An anti-active eavesdropping unmanned aerial vehicle assisted multi-cluster air computing method and device

Abstract

The application discloses an unmanned aerial vehicle auxiliary multi-cluster air computing method and device resisting active eavesdropping; in the unmanned aerial vehicle auxiliary air computing system of multiple active eavesdroppers, the application divides the sensors based on the branch and bound framework, outputs an optimal cell cluster strategy, and realizes effective and safe spatial isolation of the sensors and the eavesdroppers; the joint optimization problem is established under the conditions of meeting the safety constraints of the eavesdroppers and the maximum transmission power constraints of the sensors, and the target is to minimize the average mean square error between the received signals and the aggregated signals of the unmanned aerial vehicles; the joint cell cluster strategy is combined, the joint optimization problem is solved through an alternating optimization algorithm, the scheduling coefficient, the sensor transmission coefficient, the unmanned aerial vehicle noise reduction factor and the unmanned aerial vehicle hovering position are optimized, and the calculation precision is improved; the application significantly improves the calculation precision while ensuring safety, and provides an effective solution for a dense Internet of Things environment with eavesdropping risks.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, and in particular to a method and apparatus for anti-active eavesdropping unmanned aerial vehicle-assisted multi-cluster aerial computing. Background Technology

[0002] Over-the-air computing is an emerging paradigm of communication-computing convergence. Its core idea is to leverage the inherent broadcast and signal superposition characteristics of wireless channels to directly perform function calculations on data during signal transmission. Unlike the traditional communication system's separate processing approach of independent transmission followed by centralized computation, over-the-air computing allows all devices to share the same time-frequency resources for simultaneous transmission. Through carefully designed pre-processing and post-processing, the desired aggregation function value is directly obtained at the fusion center. This architecture is particularly suitable for distributed data aggregation tasks in scenarios such as the Internet of Things (IoT) and edge computing, including distributed machine learning model updates, environmental monitoring data statistics, and distributed sensor network data fusion. Compared to traditional orthogonal multiple access (OMA) technologies, over-the-air computing reduces the communication load from being proportional to the number of devices to a constant, significantly improving spectral efficiency and reducing communication latency, providing a new technological path for efficient collaboration of large-scale IoT devices.

[0003] With the rapid development of drone technology, aerial computing systems using drones as mobile convergence hubs have attracted widespread attention. Drones, with their high mobility and flexible deployment capabilities, can dynamically adjust their spatial position to optimize wireless channel quality with ground equipment. Especially in urban environments or complex terrains, drones can establish line-of-sight transmission links, effectively avoiding the impact of obstacles on communication quality. Furthermore, the three-dimensional mobility of drones provides a new dimension for optimizing aerial computing systems. Through reasonable trajectory design, areas with poor channel conditions can be actively avoided, and channel differences between different devices can be balanced, thereby mitigating the bottleneck effect of traditional ground-based aerial computing systems. Compared with traditional fixed base stations, drone-assisted aerial computing systems show significant improvements in both computing accuracy and coverage.

[0004] However, the open nature of wireless channels presents severe security challenges to in-flight computing systems. Since all devices share the same transmission resources, transmission signals naturally superimpose in the air, allowing any eavesdropper within range to receive the aggregated signal, potentially leading to the leakage of sensitive data. Physical layer security, a lightweight security method that does not rely on encryption algorithms, leverages the physical characteristics of wireless channels to ensure information security. Its core idea is to improve the quality of legitimate channels while deteriorating the transmission conditions of eavesdropping channels through reasonable resource allocation and signal design. In in-flight computing scenarios, physical layer security can securely transform the objective into ensuring the computational accuracy of legitimate fusion centers while preventing eavesdroppers from accurately reconstructing the aggregation function value from the received signal. However, effective solutions are still lacking when facing coordinated attacks from multiple active eavesdroppers, particularly in areas such as multi-slot joint optimization, security-aware partitioning, and dynamic trajectory design, where it is often difficult to balance computational accuracy and security performance. Summary of the Invention

[0005] This invention provides a method and apparatus for anti-active eavesdropping drone-assisted multi-cluster aerial computing, which solves the technical problem that existing aerial computing systems struggle to balance computational accuracy and security performance in scenarios facing multiple active eavesdroppers.

[0006] This invention provides a method for anti-active eavesdropping UAV-assisted multi-cluster aerial computing, the method comprising:

[0007] Acquire system information from an aerial computing system consisting of drones, multiple sensors, and multiple eavesdroppers;

[0008] Based on the channel information in the system information, establish channel gain models between the UAV and the sensor, between the eavesdropper and the sensor, between the UAV and the eavesdropper, and between each of the eavesdroppers;

[0009] A branch-and-bound framework is used to divide sensor cell clusters based on the system information to obtain a cell clustering strategy;

[0010] Based on the cell clustering strategy and all the channel gain models, under the conditions of satisfying the eavesdropper security constraints and the sensor maximum transmission power constraints, a joint optimization problem is established with the objective of minimizing the average mean square error between the received signal and the aggregated signal of the UAV.

[0011] The joint optimization problem is solved iteratively using an alternating optimization algorithm, which outputs the optimal scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position.

[0012] Optionally, the step of using a branch-and-bound framework to partition sensor cell clusters based on the system information to obtain a cell cluster strategy includes:

[0013] Using a branch-and-bound framework, based on the system information, a cell cluster model is established with the objective of minimizing the maximum coverage radius of all cell clusters and the constraint that the distance between the center of each cell cluster and the location of all eavesdroppers is greater than a preset security radius threshold.

[0014] By constructing a search tree to traverse all candidate partitions in the cell cluster model, the candidate partitions are optimized based on a depth-first search strategy and pruning techniques.

[0015] The non-convex optimization problem of the optimized candidate partition is solved by the minimum enclosing circle process, and the cell cluster strategy is output. The cell cluster strategy includes the cell cluster to which each sensor belongs, the coverage radius of each cell cluster, and the center position of each cell cluster.

[0016] Optionally, the step of iteratively solving the joint optimization problem using an alternating optimization algorithm to output the optimal scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position includes:

[0017] Initialize the optimization variables for the joint optimization problem; the optimization variables include initial scheduling coefficients, sensor transmission coefficients, UAV noise reduction factor, and UAV hovering position; wherein, under the initial scheduling coefficients, the initial UAV hovering position is determined by the center position in the cell cluster strategy;

[0018] The mean square error between the initial received signal and the aggregated signal of the UAV is calculated based on the initial scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position. The parameters of the joint optimization problem are optimized by an alternating optimization algorithm, and the updated scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position are obtained by alternating optimization.

[0019] The mean square error between the received signal and the aggregated signal of the updated drone is calculated based on the updated scheduling coefficient, sensor transmission coefficient, drone noise reduction factor, and drone hovering position.

[0020] If the absolute value of the difference between the average mean square error between the updated UAV's received signal and the aggregated signal and the initial average mean square error between the UAV's received signal and the aggregated signal is greater than a preset convergence threshold, then the updated scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position are used as the initial scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position, and the process jumps to execute the step of calculating the initial average mean square error between the UAV's received signal and the aggregated signal based on the initial scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position.

[0021] If the absolute value of the difference between the mean square error between the received signal and the aggregated signal of the updated UAV and the mean square error between the received signal and the aggregated signal of the initial UAV is less than or equal to the preset convergence threshold, then the updated scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position are used as the optimal scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position and output.

[0022] Optionally, the step of using an alternating optimization algorithm to optimize the parameters of the joint optimization problem, and obtaining updated scheduling coefficients, sensor transmission coefficients, UAV noise reduction factors, and UAV hovering positions through alternating optimization, includes:

[0023] Under the alternating optimization algorithm, the joint optimization problem is transformed into a scheduling coefficient subproblem. Based on the initial sensor transmission coefficients, UAV noise reduction factor and UAV hovering position, the scheduling coefficient subproblem is exhaustively searched to obtain the updated scheduling coefficients.

[0024] Based on the calculation formula of the average mean square error between the received signal and the aggregated signal of the UAV, the updated UAV noise reduction factor is solved according to the updated scheduling coefficient, the initial sensor transmission coefficient and the UAV hovering position.

[0025] The joint optimization problem is transformed into a sensor transmission coefficient subproblem. The sensor transmission coefficient subproblem is solved using a continuous convex approximation method based on the updated scheduling coefficients, the UAV noise reduction factor, and the initial UAV hovering position, to obtain the updated sensor transmission coefficients.

[0026] The joint optimization problem is transformed into a UAV hovering position subproblem. The gradient descent algorithm is used to solve the UAV hovering position subproblem based on the updated scheduling coefficients, UAV noise reduction factor and sensor transmission coefficient to obtain the updated UAV hovering position.

[0027] Optionally, the joint optimization problem P0 is represented as follows:

[0028] ;

[0029] In the formula: Let n be the hovering position of the UAV in time slot n. Let m be the set of sensors for cell cluster m. Let m be the coverage radius of the cell cluster. Let n be the sensor transmission coefficient in time slot n. Let be the noise reduction factor of the UAV in time slot n. Let m be the scheduling coefficient of cell cluster m in time slot n. Let be the average mean square error between the received signal and the aggregated signal of the UAV in time slot n; Let n be the noise reduction factor for the eavesdropper in time slot n. Let n be the mean square error between the eavesdropper's received signal and the aggregated signal in time slot n. To preset a safety threshold, This represents the sensor's maximum transmission power. For the first The fixed horizontal coordinates of each sensor; M is the total number of cells in the cluster.

[0030] Optionally, the mean square error between the received signal and the aggregated signal of the UAV is calculated as follows:

[0031] ;

[0032] In the formula: For the UAV's received signal in time slot n, This represents the aggregated signal obtained by the UAV at cell cluster m in time slot n; To represent the channel gain between the UAV and the k-th sensor in time slot n, The eavesdropper's transmission power, Let the channel gain between the UAV and the vth eavesdropper be in time slot n. Let n be the additive white Gaussian noise power at the UAV in time slot n;

[0033] The average mean square error between the eavesdropper's received signal and the aggregated signal is calculated as follows:

[0034] ;

[0035] In the formula: The signal received by the eavesdropper in time slot n; Let v be the channel gain between the v-th eavesdropper and the k-th sensor in time slot n. For the vth eavesdropper and the th eavesdropper in time slot n Channel gain between eavesdroppers For the self-interference channel gain of the eavesdropper in time slot n, Let be the additive white Gaussian noise power at the eavesdropper's location in time slot n.

[0036] This invention also provides a drone-assisted multi-cluster aerial computing device resistant to active eavesdropping, the device comprising:

[0037] The information acquisition module is used to acquire system information from an airborne computing system consisting of a drone, multiple sensors, and multiple eavesdroppers.

[0038] The channel gain model establishment module is used to establish channel gain models between the UAV and the sensor, the eavesdropper and the sensor, the UAV and the eavesdropper, and each of the eavesdroppers based on the channel information in the system information;

[0039] The cell cluster strategy generation module is used to divide sensor cell clusters according to the system information using a branch and bound framework to obtain cell cluster strategies.

[0040] The joint optimization problem establishment module is used to establish a joint optimization problem with the objective of minimizing the average mean square error between the received signal and the aggregated signal of the UAV, based on the cell clustering strategy and all the channel gain models, under the conditions of satisfying the eavesdropper security constraint and the sensor maximum transmission power constraint.

[0041] The optimal strategy output module is used to iteratively solve the joint optimization problem using an alternating optimization algorithm, and output the optimal scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position.

[0042] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the UAV-assisted multi-cluster aerial computing method as described above.

[0043] This invention also provides a computer-readable storage medium storing a computer program or instructions thereon, which, when executed by a processor, implements the steps of the UAV-assisted multi-cluster aerial computing method described above.

[0044] This invention also provides a computer program product, including a computer program or instructions, which, when executed by a processor, implement the steps of any of the above-described UAV-assisted multi-cluster aerial computing methods.

[0045] As can be seen from the above technical solutions, the present invention has the following advantages:

[0046] This invention provides a method and apparatus for anti-active eavesdropping UAV-assisted multi-cluster aerial computing. The method includes: acquiring system information of an aerial computing system composed of a UAV, multiple sensors, and multiple eavesdroppers; establishing channel gain models between the UAV and sensors, between eavesdroppers and sensors, between the UAV and eavesdroppers, and among the eavesdroppers based on channel information in the system information; using a branch-and-bound framework to divide the sensor cell clusters according to the system information to obtain a cell clustering strategy; based on the cell clustering strategy and all channel gain models, establishing a joint optimization problem with the objective of minimizing the average mean square error between the UAV's received signal and the aggregated signal, under the conditions of satisfying eavesdropper security constraints and sensor maximum transmission power constraints; and using an alternating optimization algorithm to iteratively solve the joint optimization problem, outputting the optimal scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position.

[0047] In a drone-assisted aerial computing system with multiple active eavesdroppers, this invention partitions sensors based on a branch-and-bound framework, outputting an optimal cell clustering strategy to achieve effective secure spatial isolation between sensors and eavesdroppers. Under the conditions of satisfying eavesdropper security constraints and sensor maximum transmission power constraints, a joint optimization problem is established with the objective of minimizing the average mean square error between the drone's received signal and the aggregated signal. The joint cell clustering strategy is used to solve the joint optimization problem through an alternating optimization algorithm to optimize scheduling coefficients, sensor transmission coefficients, drone noise reduction factors, and drone hovering positions, thereby improving computational accuracy. This invention significantly improves computational accuracy while ensuring security, providing an effective solution for dense IoT environments with eavesdropping risks. Attached Figure Description

[0048] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 A flowchart illustrating the steps of an anti-active eavesdropping UAV-assisted multi-cluster aerial computing method provided in an embodiment of the present invention;

[0050] Figure 2 A model diagram of an unmanned aerial vehicle (UAV) assisted multi-cluster aerial computing system provided in an optional embodiment of the present invention;

[0051] Figure 3 The partitioning optimization diagram is provided for the baseline scheme based on rectangular partitioning in the embodiments of the present invention;

[0052] Figure 4 The partitioning optimization diagram of the baseline scheme based on the K-means algorithm provided in the embodiments of the present invention;

[0053] Figure 5 A partitioning optimization diagram of the benchmark scheme for a constant UAV position provided in an embodiment of the present invention;

[0054] Figure 6 This is a partitioning optimization diagram based on a branch-and-bound framework provided in an embodiment of the present invention;

[0055] Figure 7 The average values ​​under different FCs provided in the embodiments of the present invention With maximum transmission power Relationship diagram;

[0056] Figure 8 This is a structural block diagram of a drone-assisted multi-cluster aerial computing device that resists active eavesdropping, provided as an embodiment of the present invention. Detailed Implementation

[0057] This invention provides a method and apparatus for anti-active eavesdropping drone-assisted multi-cluster aerial computing, which solves the technical problem that existing aerial computing systems struggle to balance computational accuracy and security performance in scenarios facing multiple active eavesdroppers.

[0058] To make the objectives, features, and advantages of this invention more apparent and understandable, the technical solutions of the embodiments of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described below are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0059] It should be noted that, in the optional embodiments of the present invention, the data related to object information, etc., requires the permission or consent of the object when the embodiments of the present invention are applied to specific products or technologies. Furthermore, the collection, use, and processing of the relevant data must comply with the relevant laws, regulations, and standards of the relevant countries and regions. In other words, if the embodiments of the present invention involve data related to an object, it needs to be obtained with the object's authorization and consent, the authorization and consent of relevant departments, and in accordance with the relevant laws, regulations, and standards of the country and region. If the embodiments involve personal information, the acquisition of all personal information requires the individual's consent. If sensitive information is involved, the separate consent of the information subject is required. The embodiments also need to be implemented with the object's authorization and consent.

[0060] Please see Figure 1This invention provides a method for anti-active eavesdropping UAV-assisted multi-cluster aerial computing, the method comprising:

[0061] Step 101: Obtain system information of an airborne computing system consisting of a drone, multiple sensors, and multiple eavesdroppers;

[0062] Please see Figure 1 This method is applied to a UAV-assisted multi-cluster aerial computing system, which includes a UAV, multiple sensors, and multiple eavesdroppers (EVEs); wherein the UAV acts as the aggregation center to achieve aggregation over a sufficiently long given duration. Internal polymerization originating from the ground Data from a single-antenna sensor, A single-antenna active eavesdropper attempts to interfere with and intercept aggregated messages.

[0063] Understandably, the system information includes the fixed location information of the sensors, the fixed location information of the eavesdropper, the flight altitude of the drone and the flight altitude of the eavesdropper, as well as the preset safety radius threshold and the preset number of cell clusters. The preset safety radius threshold is used to limit the coverage radius of subsequent cell cluster divisions.

[0064] Specifically, assuming all nodes (including drones, sensors, and eavesdroppers) are deployed in a three-dimensional Cartesian coordinate system, sensors are randomly distributed on the ground, and the horizontal position coordinates of the drones and eavesdroppers are fixed, with their flight altitudes being respectively... and ;No. The fixed horizontal position coordinates of the sensors are , No. The horizontal coordinates of the eavesdropper's projection on the ground are: .

[0065] Since the mission duration T is long enough, the flight time of the UAV from one location to another can be omitted. Therefore, we only need to consider optimizing the positioning or hovering position. This means optimizing the UAV within the time slot. The horizontal position coordinates at point are represented as .

[0066] Step 102: Based on the channel information in the system information, establish channel gain models between the UAV and the sensor, between the eavesdropper and the sensor, between the UAV and the eavesdropper, and between each eavesdropper.

[0067] It should be noted that the channel information in the system information includes fixed location information, the fixed location information of the eavesdropper, the flight altitude of the drone, and the flight altitude of the eavesdropper;

[0068] Channel gain between the drone and the k-th sensor It can be modeled as:

[0069] (1)

[0070] In the formula: This represents the channel gain at a reference distance of 1 meter. Indicates in The channel phase shift between the UAV and the k-th sensor at the time slot;

[0071] No. Channel gain between the eavesdropper and the k-th sensor It can be modeled as:

[0072] (2)

[0073] In the formula: In order to be in Time slot The channel phase shift between the eavesdropper and the k-th sensor.

[0074] Channel gain between the drone and the vth eavesdropper It can be modeled as:

[0075] (3)

[0076] In the formula: In order to be in The channel phase shifts between the drone and the vth eavesdropper at the time slot.

[0077] The vth eavesdropper and the... Channel gain between eavesdroppers It can be modeled as:

[0078] (4)

[0079] In the formula: In order to be in The vth eavesdropper and the 1st eavesdropper at the time slot The channel phase shift of the eavesdropper.

[0080] Step 103: Using a branch-and-bound framework, sensor cell clusters are divided based on system information to obtain a cell cluster strategy.

[0081] It is worth noting that in an environment with multiple sensors randomly distributed and multiple eavesdroppers, there may be situations where the eavesdroppers are very close to the sensors. If the data is not divided into cell clusters, on the one hand, the drone will be affected by huge interference signals when receiving sensors that are very close to the eavesdroppers, and on the other hand, it will be difficult to meet the security constraints and achieve secure transmission.

[0082] This embodiment uses a branch-and-bound framework to divide the sensor cell cluster based on system information, that is... Each sensor is divided into A coverage radius of The cluster of cells solves the interference problem in dense device environments through spatial isolation and resource optimization, thereby improving computing efficiency and security.

[0083] More specifically, each cell cluster contains a set of sensors: And each sensor belongs to only one cell cluster: ; A cluster of cells can cover all sensors, and each cell must contain one sensor: .

[0084] In one specific implementation, step 103 may include the following steps:

[0085] S11. Using a branch and bound framework, based on system information, establish a cell cluster model with the objective of minimizing the maximum coverage radius of all cell clusters and the constraint that the distance between the center of each cell cluster and the location of all eavesdroppers is greater than a preset security radius threshold.

[0086] S12. By constructing a search tree, all candidate partitions in the cell cluster model are traversed, and the candidate partitions are optimized based on the depth-first search strategy and pruning techniques.

[0087] S13. Solve the non-convex optimization problem of the optimized candidate partition using the minimum enclosing circle process, and output the cell cluster strategy. The cell cluster strategy includes the cell cluster to which each sensor belongs, the coverage radius of each cell cluster, and the center position of each cell cluster (VBS Center).

[0088] This specific embodiment employs a branch-and-bound framework to solve a high-level N-center problem, aiming to achieve optimal sensor partitioning and UAV hovering position deployment. This branch-and-bound framework seeks... Given a cluster of coverage disks (i.e., cell clusters), and considering the condition that the distance between the center of each cell cluster and the positions of all eavesdroppers is greater than a preset security radius threshold, the maximum coverage radius of all disks is minimized. The mathematical model of the cell cluster model can then be expressed as follows:

[0089] (5)

[0090] In the formula: and They represent the first The center position and coverage radius of each disk; the constraint of equation (5) ensures that the center of each disk maintains at least a preset safety radius threshold from all eavesdroppers. The distance is such that all sensors are completely covered and do not overlap.

[0091] To solve the above cell cluster model, a branch-and-bound framework is used to construct a search tree to traverse possible partitioning schemes for the cell cluster model. For each candidate partition, the advanced minimum bounding circle procedure is invoked to solve the non-convex optimization problem, i.e.:

[0092] (6)

[0093] The advanced minimum enclosing circle process employs a sequential convex programming approach, linearizing non-convex constraints through a first-order Taylor expansion and iteratively solving the convex approximation problem until convergence, outputting a cell clustering strategy. During the solution process, a depth-first search strategy and pruning techniques are used to significantly reduce computational complexity, achieving efficient partitioning while ensuring optimal solution.

[0094] In scenarios with dense edge devices (i.e., sensors) and multiple active eavesdroppers, this embodiment treats the partitioning of edge devices as an advanced N-centered problem and proposes a solution method based on a branch and bound framework. This method involves constructing a search tree to traverse possible partitioning schemes. For each candidate partition, an advanced minimum enclosing circle process is invoked to solve the non-convex optimization problem.

[0095] Step 104: Based on the cell cluster strategy and all channel gain models, and under the conditions of satisfying the eavesdropper security constraints and the maximum transmission power constraints of the sensors, establish a joint optimization problem with the objective of minimizing the average mean square error between the UAV's received signal and the aggregated signal.

[0096] In this embodiment, the sensors are partitioned according to the cell cluster strategy, and the sensors within each cell cluster are aggregated. Data transmitted from ground sensors within a time slot; in aerial computing missions, assuming the sensors Send preprocessing signal Then the drone in the 1st Aggregated signals obtained from individual cell clusters It can be represented as:

[0097] (7)

[0098] Among them, the data from each sensor To ensure that subsequent aerial calculations are not affected by the statistical characteristics of the data itself and to meet transmission power constraints, the data from each sensor needs to be normalized to conform to zero mean and unit variance.

[0099] At the same time, the drone will receive from Interference noise emitted by an eavesdropper and additive white Gaussian noise Therefore, the drone in the Signal received in each time slot Represented as:

[0100] (8)

[0101] In the formula: Let m be the scheduling coefficient of cell cluster m at time slot n. and , Let n be the sensor transmission coefficient in time slot n. The transmission power of the eavesdropper.

[0102] set up This indicates the sensor's maximum transmission power. In the The transmission power constraint for each time slot is expressed as:

[0103] (9)

[0104] Because eavesdroppers operate in full-duplex mode, the data they receive is interfered with by the data they themselves send. We use... To represent the self-interference channel, therefore, the first The eavesdropper was in the first Signal received in each time slot for:

[0105] (10)

[0106] in, The additive white Gaussian noise at the location of the eavesdropper.

[0107] After processing the received signals using a noise reduction factor at the drone and the eavesdropper, the estimated received signal representation is as follows:

[0108] (11)

[0109] (12)

[0110] In the formula: For the UAV's received signal in time slot n, Let n be the noise reduction factor for the UAV in time slot n; For the signal received by eavesdropper v in time slot n, Let v be the noise reduction factor for the eavesdropper in time slot n.

[0111] To evaluate the performance of in-flight computing, a commonly used performance metric is the root mean square error (MSE). Furthermore, to maximize the received signal power at the UAV location, the sensor... The signal phase is set to This ensures phase alignment of the signal at the drone end. Meanwhile, we employ a worst-case assumption that the signal phase at the eavesdropper's location is also aligned.

[0112] Received signals calculated in the air from drones and eavesdroppers and The aggregated signals sent by the sensors respectively Root mean square error (RMSE) calculation is performed to establish the average MMS error between the UAV's received signal and the aggregated signal. The calculation method and the mean square error between the eavesdropper's received signal and the aggregated signal. The calculation method is as follows:

[0113] (13)

[0114] as well as,

[0115] (14)

[0116] Our goal is to optimize cell cluster strategies and scheduling coefficients throughout the entire mission. Noise reduction factor of drones Sensor transmission coefficient and the drone hovering position In the presence of eavesdropper noise interference, the joint optimization problem P0 is established to minimize the mean square error between the received signal and the aggregated signal at the UAV, while satisfying the eavesdropper security constraints and the sensor maximum transmission power constraints, and is expressed as follows:

[0117] (15)

[0118] In the formula: The preset security threshold represents the minimum acceptable mean square error at each time slot eavesdropper.

[0119] Step 105: The joint optimization problem is solved iteratively using an alternating optimization algorithm, and the optimal scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position are output.

[0120] This embodiment uses an alternating optimization algorithm to iteratively solve the joint optimization problem, with scheduling coefficients... Drone noise reduction factor Sensor transmission coefficient and drone hovering position To optimize variables through iterative optimization, the system outputs the optimal scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position. Under the security constraint of ensuring that the computing performance of eavesdroppers deteriorates, the system significantly improves the computing accuracy of legitimate receivers and effectively addresses complex threat scenarios involving coordinated attacks by multiple eavesdroppers.

[0121] In one specific implementation, step 105 may include the following steps:

[0122] S21. Initialize the optimization variables for the joint optimization problem; the optimization variables include the initial scheduling coefficients, sensor transmission coefficients, UAV noise reduction factor, and UAV hovering position; wherein, under the initial scheduling coefficients, the initial UAV hovering position is determined by the center position in the cell cluster strategy;

[0123] S22. Calculate the average mean square error between the initial received signal and the aggregated signal of the UAV based on the initial scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor and UAV hovering position; use the alternating optimization algorithm to optimize the parameters of the joint optimization problem, and obtain the updated scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor and UAV hovering position through alternating optimization.

[0124] S23. Calculate the average mean square error between the received signal and the aggregated signal of the updated UAV based on the updated scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor and UAV hovering position.

[0125] S24. If the absolute value of the difference between the average mean square error between the updated UAV's received signal and the aggregated signal and the initial average mean square error between the UAV's received signal and the aggregated signal is greater than a preset convergence threshold, then the updated scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position are used as the initial scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position, and the process jumps to the step of calculating the initial average mean square error between the UAV's received signal and the aggregated signal based on the initial scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position.

[0126] S25. If the absolute value of the difference between the average mean square error between the updated UAV's received signal and the aggregated signal and the initial average mean square error between the UAV's received signal and the aggregated signal is less than or equal to the preset convergence threshold, then the updated scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position are used as the optimal scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position and output.

[0127] Among them, the alternating optimization algorithm has good convergence and practicality, providing a reliable solution for security-sensitive IoT data aggregation applications. Furthermore, to prevent the algorithm from getting stuck in multiple loops due to failure to meet convergence conditions, thus wasting computational resources, in practical applications, if the number of iterations reaches a preset maximum number of iterations... Then, the currently updated scheduling coefficients, sensor transmission coefficients, UAV noise reduction factor, and UAV hovering position are used as the optimal scheduling coefficients, sensor transmission coefficients, UAV noise reduction factor, and UAV hovering position and output.

[0128] In one specific implementation, the alternating optimization algorithm is used to optimize the parameters of the joint optimization problem. The process of obtaining the updated scheduling coefficients, sensor transmission coefficients, UAV noise reduction factor, and UAV hovering position through alternating optimization includes the following steps:

[0129] S31. Under the alternating optimization algorithm, the joint optimization problem is transformed into a scheduling coefficient subproblem. Based on the initial sensor transmission coefficient, UAV noise reduction factor and UAV hovering position, the scheduling coefficient subproblem is exhaustively searched to obtain the updated scheduling coefficient.

[0130] S32. Based on the calculation formula of the average mean square error between the received signal and the aggregated signal of the UAV, the updated UAV noise reduction factor is solved according to the updated scheduling coefficient, the initial sensor transmission coefficient and the UAV hovering position.

[0131] S33. The joint optimization problem is transformed into a sensor transmission coefficient subproblem. The sensor transmission coefficient subproblem is solved by using the continuous convex approximation method based on the updated scheduling coefficients, the UAV noise reduction factor and the initial UAV hovering position, to obtain the updated sensor transmission coefficients.

[0132] S34. The joint optimization problem is transformed into a UAV hovering position subproblem. The gradient descent algorithm is used to solve the UAV hovering position subproblem based on the updated scheduling coefficient, UAV noise reduction factor and sensor transmission coefficient to obtain the updated UAV hovering position.

[0133] In step S31 above, the joint optimization problem is transformed into a scheduling coefficient subproblem P3, that is:

[0134] (16)

[0135] In the With a fixed initial sensor transmission coefficient in each time slot Drone noise reduction factor and the drone hovering position An exhaustive search is performed on the subproblem of scheduling coefficients to obtain the updated scheduling coefficients. ,Right now:

[0136] (17)

[0137] In step S32 above, based on the calculation formula of the average mean square error between the UAV's received signal and the aggregated signal, and according to the updated scheduling coefficients... and the initial sensor transmission coefficient and drone hovering position Noise reduction factor for drones By performing differentiation, a closed-form solution is derived.

[0138] Specifically, in formula (13), for each time slot We extract and expand the mean squared error. In the calculation formula and The relevant items are:

[0139] (18)

[0140] By taking the derivative and setting it to zero, we obtain the closed-form solution of the UAV denoising factor as the updated UAV denoising factor, as shown below:

[0141] (19)

[0142] For the eavesdropper noise reduction factor, the updated scheduling coefficients are also fixed first. and the initial sensor transmission coefficient and drone hovering position Denoising factor for eavesdroppers Perform differentiation, then... Taking the derivative and setting it to zero, we obtain a closed-form solution for the eavesdropping denoising factor, which is then used as the updated eavesdropping denoising factor, as shown below:

[0143] (20)

[0144] In step S33 above, the joint optimization problem is transformed into a sensor transmission coefficient subproblem P4, namely:

[0145] (twenty one)

[0146] Using the solution obtained in step S32 Substituting the closed-form solution into (21), we get:

[0147] (twenty two)

[0148] So, the scheduling coefficient subproblem It can be rephrased as:

[0149] (twenty three)

[0150] Since the constraint in equation (21) is non-convex, this embodiment employs a continuous convex approximation algorithm to iteratively solve an approximation problem in order to address this issue. Specifically, in each iteration... In China, Applying a first-order Taylor expansion to approximate the constraint condition (21), we obtain:

[0151] (twenty four)

[0152] in,

[0153] (25)

[0154] Therefore, question P5 can be rewritten as:

[0155] (26)

[0156] At this point, the updated sensor transmission coefficients can be approximately solved using the convex optimization problem P6 from the CVX toolbox. .

[0157] In step S34 above, the joint optimization problem is transformed into the UAV hovering position subproblem P7, namely:

[0158] (27)

[0159] question It is about Given the unconstrained nonconvex optimization problem, this embodiment employs the gradient descent algorithm to solve the UAV hovering position subproblem P7 based on the updated scheduling coefficients, UAV noise reduction factor, and sensor transmission coefficients, thereby obtaining the updated UAV hovering position. .

[0160] Please refer to Table 1, which shows the algorithm flow for optimizing the scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position using the alternating optimization algorithm within the multi-time-slot joint optimization framework.

[0161] Table 1. Algorithm flow for optimizing scheduling coefficients, sensor transmission coefficients, UAV noise reduction factor, and UAV hovering position using the alternating optimization algorithm.

[0162]

[0163] This embodiment, after obtaining the cell cluster strategy partitioning scheme, establishes a joint optimization problem with the objective of minimizing the mean square error between the received signal and the aggregated signal of the UAV, under the conditions of satisfying the eavesdropper security constraints and the sensor maximum transmission power constraints. Through an alternating optimization method, scheduling coefficients, sensor transmission coefficients, UAV noise reduction factors, and UAV hovering positions are jointly designed to minimize the computational distortion of the legitimate receiver while ensuring that the mean square error of the eavesdropper is always higher than a preset security threshold. For the resulting non-convex optimization problem, a highly efficient solution method based on alternating optimization and successive convex approximation is further proposed. This method significantly improves computational accuracy while ensuring security, providing an effective solution for dense IoT environments with eavesdropping risks.

[0164] To verify the technical effects of the present invention, this embodiment also provides corresponding simulation experiments. Please refer to [link / reference]. Figures 3-7 .

[0165] In the simulation experiment, the service area of ​​the drone is a square area with a side length of 2000. 20 sensors are randomly generated in this area. The xy coordinates of the eavesdropper are (400, 0) and (1600, 1600) respectively. The drone's flight altitude is The eavesdropper flew at an altitude of The reference channel power gain is set to AWGN power set dBm, the maximum transmit power of each ground sensor is set to = 2W, the transmission power of each eavesdropper is set to 2W. = 1W, .

[0166] To evaluate the performance of the proposed solution, the present invention is compared with the following four benchmark solutions: (1) Benchmark solution with constant transmission coefficient. Each sensor communicates using maximum transmission power during in-flight computing missions, i.e. (2) Benchmark with Rectangular Clustering: The entire area is divided into rectangular regions by uniformly dividing the sensor. The drone is assigned a rectangular sub-region, with the center point of each sub-region serving as its hovering position. Figure 3 As shown, it is evenly distributed within this area. One ground sensor, (3) Baseline scheme based on K-means algorithm (K-means algorithm): This scheme uses the K-means algorithm to... Each sensor is divided into Each community, such as Figure 4 As shown. (4) Benchmark with constant UAV position The sensor partitioning is achieved using a support-bounded framework, with each drone hovering position serving as a constant sensor center for each partition. Optimize, such as Figure 5 As shown.

[0167] Figures 3 to 5 The results of sensor partitioning under three different baseline schemes and the optimal UAV position are presented. All are based on the rectangular partitioning scheme. The total distance from the ground sensors to the drone is greater than that of the K-means algorithm and branch-and-bound scheme, leading to higher path loss and thus reducing AirComp performance. The K-means algorithm is a drone-sensor association scheme based on minimizing distance. It performs well in scenarios where physical layer security is not considered. If an active eavesdropper is present, the algorithm ignores interference noise and blindly seeks the center point that minimizes the sum of Euclidean distances within the cluster. However, in scenarios with security threats, this "distance-optimal" clustering method may not be "security-optimal." This singular optimization of distance may push the drone's hovering position and sensor clustering closer to the eavesdropper, unintentionally providing a better eavesdropping channel. Therefore, as... Figure 6 As shown, our proposed solution enables better sensor correlation and drone deployment.

[0168] Figure 7 The average FC of different drone deployment centers was compared. With maximum transmission power The relationship. It can be observed that, with... The increase of all schemes average All MSE values ​​decrease. This trend is expected, as higher transmit power improves the network's ability to combat channel fading. It was also observed that the constant power benchmark had the highest average MSE and decreased the slowest with increasing power, highlighting the importance of adaptive power control and signal alignment in AirComp. Furthermore, the proposed scheme significantly outperforms both the rectangular partitioning and K-means partitioning benchmarks, as the former ignores sensor density and channel quality, while the latter fails to consider EVE location. Finally, the performance gain of the proposed scheme relative to the fixed partition center benchmark highlights the benefits of utilizing spatial degrees of freedom in UAV localization, as fixed hovering points limit the system's ability to compensate for poor channel conditions at the edge devices. In summary, the proposed scheme consistently achieves lower MSE, further confirming its advantages in airborne computing in scenarios with active eavesdroppers.

[0169] The UAV-assisted multi-cluster aerial computing device provided in the embodiments of this application is described below. The UAV-assisted multi-cluster aerial computing device described below and the UAV-assisted multi-cluster aerial computing method described above can be referred to and corresponded to each other.

[0170] Please see Figure 8 This invention also provides a drone-assisted multi-cluster aerial computing device resistant to active eavesdropping, the device comprising:

[0171] The information acquisition module 201 is used to acquire system information of an airborne computing system consisting of a drone, multiple sensors, and multiple eavesdroppers.

[0172] The channel gain model establishment module 202 is used to establish channel gain models between the UAV and the sensor, the eavesdropper and the sensor, the UAV and the eavesdropper, and each of the eavesdroppers based on the channel information in the system information;

[0173] The cell cluster strategy generation module 203 is used to divide the sensor cell clusters according to the system information using a branch and bound framework to obtain the cell cluster strategy.

[0174] The joint optimization problem establishment module 204 is used to establish a joint optimization problem with the objective of minimizing the average mean square error between the received signal and the aggregated signal of the UAV, based on the cell clustering strategy and all the channel gain models, under the conditions of satisfying the eavesdropper security constraint and the sensor maximum transmission power constraint.

[0175] The optimal strategy output module 205 is used to iteratively solve the joint optimization problem using an alternating optimization algorithm, and output the optimal scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor and UAV hovering position.

[0176] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the UAV-assisted multi-cluster aerial computing method as described above.

[0177] This invention also provides a computer-readable storage medium storing a computer program or instructions thereon, which, when executed by a processor, implements the steps of the UAV-assisted multi-cluster aerial computing method described above.

[0178] This invention also provides a computer program product, including a computer program or instructions, which, when executed by a processor, implement the steps of any of the above-described UAV-assisted multi-cluster aerial computing methods.

[0179] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0180] Furthermore, the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion, such that a process, method, system, product, or apparatus that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or units that are not explicitly listed or that are inherent to such process, method, product, or apparatus.

[0181] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection between apparatuses or units through some interfaces, and may be electrical, mechanical, or other forms.

[0182] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0183] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0184] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0185] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. 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 of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for UAV-assisted multi-swarm aerial computing resistant to active eavesdropping, characterized in that, The method includes: Acquire system information from an aerial computing system consisting of drones, multiple sensors, and multiple eavesdroppers; Based on the channel information in the system information, establish channel gain models between the UAV and the sensor, between the eavesdropper and the sensor, between the UAV and the eavesdropper, and between each of the eavesdroppers; A branch-and-bound framework is used to divide sensor cell clusters based on the system information to obtain a cell clustering strategy; Based on the cell clustering strategy and all the channel gain models, under the conditions of satisfying the eavesdropper security constraints and the sensor maximum transmission power constraints, a joint optimization problem is established with the objective of minimizing the average mean square error between the received signal and the aggregated signal of the UAV. The joint optimization problem is solved iteratively using an alternating optimization algorithm, and the optimal scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position are output. The joint optimization problem P0 is represented as follows: ； In the formula: In time slot n The hovering position of the drone below For cell clusters m A collection of sensors, For cell clusters m The coverage radius, In time slot n The sensor transmission coefficient below, In time slot n The noise reduction factor of the drone is below. In time slot n The following cell cluster m The scheduling coefficient, In time slot n The mean square error between the received signal and the aggregated signal from the drone. In time slot n The noise reduction factor of the eavesdropper, In time slot n The mean square error between the received signal and the aggregated signal of the eavesdropper. To preset a safety threshold, This is the sensor's maximum transmission power. For the first The fixed horizontal coordinates of each sensor M This represents the total number of cells in the cluster. The average mean square error between the received signal and the aggregated signal of the UAV is calculated as follows: ； In the formula: In time slot n The drone receives signals. In time slot n Drone cluster in the community m The aggregated signal obtained at the location; In time slot n The drone and the first k Channel gain between sensors The eavesdropper's transmission power, In time slot n The drone and the first v Channel gain between eavesdroppers Let n be the additive white Gaussian noise power at the UAV in time slot n; The mean square error between the eavesdropper's received signal and the aggregated signal is calculated as follows: ； In the formula: In time slot n The eavesdropper's received signal; In time slot n Next v The eavesdropper and the first k Channel gain between sensors In time slot n Next v The eavesdropper and the first Channel gain between eavesdroppers In time slot n The self-interference channel gain of the eavesdropper In time slot n The power of additive white Gaussian noise at the location of the eavesdropper.

2. The UAV-assisted multi-cluster aerial computing method according to claim 1, characterized in that, The step of using a branch-and-bound framework to divide sensor cell clusters based on the system information to obtain a cell cluster strategy includes: Using a branch-and-bound framework, based on the system information, a cell cluster model is established with the objective of minimizing the maximum coverage radius of all cell clusters and the constraint that the distance between the center of each cell cluster and the location of all eavesdroppers is greater than a preset security radius threshold. By constructing a search tree to traverse all candidate partitions in the cell cluster model, the candidate partitions are optimized based on a depth-first search strategy and pruning techniques. The non-convex optimization problem of the optimized candidate partition is solved by the minimum enclosing circle process, and the cell cluster strategy is output. The cell cluster strategy includes the cell cluster to which each sensor belongs, the coverage radius of each cell cluster, and the center position of each cell cluster.

3. The UAV-assisted multi-cluster aerial computing method according to claim 2, characterized in that, The step of iteratively solving the joint optimization problem using an alternating optimization algorithm to output the optimal scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position includes: Initialize the optimization variables for the joint optimization problem; the optimization variables include initial scheduling coefficients, sensor transmission coefficients, UAV noise reduction factor, and UAV hovering position; wherein, under the initial scheduling coefficients, the initial UAV hovering position is determined by the center position in the cell cluster strategy; The mean square error between the initial received signal and the aggregated signal of the UAV is calculated based on the initial scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position. The parameters of the joint optimization problem are optimized by an alternating optimization algorithm, and the updated scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position are obtained by alternating optimization. The mean square error between the received signal and the aggregated signal of the updated drone is calculated based on the updated scheduling coefficient, sensor transmission coefficient, drone noise reduction factor, and drone hovering position. If the absolute value of the difference between the average mean square error between the updated UAV's received signal and the aggregated signal and the initial average mean square error between the UAV's received signal and the aggregated signal is greater than a preset convergence threshold, then the updated scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position are used as the initial scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position, and the process jumps to execute the step of calculating the initial average mean square error between the UAV's received signal and the aggregated signal based on the initial scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position. If the absolute value of the difference between the mean square error between the received signal and the aggregated signal of the updated UAV and the mean square error between the received signal and the aggregated signal of the initial UAV is less than or equal to the preset convergence threshold, then the updated scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position are used as the optimal scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position and output.

4. The UAV-assisted multi-cluster aerial computing method according to claim 3, characterized in that, The step of using an alternating optimization algorithm to optimize the parameters of the joint optimization problem, and obtaining updated scheduling coefficients, sensor transmission coefficients, UAV noise reduction factors, and UAV hovering positions through alternating optimization, includes: Under the alternating optimization algorithm, the joint optimization problem is transformed into a scheduling coefficient subproblem. Based on the initial sensor transmission coefficients, UAV noise reduction factor and UAV hovering position, the scheduling coefficient subproblem is exhaustively searched to obtain the updated scheduling coefficients. Based on the calculation formula of the average mean square error between the received signal and the aggregated signal of the UAV, the updated UAV noise reduction factor is solved according to the updated scheduling coefficient, the initial sensor transmission coefficient and the UAV hovering position. The joint optimization problem is transformed into a sensor transmission coefficient subproblem. The sensor transmission coefficient subproblem is solved using a continuous convex approximation method based on the updated scheduling coefficients, the UAV noise reduction factor, and the initial UAV hovering position, to obtain the updated sensor transmission coefficients. The joint optimization problem is transformed into a UAV hovering position subproblem. The gradient descent algorithm is used to solve the UAV hovering position subproblem based on the updated scheduling coefficients, UAV noise reduction factor and sensor transmission coefficient to obtain the updated UAV hovering position.

5. A drone-assisted multi-cluster aerial computing device resistant to active eavesdropping, characterized in that, The apparatus is used to implement the anti-active eavesdropping UAV-assisted multi-cluster aerial computing method as described in any one of claims 1-4, the apparatus comprising: The information acquisition module is used to acquire system information from an airborne computing system consisting of a drone, multiple sensors, and multiple eavesdroppers. The channel gain model establishment module is used to establish channel gain models between the UAV and the sensor, the eavesdropper and the sensor, the UAV and the eavesdropper, and each of the eavesdroppers based on the channel information in the system information; The cell cluster strategy generation module is used to divide sensor cell clusters according to the system information using a branch and bound framework to obtain cell cluster strategies. The joint optimization problem establishment module is used to establish a joint optimization problem with the objective of minimizing the average mean square error between the received signal and the aggregated signal of the UAV, based on the cell clustering strategy and all the channel gain models, under the conditions of satisfying the eavesdropper security constraint and the sensor maximum transmission power constraint. The optimal strategy output module is used to iteratively solve the joint optimization problem using an alternating optimization algorithm, and output the optimal scheduling coefficient, sensor transmission coefficient, UAV noise reduction factor, and UAV hovering position.

6. A computer device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the UAV-assisted multi-cluster aerial computing method as described in any one of claims 1-4.

7. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed by the processor, they implement the steps of the UAV-assisted multi-cluster aerial computing method as described in any one of claims 1-4.

8. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by the processor, they implement the steps of the UAV-assisted multi-cluster aerial computing method as described in any one of claims 1-4.

Images