Unmanned aerial vehicle path privacy protection and task optimization method

Abstract

The invention relates to the technical field of path privacy protection and task scheduling of unmanned aerial vehicles in smart cities, in particular to a path protection and task optimization method combining differential privacy budget allocation and false task point insertion. The method comprises the following steps: firstly, generating an initial unmanned aerial vehicle path based on a task demand, calculating global sensitivity of each path segment, and dynamically allocating a total privacy budget to different path segments according to the sensitivity; performing differential privacy perturbation on the coordinates of the path segment by using a Laplace mechanism, and inserting false task points on a perturbation path to confuse potential attackers; and finally, constructing a weighted optimization model of energy consumption and path offset, and reconstructing a disturbance path to meet task time limit and energy consumption constraints. Experimental results show that the method significantly reduces energy consumption while improving the privacy protection rate, and is suitable for various application scenarios such as unmanned aerial vehicle parcel delivery and Internet of Things data acquisition in smart cities.

Description

Technical Field

[0001] This invention belongs to the field of drone path planning and privacy protection technology, specifically involving a drone path privacy protection and task optimization method based on differential privacy budget allocation and sham task point insertion, which can be widely applied to tasks such as package delivery, environmental monitoring, and IoT data collection in smart cities. Background Technology

[0002] With the widespread application of unmanned aerial vehicles (UAVs) in smart city environments, they have played a vital role in areas such as express delivery, emergency rescue, and IoT data collection. However, when UAVs perform missions, their flight paths and mission point information may be collected and analyzed through external observation, wireless communication interception, and other methods, thereby exposing users' privacy information and even posing security risks.

[0003] Existing privacy protection methods mainly fall into two categories: encrypted communication and trajectory perturbation. Encrypted communication can protect the data transmitted over the link, but it cannot hide the physical flight path of the drone. Trajectory perturbation methods, on the other hand, hide the real mission point location by modifying the path, including inserting spoofed mission points and randomly shifting the path. However, purely engineering-based methods, such as inserting spoofed mission points, lack formal privacy guarantees and are difficult to quantify the level of protection. While purely algorithmic methods, such as differential privacy path perturbation, have strict mathematical privacy boundaries, they ignore energy consumption and time constraints during mission execution, resulting in low efficiency in actual deployment.

[0004] Differential privacy, as a rigorous privacy-preserving mathematical model, has been introduced into UAV path protection in recent years. It achieves obfuscation of location data by adding random noise that follows a probability distribution to the path coordinates. However, in actual UAV missions, simple differential privacy perturbations can lead to increased path offsets and even affect mission reachability; similarly, relying solely on spurious mission points may be vulnerable to pattern matching algorithms due to a lack of coordinate perturbation.

[0005] Therefore, there is an urgent need for a method that can organically combine differential privacy perturbation with spoof task point insertion mechanism, so as to ensure the strength of privacy protection while taking into account path feasibility, task completion rate and energy consumption optimization. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for protecting and optimizing UAV path privacy by combining differential privacy perturbation and spurious task point insertion. Through segmented sensitivity analysis, dynamic privacy budget allocation, differential privacy position perturbation, dynamic spurious task point insertion, and path optimization reconstruction, this method achieves high-strength privacy protection for UAV flight paths while maintaining high mission execution efficiency within flight energy consumption and mission time limits.

[0007] To achieve the above objectives, the technical solution adopted by the present invention is: a method for protecting the privacy of unmanned aerial vehicle (UAV) paths and optimizing tasks, comprising the following steps:

[0008] S1. Generate the initial path for the task;

[0009] S2. Conduct path sensitivity calculation: Perform privacy sensitivity assessment on each segment of the initial path to quantify the potential privacy leakage risk during task execution;

[0010] S3. Privacy Budget Allocation: Set an overall privacy protection budget and allocate the budget proportionally according to the sensitivity of each path segment to match the protection strength with the path risk;

[0011] S4. Differential Privacy Perturbation: For path segments allocated to different privacy budgets, their geographical coordinates are randomly perturbed to generate perturbed paths;

[0012] S5. Enhanced privacy through spoofing task points: In the perturbation path, several spoofing task points are dynamically inserted based on the remaining privacy budget and task time margin.

[0013] S6. Path Optimization and Reconstruction: After the disturbance path and false mission point insertion are completed, a weighted optimization model that comprehensively considers flight energy consumption and path offset is constructed. Under the condition of meeting the mission time limit and energy consumption constraints, the path is optimized as a whole to generate the final flight path.

[0014] In step S1, the initial flight path of the UAV is generated using a path planning algorithm based on the set of target points allocated by the task scheduling center.

[0015] Step S4 includes:

[0016] S41. Select a perturbation mechanism: Use either the Laplace mechanism or the Gaussian mechanism to perturb the path coordinates;

[0017] S42, Path Coordinate Noise Injection: Treat the starting and ending coordinates of the path segment as perturbation units, and add random noise that follows a Laplace distribution according to the allocated privacy budget. The noise amplitude is directly proportional to the sensitivity and inversely proportional to the budget.

[0018] S43. Flight Accessibility Correction: If the coordinates fall into an unflyable area after the disturbance, the constrained projection method is used to adjust them to the nearest flyable position while preserving the random disturbance characteristics.

[0019] S44. Maintain flight path continuity: After independently perturbing each segment of the path, smoothing is performed to obtain a smooth perturbed path trajectory.

[0020] In step S41, the Laplace mechanism, which conforms to the definition of differential privacy, is used first to perturb the path coordinates; when the location sensitivity is extremely high or the perturbation requirement is small, the Gaussian mechanism is switched to perturb the path coordinates.

[0021] Step S5 includes: selecting locations that are similar to the actual task points but unrelated to the task, based on the geometric shape of the disturbance path and the set of candidate stopping points, as false task points;

[0022] Parameters for fake task points are set to be similar to those of real task points in terms of task type and dwell time, making it difficult for external observers to distinguish between real and fake tasks based on dwell time characteristics.

[0023] The confirmation of the number of fake task points includes: dynamically determining the number of fake task points by combining the remaining privacy budget and the remaining task time.

[0024] After inserting fake task points, the path is locally adjusted to make the real and fake task points interspersed and distributed, avoiding concentration at the end of the path or in specific areas, thus reducing the success rate of pattern recognition attacks.

[0025] In step S3, the allocation of the privacy budget for the drone mission adopts a dynamic risk perception method, which automatically adjusts the privacy budget ratio of each mission segment based on the importance of the mission, the execution progress, and the drone status.

[0026] The allocation of the privacy budget in step S3 includes:

[0027] 1) Budget Vector Initialization: The system first establishes the privacy budget vector for all task points in the task set M. Initially, all values ​​are set to zero.

[0028] 2) Task risk calculation and correction: For each task point in set M... The base risk value is calculated based on the sensitivity of the mission location, the mission type, and information that external attackers may exploit. According to the planned execution ratio of the task. and the system-defined time decay coefficient The correction factor is obtained. When the task has not yet started or is at a low progress, the correction factor is increased to allocate more budget. The base risk value is multiplied by the correction factor to obtain the comprehensive risk weight.

[0029] 3) The system will allocate a total privacy budget. The privacy budget is allocated proportionally according to the comprehensive risk weight. The higher the weight of the task segment, the more privacy budget it receives, thereby introducing stronger protection in differential privacy perturbation.

[0030] 4) All task points are assigned a comprehensive risk weight. Sort tasks from largest to smallest to generate a high-risk task priority queue. In the subsequent path disturbance and false task insertion process, high-risk task points should be protected first;

[0031] 5) The system monitors the remaining energy of the drone in real time. Remaining task time Status parameters: When a status change is detected to exceed a set threshold, steps 2–4 are re-executed to ensure that the budget allocation is consistent with the current task environment.

[0032] 6) Finally, the system outputs the allocated privacy budget vector. and high-risk task priority queue .

[0033] In step S2, path sensitivity analysis, a path query function f is defined, whose input is the complete UAV task path and whose output is a feature vector. These correspond to the total path distance, total flight time, and total energy consumption, respectively.

[0034] The sensitivity calculation process fixes all path segments except for the i-th segment, replaces the i-th segment with a neighboring feasible path segment, obtains the characteristic change amount through flight simulation, and calculates the comprehensive sensitivity:

[0035]

[0036] in, , , For fixed weighting coefficients, and satisfying , , , These represent the changes in total mission distance, total flight time, and total energy consumption, respectively.

[0037] The advantages of this invention are as follows: 1. A dual-layer privacy protection mechanism enhances resistance to inference: This invention is the first to organically combine differential privacy perturbation with a false task point insertion mechanism, forming a dual-layer protection structure that confuses path coordinate perturbation with task target obfuscation. Differential privacy perturbation can effectively obscure the actual position of the UAV, preventing the position data from being directly reconstructed; false task point insertion can break the recognizability of trajectory patterns, significantly increasing the difficulty for attackers to infer the real task target, thus possessing high resistance to inference under various attack models.

[0038] 2. Sensitivity-driven privacy budget allocation to reduce path offset: Unlike existing methods that uniformly allocate privacy budgets, this invention uses path sensitivity analysis to allocate more budget to path segments with higher privacy risks. This ensures privacy strength at high-risk locations while reducing excessive disturbance to low-risk path segments, thereby reducing the overall path offset and ensuring the feasibility and stability of the flight path.

[0039] 3. Task-constrained path optimization for improved execution efficiency: After privacy protection processing, this invention introduces an optimization model that comprehensively considers flight energy consumption and mission time limits to globally reconstruct disturbed paths. This process not only ensures that the path meets mission requirements in terms of both energy consumption and time, but also minimizes the decrease in flight efficiency caused by privacy processing, guaranteeing timely and efficient mission completion.

[0040] 4. Enhanced System Robustness and Security: Through the mathematically provable protection mechanism of differential privacy and the strategic insertion of spoofed task points, this invention can maintain a high level of privacy protection and task execution success rate when facing various threats such as trajectory reconstruction attacks, statistical analysis attacks, and correlation analysis attacks, thereby significantly improving the overall robustness and operational security of the UAV system. Attached Figure Description

[0041] The following is a brief explanation of the contents of each of the accompanying drawings and the markings in the drawings:

[0042] Figure 1 This is a schematic diagram of the overall framework of the method of the present invention, including modules for original path generation, sensitivity analysis, privacy budget allocation, differential privacy perturbation, spoof task point insertion, and path optimization.

[0043] Figure 2 This is a schematic diagram of the differential privacy perturbation mechanism, which shows the process of randomly perturbing UAV pathpoints based on the Laplace mechanism.

[0044] Figure 3 This is a schematic diagram of spurious task point insertion, showing the mixed distribution of real and spurious task points in the perturbation path.

[0045] Figure 4 This is a schematic diagram of the path optimization results, showing the final flight path of the UAV under privacy protection and energy consumption constraints; Figure 5 The figure shows the simulation test results comparing this solution with existing technical solutions. Detailed Implementation

[0046] The specific embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and the description of the preferred embodiments.

[0047] This invention relates to the field of drone path privacy protection and task scheduling technology in smart cities, and particularly to a path protection and task optimization method combining differential privacy budget allocation and spoofed task point insertion. The method first generates an initial drone path based on task requirements and calculates the global sensitivity of each path segment, dynamically allocating the total privacy budget to different path segments according to the sensitivity. Then, it uses the Laplace mechanism to differentially privacy-perturb the path segment coordinates and inserts spoofed task points on the perturbed path to confuse potential attackers. Finally, it constructs a weighted optimization model of energy consumption and path offset to reconstruct the perturbed path to meet task time and energy consumption constraints. Experimental results show that this invention significantly reduces energy consumption while improving privacy protection, and is suitable for various application scenarios in smart cities, such as drone package delivery and IoT data collection. The specific solution is as follows:

[0048] like Figure 1 As shown, a path protection and task optimization method combining differential privacy budget allocation and spoof task point insertion includes the following steps:

[0049] S1. Generate Initial Flight Path: Based on the set of target points allocated by the task scheduling center, generate the initial flight path of the UAV using a path planning algorithm (such as Dijkstra's algorithm). This path comprehensively considers the spatial distribution of task points, task priority, and flight distance, providing a basic path for subsequent privacy protection processing.

[0050] S2. Conduct path sensitivity calculations: Perform a privacy sensitivity assessment on each segment of the initial path to quantify the potential privacy leakage risk during task execution. Path segments with higher sensitivity are more easily inferred by attackers from the true task information, therefore requiring more privacy protection resources to be allocated in subsequent steps.

[0051] S3. Privacy Budget Allocation: Set an overall privacy protection budget and allocate it proportionally according to the sensitivity of each path segment to match the protection strength with the path risk. High-risk path segments receive more budget, while low-risk path segments receive less budget, in order to balance protection effectiveness and task efficiency.

[0052] S4. Differential Privacy Perturbation: For path segments allocated to different privacy budgets, the geographic coordinates are randomly perturbed using a differential privacy mechanism (such as the Laplace mechanism) to generate a perturbed path, which effectively obscures the true flight position and prevents external inference of path details.

[0053] S5. Enhanced Privacy Through Fake Mission Point Insertion: Within the perturbation path, several fake mission points are dynamically inserted based on the remaining privacy budget and mission time margin. These mission points are indistinguishable from real mission points in external observation, significantly increasing the difficulty for attackers to analyze the path. The number and location of insertions are dynamically adjusted according to the real-time mission situation to strengthen privacy protection without significantly increasing flight costs.

[0054] S6. Path Optimization and Reconstruction: After the perturbation path and spurious mission point insertion are completed, a weighted optimization model that comprehensively considers flight energy consumption and path offset is constructed. Under the condition of meeting mission time and energy consumption constraints, the path is optimized as a whole to generate the final flight path, which has both high privacy and ensures the successful completion of the mission.

[0055] Secondly, after allocating privacy budgets to path segments, this step performs differential privacy perturbation processing for each path segment, specifically including:

[0056] S41. Select perturbation mechanism: Prioritize using the Laplace mechanism that conforms to the differential privacy definition to perturb the path coordinates; in cases of extremely high location sensitivity or low perturbation requirements, the Gaussian mechanism can be switched to reduce additional energy consumption.

[0057] S42, Path Coordinate Noise Injection: Treat the starting and ending coordinates of the path segment as perturbation units, and add random noise that follows a Laplace distribution according to the allocated privacy budget. The noise amplitude is directly proportional to the sensitivity and inversely proportional to the budget.

[0058] S43. Flight Accessibility Correction: If the coordinates after the disturbance fall into a non-flyable area (such as a building or no-fly zone), the constrained projection method is used to adjust them to the nearest flyable position while preserving the random disturbance characteristics.

[0059] S44. Maintain flight path consistency: After independently perturbing each segment of the path, smoothing is performed to reduce abrupt turns or sharp bends, ensuring that the flight trajectory is physically feasible and stable.

[0060] Finally, after completing the path perturbation, the real mission information is further obfuscated by adding false mission points into the flight path, specifically including:

[0061] S51. Insertion location selection: Based on the geometry of the disturbance path and the set of candidate stopping points, priority is given to selecting locations that are similar to the actual task point distribution but are irrelevant to the task.

[0062] S52. Determine the number: Dynamically determine the number of fake mission points by combining the remaining privacy budget and mission time slack; increase the number when time is ample to enhance privacy, and decrease it when time is tight to control flight costs;

[0063] S53. Task Attribute Forgery: Set parameters for fake task points that are similar to those of real task points in terms of task type, dwell time, etc., making it difficult for external observers to distinguish between real and fake tasks by dwell characteristics.

[0064] S54. Optimize the overall path structure: After inserting fake task points, make local adjustments to the path so that real and fake task points are interspersed and distributed, avoiding concentration at the end of the path or in specific areas, thereby reducing the success rate of pattern recognition attacks.

[0065] like Figure 1 As shown, this invention proposes a method for protecting the privacy of UAV paths, comprising six main steps: path generation, path sensitivity analysis, differential privacy budget allocation, path perturbation handling, insertion of false mission points, and path optimization and reconstruction. This method effectively reduces the probability of external observers inferring the true mission path through a two-layer privacy protection mechanism, while simultaneously considering flight energy consumption and mission completion efficiency.

[0066] The method specifically includes the following steps:

[0067] In step S1, which involves generating the original path and defining constraints, the system receives a set of task points assigned by the task center.

[0068] Where T is the set of task points, t1-t n This is the task point.

[0069] First, the mission points are mapped onto a 3D city flight map, including information such as flyable corridors, no-fly zones, and building outlines. Based on the UAV's maximum flight radius, unit energy consumption function, flight speed limits, and mission completion time, the Dijkstra algorithm is used to generate the initial path. .

[0070] Flight energy consumption can be calculated using the following formula:

[0071]

[0072] in, For hovering power, For flight power, and These are hovering time and flight time, respectively. To improve the robustness of the path under subsequent disturbances, a set of alternative waypoints is introduced during the path generation phase for replacement by subsequent disturbances or spurious mission points.

[0073] Path planning can be generated using shortest path search algorithms (such as Dijkstra's algorithm) and combined with obstacle maps to ensure that the path does not enter no-fly zones. For missions requiring high-altitude flight, 3D spatial modeling can also be considered to meet airspace management requirements.

[0074] In this embodiment, the allocation of the drone mission privacy budget employs a dynamic risk perception method. Based on the mission's importance, execution progress, and drone status, the privacy budget ratio for each mission segment is automatically adjusted, thereby ensuring strong privacy protection while reducing energy consumption and maintaining mission execution efficiency. This method includes the following steps:

[0075] 1) Budget vector initialization

[0076] The system first establishes a privacy budget vector for all task points in the task set M. Here, "task point" refers to the specific geographical location that the drone needs to reach during the mission, such as a package delivery receiving point, an environmental monitoring sampling point, or an IoT data collection device point. This budget vector is initially set to zero to prepare for subsequent allocation.

[0077] 2) Task risk calculation and correction: For each task point in set M... The base risk value is calculated based on the sensitivity of the mission location, the mission type, and information that external attackers may exploit. According to the planned execution ratio of the task. and the system-defined time decay coefficient The correction factor is obtained. When the task has not yet started or is at a low progress, the correction factor is increased to allocate more budget. The base risk value is multiplied by the correction factor to obtain the comprehensive risk weight.

[0078] 3) The system will allocate a total privacy budget. The privacy budget is allocated proportionally based on the comprehensive risk weight. Task segments with higher weights receive more privacy budget, thereby introducing stronger protection in differential privacy perturbations.

[0079] 4) All task points are assigned a comprehensive risk weight. Sort tasks from largest to smallest to generate a high-risk task priority queue. This queue is an ordered queue of task point sets, used to determine the priority of protection during subsequent path perturbations and spoof task insertions, while the privacy budget vector... This refers to the specific budget value allocated to each task point.

[0080] 5) The system monitors the remaining energy of the drone in real time. Remaining task time Status parameters. When a status change exceeds a set threshold, steps 2–4 are re-executed to ensure that the budget allocation remains consistent with the current mission environment. "Status change exceeding a set threshold" here refers to a situation where, due to unforeseen circumstances such as increased energy consumption caused by high wind resistance or delays in mission point processing, the actual status parameters of the drone differ from the baseline value at the time of the last budget allocation by an amount exceeding the preset tolerance range (e.g., energy decrease exceeding 10%). Once this threshold is exceeded, the system will automatically re-execute steps 2–4 to dynamically adjust the privacy budget allocation, ensuring that the new protection scheme remains feasible and efficient in the current mission environment.

[0081] 6) Finally, the system outputs the allocated privacy budget vector. and high-risk task priority queue This is used by the subsequent differential privacy perturbation module and spoof task point insertion module. It is implemented using Algorithm 1, which is detailed below:

[0082] Algorithm 1: Dynamic Risk Awareness Privacy Budget Adaptive Allocation (DRAPBA)

[0083] Input: Task set M, risk assessment result Ri, UAV state parameters ( Total privacy budget Risk weighting coefficient .

[0084] Output: Privacy pre-screening allocation vector for each task High-risk task priority queue .

[0085] 1. Initialize the budget allocation vector Empty;

[0086] 2. For each task point ∈M execute;

[0087] (1) Calculate the basic risk value ;

[0088] (2) Calculate the task schedule correction factor ;

[0089] (3) Calculate the overall risk weight ;

[0090] 3. Allocate the privacy budget proportionally:

[0091] * .

[0092] 4. According to Generate a priority queue for high-risk tasks by sorting them from largest to smallest. .

[0093] 5. If the change in UAV status parameters exceeds the threshold, return to step 2 to recalculate.

[0094] 6. Output and .

[0095] Furthermore, in step S2, path sensitivity analysis, a path query function f is defined, whose input is the complete UAV task path and whose output is a feature vector. These correspond to the total path distance, total flight time, and total energy consumption, respectively.

[0096] The sensitivity calculation process fixes all path segments except for the i-th segment, replaces the i-th segment with a neighboring feasible path segment, obtains the characteristic change amount through flight simulation, and calculates the comprehensive sensitivity:

[0097]

[0098] in, , , For fixed weighting coefficients (satisfying) ), , , These represent the changes in total mission distance, total flight time, and total energy consumption, respectively.

[0099] Sensitivity reflects the impact of a path segment's leakage on the overall task privacy; a higher value indicates a more sensitive segment. Sensitivity calculation employs a three-dimensional weighted model to ensure that the combined effects of distance, time, and energy consumption are considered across different task types.

[0100] Furthermore, in step S3, differential privacy budget allocation, let the total privacy budget be... Budget allocation follows the principle of "high sensitivity segment, high protection," and introduces a time decay factor for task execution order. :

[0101]

[0102] Where γ>0 is the time decay coefficient, and i represents the index of the execution order of the path segment.

[0103] The budget is proportionally divided into disturbance budgets. With fake mission budget :

[0104]

[0105] in A fixed scaling factor, typically 0.6–0.7, is used to ensure that both coordinate perturbation and spoofing insertion have sufficient budget support, achieving dual-layer privacy protection.

[0106] Figure 2 This diagram illustrates a differential privacy perturbation mechanism, demonstrating how a drone randomly perturbs its path coordinates based on the Laplace distribution during mission execution. The solid line represents the original, unprocessed path, the dashed line represents the perturbated flight path, and the gray area indicates a no-fly zone. By adding random noise following a Laplace distribution to the path coordinates, a spatial deviation is created between the perturbed path and the actual path, effectively hiding the true mission location. If a path point falls into a no-fly zone during perturbation, an accessibility correction algorithm adjusts it to the nearest flyable location, ensuring flight safety and mission feasibility.

[0107] Furthermore, in step S4, path perturbation and reachability correction, for each path segment coordinate, based on the perturbation budget... Add random noise that satisfies differential privacy, where, Represents the original coordinates of the i-th path segment, while These are the perturbation coordinates after adding Laplace noise. This is intended to effectively obscure the drone's true flight path through randomization.

[0108]

[0109] in This represents Laplace-distributed noise with a mean of 0 and a scale of b. For path segments located in areas with high repetition or high-density observations, a Gaussian mechanism is used.

[0110] .

[0111] Here, we use a value with a mean of 0 and a variance of 0. The mechanism introduces Gaussian noise to perturb the original path coordinates, thus mitigating the normally distributed noise. Generate new perturbation coordinates Here, δ is a manually set tolerance probability parameter used to adjust the stringency of privacy protection. This parameter is pre-set according to privacy protection needs to define the stringency of privacy protection. A lower value represents stricter and more reliable privacy protection.

[0112] After the disturbance is completed, reachability correction is performed: if the disturbance coordinates are located in a no-fly zone, the constraint projection is applied. The point is mapped to the nearest flyable location, ensuring that the corrected point still satisfies the differential privacy distribution properties. Finally, Bézier curve or B-spline smoothing is performed to reduce sharp turns and ensure the physical feasibility and stability of UAV flight.

[0113] Furthermore, in step S5, the spoofing task point insertion and obfuscation process, a spoofing task budget is used. Insert a fake task point by following these steps:

[0114] 1) Location Selection: Location selection is performed based on a candidate point set C, which is a pre-defined database of alternative waypoints containing multiple geographical locations, used to provide potential false mission points. The system calculates the similarity between each point in this set and the real mission points in terms of spatial and mission attributes, and selects the point with the highest similarity that is located within a flyable area as the false mission point location.

[0115] 2) Determination of the number of fake missions: The number of fake missions is determined by the remaining budget and the remaining flight time of the UAV.

[0116]

[0117] in, Privacy budget required to generate a fake task point.

[0118] 3) Attribute confusion: Assigning fake task points the same task type, dwell time and operation mode as real tasks, so that external observers cannot distinguish between real and fake through behavior patterns.

[0119] 4) Path reconstruction: The task access order is adjusted by using local optimization algorithms to make false and real task points intertwined in time and space, thus preventing pattern recognition attacks.

[0120] Figure 3 The diagram illustrates the interleaved distribution of real and fake mission points along the UAV flight path after differential privacy perturbation. This design, by mixing fake mission points that are highly similar to the real mission in spatial location, mission type, and dwell time, effectively increases the difficulty for attackers to infer the location of the real mission, thus significantly enhancing path privacy protection.

[0121] Furthermore, in step S6, path optimization and reconstruction and multi-machine collaboration, after performing perturbation and spurious task insertion, a dual-objective optimization model for energy consumption and path fidelity is constructed:

[0122] .

[0123] in It is a path The total length, 'a' is the energy consumption weighting coefficient, and the final path is used. Compared to the original path The offset is measured by the squared Euclidean distance between them, and b is a fidelity weighting coefficient. The optimization algorithm uses mixed integer programming, genetic algorithms, or ant colony optimization to generate the optimal path. .

[0124] In multi-UAV scenarios, introduce task allocation variables. Let i represent the drone i executing task j. In the optimization model, task non-conflict constraints and communication synchronization delay constraints are added to achieve optimal scheduling for multi-drone collaboration.

[0125] This invention first introduces the limitations of current trajectory privacy protection methods, including insufficient consideration of semantic and spatiotemporal information, and the challenges in balancing privacy and utility. Then, it proposes a differential privacy-based generative adversarial network (GAN) method to resist semantic space reasoning attacks and achieve a balance between privacy protection level and quality of service.

[0126] Figure 4 This diagram illustrates the optimized flight path for drones under privacy and energy consumption constraints. It reflects the final execution route obtained through an optimization algorithm that comprehensively considers task time limits, energy consumption models, and path offsets after perturbation and spurious task point insertion. This path achieves a balance between privacy protection and flight energy consumption while maintaining high task completion efficiency, ensuring the feasibility and stability of the solution in real-world smart city drone applications.

[0127] To verify the effectiveness and superiority of the proposed method, a simulation platform for UAV path privacy protection and task execution was constructed. Under a unified task scenario, the performance of the proposed method and several common path privacy protection strategies were tested, with a focus on comparative analysis of four key metrics: inference accuracy, energy consumption, task completion rate, and path deviation. Experimental parameters included different privacy budgets ε ranging from 0.5 to 5, and the experiments were repeatedly run in environments with multiple task point distributions to ensure the stability and representativeness of the results.

[0128] Experimental results are as follows Figure 5As shown, the inference accuracy of this invention is consistently significantly lower than the comparison strategy, indicating that external attackers have the lowest success rate in inferring from path data, demonstrating the advantage of the two-layer privacy protection mechanism. Regarding energy consumption, the curve of this invention is the lowest overall, proving that while achieving strong privacy protection, it effectively suppresses additional energy consumption growth. This is achieved through sensitivity-driven budget allocation and path optimization reconstruction. Further observation of the task completion rate shows that this invention maintains a rate above 95% under different ε values, indicating that privacy protection does not come at the expense of task reliability. In terms of path deviation, the final flight path deviation obtained by this invention is the smallest, indicating that after reachability correction and smoothing, the path not only meets privacy requirements but also satisfies practical flight feasibility.

[0129] Experimental data show that the method of this invention outperforms existing methods in all four core metrics: it effectively improves path privacy protection, reduces energy consumption and path deviation, and ensures the reliability of task execution. This indicates that the two-layer mechanism combining differential privacy perturbation and spurious task point insertion, supplemented by a path adjustment strategy based on global optimization and reconstruction, proposed in this invention can achieve a balance between privacy protection and task efficiency in smart city UAV application scenarios, possessing high engineering application value and promotion potential.

[0130] In summary, the UAV path privacy protection and task optimization method based on differential privacy and spoof task point insertion proposed in this invention addresses the problems of insufficient privacy protection, poor path feasibility, and high energy consumption in existing technologies. It constructs a complete technical system from sensitivity analysis, dynamic budget allocation, two-layer privacy protection to global path optimization. This method can significantly improve the level of path privacy protection while ensuring task executability and controllable energy consumption, and effectively resist various trajectory inference and pattern recognition attacks.

[0131] The method proposed in this invention has good scalability and can be combined with multi-UAV collaborative scheduling, intelligent prediction models and multi-constraint real-time path optimization technology. It is applicable to various types of tasks such as package delivery, environmental monitoring, emergency rescue, and IoT data collection, providing strong technical support for the safe and efficient operation of UAVs in complex airspace in the future.

[0132] Future work on this invention can be considered to be carried out in the following aspects:

[0133] 1) Research on Multi-UAV Collaborative Privacy Protection Strategy: Based on the single-UAV privacy protection mechanism of this invention, this paper further explores privacy protection methods when multiple UAVs are collaboratively performing tasks. By designing a differential privacy budget allocation mechanism among multiple UAVs and a strategy for collaboratively generating fake task points, global privacy protection of multi-UAV paths is achieved, avoiding the leakage of sensitive information caused by the correlation of multi-UAV trajectories;

[0134] 2) Deep Learning-Based Adaptive Perturbation Optimization: Combining deep reinforcement learning, graph neural networks, and other intelligent algorithms, this method achieves real-time adaptive adjustment of perturbation amplitude, spurious task location, and quantity. It can dynamically optimize the privacy protection scheme based on environmental changes, threat level, and task priority, thereby significantly improving the system's flexibility and robustness.

[0135] 3) Path optimization model incorporating more real-world constraints: Dynamic factors such as weather changes, temporary airspace control, and real-time energy consumption feedback are introduced into the path optimization model to construct a multi-constraint optimization framework for practical application scenarios. This framework ensures that both privacy protection and task execution efficiency are met in complex and dynamic environments, enhancing the engineering feasibility of this invention.

[0136] Obviously, the specific implementation of this invention is not limited to the above-described methods. Any non-substantial improvements made using the inventive concept and technical solution of this invention are within the protection scope of this invention.

Claims

1. A method for protecting drone path privacy and optimizing missions, characterized in that: Includes the following steps: S1. Generate the initial path for the task; S2. Conduct path sensitivity calculation: Perform privacy sensitivity assessment on each segment of the initial path to quantify the potential privacy leakage risk during task execution; S3. Privacy Budget Allocation: Set an overall privacy protection budget and allocate the budget proportionally according to the sensitivity of each path segment to match the protection strength with the path risk; S4. Differential Privacy Perturbation: For path segments allocated to different privacy budgets, their geographical coordinates are randomly perturbed to generate perturbed paths; S5. Enhanced privacy through spoofing task points: In the perturbation path, several spoofing task points are dynamically inserted based on the remaining privacy budget and task time margin. S6. Path Optimization and Reconstruction: After the disturbance path and false mission point insertion are completed, a weighted optimization model that comprehensively considers flight energy consumption and path offset is constructed. Under the condition of meeting the mission time limit and energy consumption constraints, the path is optimized as a whole to generate the final flight path.

2. The method for protecting drone path privacy and optimizing tasks as described in claim 1, characterized in that: In step S1, the initial flight path of the UAV is generated using a path planning algorithm based on the set of target points allocated by the task scheduling center.

3. The method for protecting drone path privacy and optimizing missions as described in claim 1, characterized in that: Step S4 includes: S41. Select a perturbation mechanism: Use either the Laplace mechanism or the Gaussian mechanism to perturb the path coordinates; S42, Path Coordinate Noise Injection: Treat the starting and ending coordinates of the path segment as perturbation units, and add random noise that follows a Laplace distribution according to the allocated privacy budget. The noise amplitude is directly proportional to the sensitivity and inversely proportional to the budget. S43. Flight Accessibility Correction: If the coordinates fall into an unflyable area after the disturbance, the constrained projection method is used to adjust them to the nearest flyable position while preserving the random disturbance characteristics. S44. Maintain flight path continuity: After independently perturbing each segment of the path, smoothing is performed to obtain a smooth perturbed path trajectory.

4. The method for protecting drone path privacy and optimizing missions as described in claim 3, characterized in that: In step S41, the Laplace mechanism, which conforms to the definition of differential privacy, is used first to perturb the path coordinates; when the location sensitivity is extremely high or the perturbation requirement is small, the Gaussian mechanism is switched to perturb the path coordinates.

5. The method for protecting drone path privacy and optimizing missions as described in claim 1, characterized in that: Step S5 includes: selecting locations that are similar to the actual task points but unrelated to the task, based on the geometric shape of the disturbance path and the set of candidate stopping points, as false task points; Parameters for fake task points are set to be similar to those of real task points in terms of task type and dwell time, making it difficult for external observers to distinguish between real and fake tasks based on dwell time characteristics.

6. The method for protecting drone path privacy and optimizing missions as described in claim 5, characterized in that: The confirmation of the number of fake task points includes: dynamically determining the number of fake task points by combining the remaining privacy budget and the remaining task time.

7. The method for protecting drone path privacy and optimizing missions as described in claim 5, characterized in that: After inserting fake task points, the path is locally adjusted to make the real and fake task points interspersed and distributed, avoiding concentration at the end of the path or in specific areas, thus reducing the success rate of pattern recognition attacks.

8. The method for protecting drone path privacy and optimizing missions as described in claim 1, characterized in that: In step S3, the allocation of the privacy budget for the drone mission adopts a dynamic risk perception method, which automatically adjusts the privacy budget ratio of each mission segment based on the importance of the mission, the execution progress, and the drone status.

9. The method for protecting drone path privacy and optimizing missions as described in claim 8, characterized in that: The allocation of the privacy budget in step S3 includes: 1) Budget Vector Initialization: The system first establishes the privacy budget vector for all task points in the task set M. Initially, all values ​​are set to zero. 2) Task risk calculation and correction: For each task point in set M... The base risk value is calculated based on the sensitivity of the mission location, the mission type, and information that external attackers may exploit. According to the planned execution ratio of the task. and the system-defined time decay coefficient The correction factor is obtained. When the task has not yet started or is at a low progress, the correction factor is increased to allocate more budget. The base risk value is multiplied by the correction factor to obtain the comprehensive risk weight. 3) The system will allocate a total privacy budget. The privacy budget is allocated proportionally according to the comprehensive risk weight. The higher the weight of the task segment, the more privacy budget it receives, thereby introducing stronger protection in differential privacy perturbation. 4) All task points are assigned a comprehensive risk weight. Sort tasks from largest to smallest to generate a high-risk task priority queue. In the subsequent path disturbance and false task insertion process, high-risk task points should be protected first; 5) The system monitors the remaining energy of the drone in real time. Remaining task time Status parameters: When a status change is detected to exceed a set threshold, steps 2–4 are re-executed to ensure that the budget allocation is consistent with the current task environment. 6) Finally, the system outputs the allocated privacy budget vector. and high-risk task priority queue .

10. The method for protecting drone path privacy and optimizing missions as described in claim 5, characterized in that: In step S2, path sensitivity analysis, a path query function f is defined, whose input is the complete UAV task path and whose output is a feature vector. These correspond to the total path distance, total flight time, and total energy consumption, respectively. The sensitivity calculation process fixes all path segments except for the i-th segment, replaces the i-th segment with a neighboring feasible path segment, obtains the characteristic change amount through flight simulation, and calculates the comprehensive sensitivity: ； in, , , For fixed weighting coefficients, and satisfying , , , These represent the changes in total mission distance, total flight time, and total energy consumption, respectively.

Images