worst-case SINR of eavesdropping converges stably to -14.5 dB, which is far below the set safety threshold of -5 dB, thus solving the unrealistic requirement in traditional methods that the location of the eavesdropper must be perfectly known.

[0095] Furthermore, the first scoring item employs a maximum-minimum fairness criterion, ensuring that every user receives a stable communication rate under high load scenarios. In particular, users in disadvantaged positions or with poor conditions can enjoy a service quality superior to fixed or random policies. For example... Figure 5 and Figure 6 The results show that the perceived signal-to-noise ratio (SINR) and minimum user rate are consistently maintained at a high level, reflecting the balanced resource allocation of this scheme.

[0096] In one optional implementation, the sum of the communication covariance matrix and the sensing covariance matrix corresponding to any transmitter satisfies a preset maximum transmit power constraint.

[0097] In one example, for any transmitter in the network, its communication covariance matrix used to carry communication signals is... The sensing covariance matrix used for sensing detection and artificial noise interference The sum of the matrix traces must be less than or equal to a preset threshold. All devices must meet the single-point maximum transmit power constraint. , This is the maximum transmission power.

[0098] This invention, through this constraint, strictly limits the agent's decision-making behavior within the hardware's limits, enabling the covariance matrix output by the algorithm to be directly converted into a practically transmittable waveform. Since the total power is a fixed and finite budget, this constraint forces the agent to intelligently allocate power between communication and sensing functions. It guides the deep reinforcement learning model to actively learn how to balance communication rate, sensing accuracy, and the effects of artificial noise interference while meeting power constraints, thereby making efficient and feasible resource management decisions.

[0099] This embodiment also provides a computing task security offloading device integrating communication and sensing, which is used to implement the above embodiments and preferred embodiments, and will not be repeated as already described. As used below, the term "module" can be a combination of software and / or hardware that implements a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0100] This embodiment provides a computing task secure offloading device that integrates communication and sensing, such as... Figure 7 As shown, it includes:

[0101] The signal synthesis and transmission module 701 is used to transmit composite signals and receive corresponding echo signals. The composite signals include communication signals and sensing signals, and the echo signals are the echo signals corresponding to the sensing signals.

[0102] The positioning module 702 is used to estimate the eavesdropper's location based on the echo signal. The eavesdropping location includes height information and direction information.

[0103] The strategy decision module 703 is used to input system status information and eavesdropping location into the decision model, and output control strategy through the decision model. The system status information is the status information of the communication network, which is the communication network of the transmitting end and the receiving end. The control strategy is used to adjust the direction and energy of the composite signal so that the communication signal received by the receiving end is enhanced and other received signals are weakened, and the perceived signal received by the eavesdropper is enhanced and the received communication signal is weakened.

[0104] The composite signal adjustment and task offloading module 704 is used to adjust the composite signal according to the control strategy to safely offload the computing task to the receiving end.

[0105] This embodiment provides an integrated communication and sensing computing task offloading device. Through the coordinated operation of its modules, it achieves secure, efficient, and integrated task offloading. The sensing and positioning modules enable the system to proactively detect and locate potential eavesdropping threats. The strategy decision-making module makes intelligent decisions based on real-time environmental information, and the signal synthesis and transmission module precisely executes the strategy. The device achieves unified optimization through the strategy decision-making module, generating a coordinated control strategy. The signal synthesis and transmission module generates a single composite signal to simultaneously achieve both functions. This unifies communication and security protection at the signal level, avoiding resource competition and efficiency losses caused by functional separation. The device uses beamforming to precisely control the spatial distribution of signal energy, accurately focusing communication signal energy onto the legitimate receiving end while precisely directing the energy of sensing / interference signals to the eavesdropping location. This capability allows the system to improve the performance of the target link while proactively suppressing eavesdropping links, enhancing physical layer security.

[0106] The integrated communication and sensing computing task secure offloading device provided in this embodiment of the invention can execute the integrated communication and sensing computing task secure offloading method provided in any embodiment of the invention, and has the corresponding functional modules and beneficial effects for executing the method. Further functional descriptions of the above modules and units are the same as in the corresponding embodiments described above, and will not be repeated here.

[0107] Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0108] The following is a detailed reference. Figure 8 This diagram illustrates a suitable structural schematic for implementing an electronic device according to embodiments of the present invention. The electronic device may include a processor (e.g., a central processing unit, graphics processor, etc.) 801, which can perform various appropriate actions and processes based on a program stored in read-only memory (ROM) 802 or a program loaded from memory 808 into random access memory (RAM) 803. The RAM 803 also stores various programs and data required for the operation of the electronic device. The processor 801, ROM 802, and RAM 803 are interconnected via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0109] Typically, the following devices can be connected to I / O interface 805: input devices 806 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 807 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; memory devices 808 including, for example, magnetic tapes, hard disks, etc.; and communication devices 808. Communication device 808 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 8Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown, and more or fewer devices may be implemented or have instead.

[0110] In particular, according to embodiments of the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a non-transitory computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 808, or installed from a memory 808, or installed from a ROM 802. When the computer program is executed by the processor 801, it performs the functions defined in the integrated communication and sensing computing task secure offloading method of the embodiments of the present invention.

[0111] Figure 8 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of the present invention.

[0112] This invention also provides a computer-readable storage medium. The methods described above according to embodiments of the invention can be implemented in hardware or firmware, or implemented as computer code that can be recorded on a storage medium, or implemented as computer code downloaded over a network and originally stored on a remote storage medium or a non-transitory machine-readable storage medium and then stored on a local storage medium. Thus, the methods described herein can be processed by software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. The storage medium can be a magnetic disk, optical disk, read-only memory, random access memory, flash memory, hard disk, or solid-state drive, etc.; further, the storage medium can also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code. When the software or computer code is accessed and executed by the computer, processor, or hardware, the integrated communication and sensing computing task secure offloading method shown in the above embodiments is implemented.

[0113] A portion of this invention can be applied as a computer program product, such as computer program instructions, which, when executed by a computer, can invoke or provide the methods and / or technical solutions according to the invention through the operation of the computer. Those skilled in the art will understand that the forms in which computer program instructions exist in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways in which computer program instructions are executed by a computer include, but are not limited to: the computer directly executing the instructions, or the computer compiling the instructions and then executing the corresponding compiled program, or the computer reading and executing the instructions, or the computer reading and installing the instructions and then executing the corresponding installed program. Here, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible to a computer.

[0114] Although embodiments of the invention have been described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for securely offloading computational tasks integrating communication and sensing, characterized in that, Applied to the sending end, the method includes: Transmit a composite signal and receive the corresponding echo signal. The composite signal includes a communication signal and a sensing signal, and the echo signal is the echo signal corresponding to the sensing signal. The eavesdropper's location is estimated based on the echo signal, and the location includes height and direction information. The system status information and the eavesdropping location are input into the decision model, and the decision model outputs a control strategy. The system status information is the status information of the communication network, which is the communication network of the transmitting end and the receiving end. The control strategy is used to adjust the direction and energy of the composite signal so that the communication signal received by the receiving end is enhanced and other received signals are weakened, and the perception signal received by the eavesdropper is enhanced and the received communication signal is weakened. The composite signal is adjusted according to the control strategy to safely offload the task to the receiving end.

2. The method according to claim 1, characterized in that, The system status information includes the channel status information from the transmitter to the receiver, the signal-to-interference-plus-noise ratio (SIR) of the current communication link, the SIR of the current sensing link, and the instantaneous power consumption of the transmitter.

3. The method according to claim 1, characterized in that, The decision model is a deep reinforcement learning model. The process of inputting system state information and the eavesdropping location into the decision model, and then outputting a control strategy through the decision model, includes: The system state information and the eavesdropping location are used as environmental information for reinforcement learning and input into the decision model. The decision model outputs the communication covariance matrix and the perception covariance matrix as the action space. A first beamforming vector for controlling the communication signal is generated based on the communication covariance matrix; A second beamforming vector for controlling the sensing signal is generated based on the sensing covariance matrix.

4. The method according to claim 3, characterized in that, When there are multiple transmitters, the step of the decision model outputting the communication covariance matrix and the perception covariance matrix based on the system state information and the eavesdropping location includes: Receive system status information and eavesdropping locations sent by each sending end; All received system status information and all eavesdropping locations are analyzed as environmental information to obtain a set of covariance matrices that serve as the action space. The target communication covariance matrix and the target perception covariance matrix corresponding to the target transmitter in the covariance matrix set are fed back to the target transmitter.

5. The method according to claim 4, characterized in that, The step of estimating the eavesdropper's location based on the echo signal includes: The reference orientation angle and altitude information of the eavesdropper are estimated based on the echo signal; An angular uncertainty interval is modeled by sensing error and the reference orientation angle, and the angular uncertainty interval is used as the direction information of the eavesdropper.

6. The method according to claim 5, characterized in that, The reward function of the decision model includes at least one of a first scoring term, a second scoring term, a third scoring term, and a fourth scoring term; the first scoring term is determined based on the lowest data transmission rate among each transmitter; the second scoring term is determined based on the signal-to-interference-plus-noise ratio (SNR) of the sensing link at each transmitter; the third scoring term is determined based on the maximum eavesdropping SNR, which is the maximum predicted SNR within the angular uncertainty interval corresponding to each transmitter, and the eavesdropping SNR is the SNR of the eavesdropping link when the eavesdropper is eavesdropping; the fourth scoring term is determined based on the instantaneous power consumption of the transmitter.

7. The method according to claim 4, characterized in that, The sum of the communication covariance matrix and the sensing covariance matrix corresponding to any transmitting end satisfies the preset maximum transmit power constraint.

8. A computing task secure offloading device integrating communication and sensing, characterized in that... The transmitting end, the device includes: A signal synthesis and transmission module is used to transmit a composite signal and receive a corresponding echo signal. The composite signal includes a communication signal and a sensing signal, and the echo signal is the echo signal corresponding to the sensing signal. The positioning module is used to estimate the eavesdropper's location based on the echo signal, the location including height and direction information; The strategy decision module is used to input system status information and the eavesdropping location into the decision model, and output a control strategy through the decision model. The system status information is the status information of the communication network, which is the communication network of the transmitting end and the receiving end. The control strategy is used to adjust the direction and energy of the composite signal so that the communication signal received by the receiving end is enhanced and other received signals are weakened, and the perception signal received by the eavesdropper is enhanced and the received communication signal is weakened. The composite signal adjustment and task offloading module is used to adjust the composite signal according to the control strategy in order to safely offload the task to the receiving end.

9. An electronic device, characterized in that, include: A memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, the processor executing the computer instructions to perform the method of any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions for causing the computer to perform the method of any one of claims 1 to 7.

Images