A privacy protection task allocation method based on beaver triple
By combining Beaver triplet and secret sharing techniques with the NSGA-II genetic algorithm, the problem of high computational and communication overhead in privacy-preserving task allocation under multi-objective and multi-task scenarios is solved, achieving effective privacy protection and efficient task allocation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SOUTH CHINA UNIV OF TECH
- Filing Date
- 2026-03-04
- Publication Date
- 2026-06-05
Smart Images
Figure CN122153955A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the technical field of information security and privacy protection, and in particular to a privacy protection task allocation method based on Beaver triples. Background Technology
[0002] In recent years, with the development of network communication technology and the popularization of smart terminal devices, the mobile Internet and the sharing economy have developed rapidly. Traditional network-based crowdsourcing models have further evolved into new data collection and service collaboration models such as Spatial Crowdsourcing (SC) and Mobile Crowdsensing (MCS). A typical SC / MCS system generally includes three main entities: task requesters, workers (participants), and a platform server. Task requesters publish tasks through the platform, the platform selects suitable workers to participate and execute the tasks based on the task requirements, and after completing the tasks, the workers upload the results to the platform, which then aggregates and provides feedback to the task requesters. Because this model is widely used in scenarios such as food delivery, ride-hailing, geographic data sharing, environmental monitoring, and intelligent transportation, spatiotemporally sensitive data such as task and worker location information, movement trajectories, and preferences are frequently collected and processed during system operation. Without effective protection, this can easily lead to problems such as worker privacy leaks, decreased user willingness, and economic losses for task requesters. Therefore, achieving privacy protection while ensuring service quality has become an important research direction in the SC / MCS field.
[0003] In practical applications, SC / MCS increasingly presents complex scenarios involving concurrent multi-tasks and multiple requesters. Task allocation can generally be divided into offline allocation and online allocation: offline allocation performs global optimization under the condition of known task and worker location / trajectory information, while online allocation requires real-time collection of location information and dynamic decision-making. Meanwhile, researchers are also focusing on multi-task scenarios, such as multi-task allocation, multi-task data aggregation, and multi-task incentive mechanisms. Based on the number of requesters and tasks, existing data allocation methods in MCS can be divided into three categories: single-task allocation (one task and multiple task workers) and multi-task allocation (multiple tasks and multiple task workers). In recent years, esoteric techniques such as differential privacy and homomorphic encryption have been widely used for privacy protection in task allocation in swarm intelligence sensing scenarios. Differential privacy technology adds noise to sensitive data to achieve data desensitization, while homomorphic encryption uses its homomorphic computation property to perform computation on encrypted data, avoiding direct exposure of data to the sensing platform and achieving data privacy protection.
[0004] However, current differential privacy schemes suffer from reduced effectiveness of task allocation results, while homomorphic encryption suffers from significant computational and communication overhead. Furthermore, current privacy-preserving task allocation schemes primarily focus on single-objective task allocation, with limited research on privacy protection in multi-objective, multi-task allocation scenarios. Summary of the Invention
[0005] To address the shortcomings and deficiencies of existing task allocation schemes in multi-objective, multi-task scenarios, this invention proposes a privacy-preserving task allocation method based on Beaver triples. This method protects privacy through secret sharing and performs calculations on privacy data using Beaver triples. It can calculate the encrypted data uploaded by the task requester (TR) and worker (TP) under privacy protection, including the distance between the worker and the task, the effective time for the worker to execute the task, and the matching value between the worker and the task. Furthermore, through a designed task allocation algorithm, it achieves a relatively optimal task allocation result under constraints in multi-objective, multi-task allocation scenarios.
[0006] To achieve the above technical solution, this invention provides a privacy-preserving task allocation method based on Beaver triples, specifically including the following steps:
[0007] S1. System Initialization: Performed by the sensing platform server. Generate and publish public parameters for two independent servers. and The registration system generates Beaver triples using Paillier homomorphic encryption.
[0008] S2, Task Issuance: The requester sends the sensing task share to two servers of the sensing platform through secret sharing, and the platform leads the workers to select and allocate tasks.
[0009] S3. Worker uploads information: Candidate workers secretly share their personal information and send it in shares to two servers of the sensing platform.
[0010] S4. Data computation under privacy protection: The two servers implement a secure computation protocol with the assistance of Beaver triples to calculate the worker-task matching degree, the average cost of the worker to complete the task and the distance between the worker and the task in a confidential state, as well as calculate the effective execution time of the worker to complete the task.
[0011] S5. Multi-objective multi-task allocation: The perception platform calculates the average cost of workers completing tasks, the matching degree between workers and tasks, and the effective time for workers to execute tasks. Through a safe comparison protocol and the NSGA-II genetic algorithm, the task allocation result aims to minimize the total cost of workers executing tasks and maximize the value created by workers.
[0012] S6. Receive and execute tasks: The server obtains the task allocation results, and the two servers send their task shares to the workers who were assigned the tasks. The workers use the two task shares they receive to recover the task information and execute the tasks.
[0013] In step S1, the detailed process of system initialization is as follows:
[0014] S11, Perception Platform Initialization: First, the sensing platform specifies a security bit depth of 2048 and randomly generates large prime numbers p and q, satisfying ||p|| = ||q|| = 2048. Then, calculate n = p · q. =lcm(p-1,q-1), where lcm is the least common multiple of the two parameters. Choose g=N+1, have the public key pk=(n,g) and send it to... Define a function and calculate mod n, private key sk = ( , ).
[0015] S12. Generate Beaver triples based on Paillier homomorphic encryption: generate , , generate , , encryption , get , And send to , Generate random number r and calculate Send d to , calculate = +Dec(d), calculate = .satisfy This refers to the definition of Beaver. The above operations can generate a large number of Beaver triples.
[0016] In step S2, the task publishing process is as follows:
[0017] Assuming all requesters A task request was sent to the perception platform, specifying the DID as... The requester's task request is denoted as ,in These are the latitude and longitude coordinates of the mission execution location. It is the end time of the task. It's an attribute of the task. This is the task's budget. The task requester splits their task into two shares via a secret sharing mechanism. and share and will and Send to each and .
[0018] In step S3, the detailed process of the worker uploading information is as follows:
[0019] Assuming all workers A task request was sent to the perception platform, specifying the DID as... The worker's task request is recorded as ,in It is the latitude and longitude of the worker's location. It is the speed at which the workers move. It is the payment a worker receives for completing a task. It is the cost for a worker to complete a task. It's the worker's ability. The worker splits their task into two parts through secret sharing. and share and will and Send to each and .
[0020] In step S4, the data computation process under privacy protection is as follows:
[0021] S41. Calculation of Manhattan distance in dense state, for two locations. , and , ,server and Each holding shares and ,in , , = , = . and The secure comparison protocol SecCmp, constructed using Beaver triples, is used to calculate SecCmp( ) and SecCmp( ) get , and , Where A(,) represents the 0 / 1 bits of the comparison result. Then, each of the two servers subtracts the share locally to obtain the secret share of the difference. , and , ,Then and Construct separately , and , And calculate SecMul using the secure multiplication protocol. ) and SecMul ) respectively obtained , and , .at last and Calculate separately and This refers to the secret shared share of Manhattan distance.
[0022] S42, Secure Division Protocol in Compact State, Server and Each holding shares and Workers tp hold y, of which , , = , = .first Randomly select a sufficiently large random number r and compute it locally. and send Put the server on . Local computation And obtained And send s to tp, tp calculates And send q to , calculate get .
[0023] In step S5, the multi-task allocation process is as follows:
[0024] S51. Multi-objective, multi-task allocation: We define the task allocation objectives and constraints as follows, with the optimization objective being... and in For the matching degree between workers and tasks, The time spent by workers actually performing tasks. The two optimization objectives are to minimize the total cost to workers performing tasks and to maximize the value created by workers performing tasks. We also impose constraints. , , , The above constraints respectively ensure that the actual time the worker spends performing the task is greater than 0. This indicates whether a worker has been assigned, and whether the worker's compensation is less than the task budget and whether the task and worker matching value is greater than or equal to 1. Finally, we constrain a worker to accept only one task.
[0025] S52. Calculation of Key Parameters for Task Allocation under Privacy Protection: To achieve the aforementioned optimization objective function, we need to calculate the average cost for a worker to complete a task, the worker-task matching value, and the actual execution time of the worker on the task, without disclosing worker and task information. For the average cost for a worker to complete a task, we leverage the property of additive secret sharing... and Calculate separately and , Will Send to This yields c. For the worker-task matching value, we use a safe multiplication protocol based on Beaver triples to perform a dot product multiplication of the 0 / 1 values to form the worker capability vector and the task attribute vector, thus obtaining c. and , accept Sent Calculations yielded We need to calculate the time it takes for workers to complete their tasks. We can calculate the distance based on the distance calculation module in step 4), perform division using the division module, and finally obtain the result by secret-shared subtraction. .
[0026] S53. Bi-objective multi-task allocation based on NSGA-II genetic algorithm: Since each worker can be assigned at most one task, a chromosome is used to represent the task assigned to the worker. A value of -1 indicates that no task has been assigned, and other values indicate that the worker has been assigned to that task. First, an initial population is generated and initialized randomly. After initialization, the chromosome is constrained and corrected. After obtaining the initial population, multiple non-dominated fronts are calculated, and individuals at each level are sorted by crowding distance. Then, a binary tournament selection is performed based on the front level and crowding distance to generate parent individuals, and offspring are generated through crossover and mutation. Constraints are used to ensure that the offspring are feasible solutions. Through multiple iterations, a set of Pareto solutions is finally obtained. The points on the Pareto front are connected in the normalized objective space, and the point with the largest perpendicular distance to the endpoint is selected as the compromise solution and the result of task allocation.
[0027] In step S6, the detailed process of receiving and executing the task is as follows:
[0028] S61, Obtain the task allocation result in step 5) and send the task result to the server through a secure channel. .
[0029] S62, and The information share of the task is sent to the worker assigned to the task via a secure channel. and share .
[0030] S63. Workers receive task information and reconstruct the task information via share. It then executes the task, and upon completion, sends the task execution report to the perception platform server. .
[0031] Compared with the prior art, the present invention has the following advantages and beneficial effects:
[0032] 1. This invention achieves multi-task data cryptographic operations by using secret sharing and Beaver, enabling homomorphic computation in an encrypted state. Compared with existing homomorphic encryption methods using Paillier, it has lower computational overhead and communication requirements.
[0033] 2. This invention ensures the privacy of each worker and requester. Through secret sharing, it protects the privacy of task requesters and workers, ensuring that the platform cannot obtain their personal information.
[0034] 3. This invention achieves dual-objective, multi-task allocation. It ensures the validity of task allocation results while protecting privacy, and keeps computational overhead within an acceptable range, enabling more efficient task allocation. Attached Figure Description
[0035] Figure 1 This is a conceptual model diagram of the multi-task approach of the present invention.
[0036] Figure 2 This is a schematic diagram of the system model of the present invention.
[0037] Figure 3 This is a flowchart of the process of the present invention.
[0038] Figure 4 This is a schematic diagram of the system operation of the present invention. Detailed Implementation
[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art without creative effort are within the protection scope of the present invention.
[0040] Example: A privacy-preserving task allocation method based on Beaver triples.
[0041] Terminology Explanation: Secret Sharing: Secret sharing is a cryptographic technique that splits sensitive data into multiple "shares" and distributes them to multiple participants for safekeeping. A single participant holding only one share cannot recover the original data; the original data can only be reconstructed when a preset threshold number of shares are combined. Secret sharing is commonly used in scenarios such as secure multi-party computation, enabling multiple participants to complete collaborative computations without revealing their individual private inputs.
[0042] Beaver Triples: Beaver triples are a type of preprocessed random data used in secure multi-party computation to efficiently perform multiplication operations. They are typically represented as a set of three random numbers (or their shared form) (a, b, c), where c = a ⋅ b. Participants can pre-generate and distribute a secret shared share of this triple during the offline phase. During the online computation phase, this triple is used to transform "multiplication" into a small number of additions and a single interaction, thereby completing the product calculation without revealing the multiplication input, improving computational efficiency and reducing communication overhead.
[0043] Homomorphic encryption is an encryption technique that allows specific operations to be performed directly on ciphertext. The new ciphertext obtained after performing the operation on the ciphertext will have the same decryption result as the plaintext. With homomorphic encryption, data providers can provide ciphertext to a third party or computational entity for operations such as addition and multiplication without disclosing the plaintext data, thus achieving verifiable or usable encrypted computation capabilities while protecting data privacy.
[0044] This example provides a privacy-preserving task allocation method based on Beaver triples. By secretly sharing task and worker information, it achieves data security on the perception platform. It also uses Beaver triples to assist in homomorphic computation in the encrypted state and assigns tasks to appropriate workers through a designed task allocation algorithm.
[0045] like Figure 1 As shown, based on the number of requesters and tasks, existing task allocation methods in MCS can be divided into two categories: single-task allocation (one task and multiple TPs) and multi-task allocation (multiple tasks and multiple TPs). This invention implements the second category: privacy-preserving multi-task allocation in MCS.
[0046] like Figure 2 As shown, the system model of this invention includes the following entities: task requester, sensing platform, and worker. The following is a description of each entity:
[0047] Requesters: Composed of multiple requesters, who are the initiators of multiple tasks. They have task awareness needs in different regions or time periods, therefore they need to send task requests to the awareness platform. In this invention, each requester first allocates their task to two shares, and then sends them to two servers of the awareness platform respectively.
[0048] Perception Platform: The perception platform is responsible for receiving tasks from requesters and distributing them to workers. Based on the information from both the requester and the worker, the perception platform uses a secure computation protocol and task allocation algorithm to obtain the task allocation result, and then sends the task information to the corresponding worker according to the allocation result.
[0049] Workers: Responsible for completing multiple tasks issued by the sensing platform. Workers secretly share and upload information to the server, obtain task information from the server, and then execute the tasks based on the task information.
[0050] Figure 3 This briefly demonstrates the workflow of a privacy-preserving multi-task data allocation method. Figure 4 This example briefly summarizes the components used in the demonstration. The aforementioned privacy-preserving task allocation method based on Beaver triples includes the following steps:
[0051] 1) System initialization: performed by the sensing platform server. Generate and publish public parameters for two independent servers. and The system is registered and Beaver triples are generated using Paillier homomorphic encryption. The detailed system initialization process is as follows:
[0052] 1.1) Perception Platform Initialization: First, the perception platform specifies 2048 security bits and randomly generates large prime numbers p and q, satisfying ||p|| = ||q|| = 2048. Then, calculate n = p · q. =lcm(p-1,q-1), where lcm is the least common multiple of the two parameters. Choose g=N+1, have the public key pk=(n,g) and send it to... Define a function and calculate mod n, private key sk = ( , ).
[0053] 1.2) Generating Beaver triples based on Paillier homomorphic encryption: generate , , generate , , encryption , get , And send to , Generate random number r and calculate Send d to , calculate = +Dec(d), calculate = .satisfy This refers to the definition of Beaver. The above operations can generate a large number of Beaver triples.
[0054] 2) Task Issuance: Requesters send their sensing task shares to two servers on the sensing platform via a secret sharing mechanism. The platform then leads the worker selection and task allocation. The task issuance process is as follows:
[0055] Assuming all requesters A task request was sent to the perception platform, specifying the DID as... The requester's task request is denoted as ,in These are the latitude and longitude coordinates of the mission execution location. It is the end time of the task. It's an attribute of the task. This is the task's budget. The task requester splits their task into two shares via a secret sharing mechanism. and share and will and Send to each and .
[0056] 3) Worker Information Upload: Candidate workers secretly share their personal information in portions to two servers of the sensing platform. The detailed process of worker information upload in step 3) is as follows:
[0057] Assuming all workers A task request was sent to the perception platform, specifying the DID as... The worker's task request is recorded as ,in It is the latitude and longitude of the worker's location. It is the speed at which the workers move. It is the payment a worker receives for completing a task. It is the cost for a worker to complete a task. It's the worker's ability. The worker splits their task into two parts through secret sharing. and share and will and Send to each and .
[0058] 4) Privacy-Preserving Data Computation: A secure computation protocol, implemented by two servers using Beaver triples, is used to calculate, in encrypted form, the matching degree between workers and tasks, the average cost for workers to complete tasks, the distance between workers and tasks, and the effective execution time for workers to complete tasks. The detailed process of privacy-preserving data computation is as follows:
[0059] 4.1) Calculation of Manhattan distance in dense state, for two locations , and , ,server and Each holding shares and ,in , , = , = . and The secure comparison protocol SecCmp, constructed using Beaver triples, is used to calculate SecCmp( ) and SecCmp( ) get , and , Where A(,) represents the 0 / 1 bits of the comparison result. Then, each of the two servers subtracts the share locally to obtain the secret share of the difference. , and , ,Then and Construct separately , and , And calculate SecMul using the secure multiplication protocol. ) and SecMul ) respectively obtained , and , .at last and Calculate separately and This refers to the secret shared share of Manhattan distance.
[0060] 4.2) Secure division protocol in encrypted state, server and Each holding shares and Workers tp hold y, of which , , = , = .first Randomly select a sufficiently large random number r and compute it locally. and send Put the server on . Local computation And obtained And send s to tp, tp calculates And send q to , calculate get .
[0061] 5) Multi-objective, multi-task allocation: The perception platform calculates the average cost of workers completing tasks, the worker-task matching degree, and the effective time for workers to execute tasks. Through a safe comparison protocol and the NSGA-II genetic algorithm, the task allocation result aims to minimize the total cost of workers executing tasks and maximize the value created by workers. In step 5), the detailed process of multi-objective, multi-task allocation is as follows:
[0062] 5.1) Multi-objective, multi-task allocation: We define the task allocation objectives and constraints as follows, with the optimization objective being... and in For the matching degree between workers and tasks, The time spent by workers actually performing tasks. The two optimization objectives are to minimize the total cost to workers performing tasks and to maximize the value created by workers performing tasks. We also impose constraints. , , , The above constraints respectively ensure that the actual time the worker spends performing the task is greater than 0. This indicates whether a worker has been assigned, and whether the worker's compensation is less than the task budget and whether the task and worker matching value is greater than or equal to 1. Finally, we constrain a worker to accept only one task.
[0063] 5.2) Calculating Key Parameters for Task Allocation Under Privacy Protection: To achieve the aforementioned optimization objective function, we need to calculate the average cost for a worker to complete a task, the worker-task matching value, and the worker's actual execution time for the task, without disclosing worker and task information. For the average cost for a worker to complete a task, we leverage the property of additive secret sharing... and Calculate separately and , Will Send to This yields c. For the worker-task matching value, we use a safe multiplication protocol based on Beaver triples to perform a dot product multiplication of the 0 / 1 values to form the worker capability vector and the task attribute vector, thus obtaining c. and , accept Sent Calculations yielded We need to calculate the time it takes for workers to complete their tasks. We can calculate the distance based on the distance calculation module in step 4), perform division using the division module, and finally obtain the result by secret-shared subtraction. .
[0064] 5.3) Bi-objective multi-task allocation based on NSGA-II genetic algorithm: Since each worker can be assigned at most one task, a chromosome is used to represent the task assigned to the worker. A value of -1 indicates that no task has been assigned, and other values indicate that the worker has been assigned to that task. First, an initial population is generated and initialized randomly. After initialization, the chromosome is constrained and corrected. After obtaining the initial population, multiple sets of non-dominated fronts need to be calculated, and individuals at each level are sorted by crowding distance. Then, a binary tournament selection is performed based on the front level and crowding distance to generate parent individuals, and offspring are generated through crossover and mutation. Constraints and corrections are used to ensure that the offspring are feasible solutions. Through multiple rounds of iteration, a set of Pareto solutions can be obtained. The points on the Pareto front are connected in the normalized objective space, and the point with the largest perpendicular distance to the endpoint is selected as the compromise solution and the result of task allocation.
[0065] 6) Receiving and Executing Tasks: The server receives the task allocation results, and both servers send their task shares to the workers assigned to those tasks. The workers then use the two task shares they receive to recover the task information and execute the task. The detailed process of receiving and executing tasks is as follows:
[0066] 6.1) Obtain the task allocation result in step 5) and send the task result to the server through a secure channel. .
[0067] 6.2) and The information share of the task is sent to the worker assigned to the task via a secure channel. and share .
[0068] 6.2) Workers receive task information and reconstruct the task information via share. It then executes the task, and upon completion, sends the task execution report to the perception platform server. .
[0069] Based on this invention, users can ensure the security of data on the crowd-sensing task allocation platform, guaranteeing that the crowd-sensing server will not obtain users' sensitive information. This protects the privacy and security of task submitters and workers.
[0070] This invention, while considering privacy protection in task allocation, achieves a more efficient task allocation scheme. Users can ensure data security through secret sharing, preventing malicious users from obtaining their data. A secure computation module built using Beaver triples enables data computation under privacy protection, offering advantages over other homomorphic encryption schemes in terms of lower computational and communication overhead. Furthermore, the NSGA-II algorithm enables efficient allocation of dual-objective optimization tasks.
[0071] The above description is only a preferred embodiment of the present invention, but the present invention should not be limited to the content disclosed in the embodiments and drawings. Therefore, any equivalent or modified embodiments made without departing from the spirit of the present invention shall fall within the protection scope of the present invention.
Claims
1. A privacy-preserving task allocation method based on Beaver triples, characterized in that, Includes the following steps: S1. System Initialization: Performed by the sensing platform server. Generate and publish public parameters for two independent servers. and The registration system generates Beaver triples using Paillier homomorphic encryption. S2, Task Issuance: The requester sends the sensing task to two servers of the sensing platform in the form of shares through secret sharing. The platform leads the selection of workers and task allocation. S3. Worker uploads information: Candidate workers secretly share their personal information and send it in shares to two servers of the sensing platform. S4. Data computation under privacy protection: The two servers implement a secure computation protocol with the assistance of Beaver triples to calculate the worker-task matching degree, the average cost of the worker to complete the task and the distance between the worker and the task in a confidential state, as well as calculate the effective execution time of the worker to complete the task. S5. Multi-objective multi-task allocation: The perception platform calculates the average cost of workers completing tasks, the matching degree between workers and tasks, and the effective time for workers to execute tasks. Through a safe comparison protocol and the NSGA-II genetic algorithm, the task allocation result aims to minimize the total cost of workers executing tasks and maximize the value created by workers. S6. Receive and execute tasks: The server obtains the task allocation results, and the two servers send their task shares to the workers who were assigned the tasks. The workers use the two task shares they receive to recover the task information and execute the tasks.
2. The privacy-preserving task allocation method based on Beaver triples according to claim 1, characterized in that: In step S1, the detailed process of system initialization is as follows: S11, Perception Platform Initialization: First, the sensing platform specifies a security bit depth of 2048 and randomly generates large prime numbers p and q, satisfying ||p|| = ||q|| = 2048. Then, calculate n = p · q. =lcm(p-1,q-1), where lcm is the least common multiple of the two parameters. Choose g=N+1, have the public key pk=(n,g) and send it to... Define a function and calculate mod n, private key sk = ( , ). S12. Generate Beaver triples based on Paillier homomorphic encryption: generate , , generate , , encryption , get , And send to , Generate random number r and calculate Send d to , calculate = +Dec(d), calculate = .satisfy This refers to the definition of Beaver. The above operations can generate a large number of Beaver triples.
3. The privacy-preserving task allocation method based on Beaver triples according to claim 1, characterized in that: In step S2, the task publishing process is as follows: Assume all requesters A task request was sent to the perception platform, specifying the DID as... The requester's task request is denoted as ,in These are the latitude and longitude coordinates of the mission execution location. It is the end time of the task. It's an attribute of the task. This is the task's budget. The task requester splits their task into two shares via a secret sharing mechanism. and share and will and Send to each and .
4. The privacy-preserving task allocation method based on Beaver triples according to claim 1, characterized in that: In step S3, the task posting process is as follows: Assume all workers A task request was sent to the perception platform, specifying the DID as... The worker's task request is recorded as ,in It is the latitude and longitude of the worker's location. It is the speed at which the workers move. It is the reward a worker receives for completing a task. It is the cost for a worker to complete a task. It's the worker's ability. The worker splits their task into two parts through secret sharing. and share and will and Send to each and .
5. The privacy-preserving task allocation method based on Beaver triples according to claim 1, characterized in that: In step S4, the data computation process under privacy protection is as follows: S41. Calculation of Manhattan distance in dense state, for two locations. , and , ,server and Each holding shares and ,in , , = , = . and The secure comparison protocol SecCmp, constructed using Beaver triples, is used to calculate SecCmp( ) and SecCmp( ) get , and , Where A(,) represents the 0 / 1 bits of the comparison result. Then, each of the two servers subtracts the share locally to obtain the secret share of the difference. , and , ,Then and Construct separately , and , And calculate SecMul using the secure multiplication protocol. ) and SecMul ) respectively obtained , and , .at last and Calculate separately and This refers to the secret shared share of Manhattan distance. S42, Secure Division Protocol in Compact State, Server and Each holding shares and Workers tp hold y, of which , , = , = .first Randomly select a sufficiently large random number r and compute it locally. and send Put the server on . Local computation And obtained And send s to tp, tp calculates And send q to , calculate get .
6. The privacy-preserving task allocation method based on Beaver triples according to claim 1, characterized in that: In step S5, the multi-task allocation process is as follows: S51. Multi-objective, multi-task allocation: We define the task allocation objectives and constraints as follows, with the optimization objective being... and in For the matching degree between workers and tasks, The time spent by workers actually performing tasks. The two optimization objectives are to minimize the total cost to workers performing tasks and to maximize the value created by workers performing tasks. We also impose constraints. , , , The above constraints respectively ensure that the actual time the worker spends performing the task is greater than 0. This indicates whether a worker has been assigned a task, and whether the worker's compensation is less than the task budget and the task-worker matching value is greater than or equal to 1. In this task assignment scenario, we stipulate that a worker can only accept one task. S52. Calculation of Key Parameters for Task Allocation under Privacy Protection: To achieve the aforementioned optimization objective function, we need to calculate the average cost for a worker to complete a task, the worker-task matching value, and the actual execution time of the worker on the task, without disclosing worker and task information. For the average cost for a worker to complete a task, we leverage the property of additive secret sharing... and Calculate separately and , Will Send to This yields c. For the worker-task matching value, we use a safe multiplication protocol based on Beaver triples to perform a dot product multiplication of the 0 / 1 values to form the worker capability vector and the task attribute vector, thus obtaining c. and , accept Sent Calculations yielded We need to calculate the time it takes for workers to complete their tasks. We can calculate the distance based on the distance calculation module in step 4), perform division using the division module, and finally obtain the result by secret-shared subtraction. . S53. Bi-objective multi-task allocation based on NSGA-II genetic algorithm: Since each worker can be assigned at most one task, a chromosome is used to represent the task assigned to the worker. A value of -1 indicates that no task has been assigned, and other values indicate that the worker has been assigned to that task. First, an initial population is generated and initialized randomly. After initialization, the chromosome is constrained and corrected. After obtaining the initial population, multiple non-dominated fronts are calculated, and individuals at each level are sorted by crowding distance. Then, a binary tournament selection is performed based on the front level and crowding distance to generate parent individuals, and offspring are generated through crossover and mutation. Constraints are used to ensure that the offspring are feasible solutions. Through multiple iterations, a set of Pareto solutions is finally obtained. The points on the Pareto front are connected in the normalized objective space, and the point with the largest perpendicular distance to the endpoint is selected as the compromise solution and the result of task allocation.
7. The privacy-preserving task allocation method based on Beaver triples according to claim 1, characterized in that... In step S6, the detailed process of receiving and executing the task is as follows: S61, Obtain the task allocation result in step 5) and send the task result to the server through a secure channel. . S62, and The information share of the task is sent to the worker assigned to the task via a secure channel. and share . S63. Workers receive task information and reconstruct the task information via share. It then executes the task, and upon completion, sends the task execution report to the perception platform server. .