A worker selection method and system based on privacy protection in multi-task scenarios

By considering the benefits of workers' collaboration and privacy sensitivity in the task allocation stage, decompose tasks and use edge computing and genetic-taboo search algorithms, the problem of insufficient privacy protection and task decomposition in traditional task allocation methods is solved, and the multi-task scenario is realized efficient and accurate task allocation.

CN118660289BActive Publication Date: 2024-11-01YANTAI UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411131725.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-08-19
Publication Date
2024-11-01
Estimated Expiration
2044-08-19

AI Technical Summary

Technical Problem

Traditional task allocation methods ignore workers' privacy protection needs in the multi-task scenarios that worker collaboratively complete multi-task scenarios, and cannot protect workers' task competition information and perceived data at the same time. In addition, the calculation of task decomposition and privacy sensitivity lead to inaccurate and efficient task allocation. Low.

Method used

By taking into account the worker's collaboration benefits, capability benefits, and privacy sensitivity during the task allocation phase, using edge computing and encryption techniques to decompose tasks into subtasks and selecting the best collaborative worker group using a genetic-taboo search algorithm with backpack constraints, Ensure privacy protection and task completion benefits.

Benefits of technology

It realizes the accurate completion of task allocation while protecting workers' privacy, improves system execution efficiency, and ensures the quality of uploaded data and task completion benefits.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118660289B_ABST
    Figure CN118660289B_ABST
Patent Text Reader

Abstract

The present invention relates to the cross-technical field of crowd intelligence perception and artificial intelligence, and in particular to a worker selection method and system based on privacy protection in a multi-task scenario. It includes obtaining worker information and platform authentication keys; calculating the collaborative benefits between workers based on the worker information and platform authentication keys; obtaining task information, and decomposing the task information to obtain subtasks; calculating the worker's ability benefits for completing subtasks, and encrypting the ability benefits by giving the workers a false identity and encryption key; determining the candidate worker group for each subtask based on the ability benefits and collaborative benefits, and calculating the worker's task completion benefits for each subtask; under the subtask budget constraint, selecting the best collaborative worker group for each subtask based on the worker's task completion benefits. This scheme can not only protect the worker's task competition information and perception data privacy at the same time, but also ensure the accuracy of task allocation and the quality of uploaded data.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the interdisciplinary technical field of crowd intelligence perception and artificial intelligence, and in particular to a worker selection method and system based on privacy protection in a multi-task scenario. Background Art

[0002] In recent years, with the widespread deployment of 5G mobile networks, the popularity of mobile devices, and the increasing abundance of various sensors (such as temperature sensors, location sensors, etc.), mobile crowd sensing (MCS) technology has developed rapidly. MCS not only makes full use of the power and wisdom of the group to efficiently solve problems, but also becomes a powerful tool to meet the needs of large-scale data collection and processing. It has shown significant application value in many fields, including real-time collection of navigation data, accurate monitoring of traffic flow, and efficient labeling of artificial intelligence data sets.

[0003] With the exponential growth of the number of mobile devices, more and more people are participating in the crowdsourcing process. These crowdsourcing participants participate in the crowdsourcing process in different regions and at different times. As the number of task executors and the coverage of crowdsourcing tasks continues to expand, the scope of crowdsourcing services is also expanding. Therefore, the research on mobile crowdsourcing still has the following problems: (1) Traditional task allocation methods either focus on the scenario of a single worker completing a single task or on the scenario of multiple workers collaborating to complete a single task, ignoring the situation of multiple workers collaborating to complete multiple tasks. (2) Traditional task allocation methods often ignore the impact of workers' privacy protection needs on the benefits of workers' task completion when calculating the benefits of workers for tasks, and usually only focus on the allocation of a single complete task without considering the decomposability of tasks. (3) Traditional privacy protection schemes either only protect the perception data uploaded by workers or only protect the location information of users during task allocation. They cannot protect workers' task competition information and perception data at the same time, and ignore other task competition information submitted by workers during task allocation. Therefore, how to recruit a group of suitable collaborative workers for multiple tasks while considering the task decomposition and the privacy sensitivity of workers to different sub-tasks, and how to accurately complete the task allocation process while fully protecting the workers' privacy (including task competition information and perception data information) and ensuring the quality of uploaded perception data are one of the main problems that mobile crowdsourcing urgently needs to solve. Summary of the invention

[0004] In order to solve the above-mentioned problems, the present invention provides a worker selection method and system based on privacy protection in a multi-task scenario. By fully considering the collaborative benefits, ability benefits and privacy sensitivity weakening effects of workers in the task allocation stage, suitable collaborative workers are selected to complete all tasks under the constraints of the task budget. At the same time, while effectively protecting the workers' task competition privacy information and perception data privacy information, the accuracy and effectiveness of task allocation can be guaranteed and the quality of submitted perception data can be guaranteed. The execution efficiency of the entire system is also improved by decomposing and unloading computing tasks and introducing edge computing technology, alleviating the data processing pressure on the cloud platform and local workers.

[0005] In a first aspect, the present invention provides a worker selection method based on privacy protection in a multi-task scenario, which adopts the following technical solution:

[0006] A privacy-preserving worker selection method in a multi-task scenario includes:

[0007] Obtain worker information and platform authentication keys;

[0008] Calculate the collaboration benefits between workers based on worker information and platform authentication keys;

[0009] Obtain task information and decompose the task information into subtasks;

[0010] Calculate the worker's ability benefit for completing the subtask, and encrypt the ability benefit by giving the worker a false identity and encryption key;

[0011] Determine the candidate worker group for each subtask based on the capability benefit and collaboration benefit, and calculate the worker's task completion benefit for each subtask;

[0012] Under the subtask budget constraint, the best group of collaborative workers is selected for each subtask based on the workers' task completion benefits.

[0013] Furthermore, the collaboration benefit includes preference similarity benefit and social trust benefit, wherein the social trust benefit between workers is calculated based on the social information value and social trust value between workers and other workers; and the preference similarity benefit between workers is calculated by the preference similarity value between workers and other workers.

[0014] Furthermore, decomposing the task information to obtain subtasks includes dividing the task information into a number of subtasks according to the location, bid and type of the task, wherein the worker chooses to complete a subtask from the same or different task instead of completing the entire task.

[0015] Furthermore, the calculating of the worker's ability benefit for completing the subtask includes calculating the worker's distance benefit, experience benefit, and privacy protection weakening benefit for completing the subtask based on the worker information.

[0016] Furthermore, the calculation of the worker's task completion benefit for each subtask includes calculating the worker's task completion benefit for each subtask based on the collaboration benefit and capability benefit of the workers in the candidate worker group as the service quality of the worker completing the subtask.

[0017] Furthermore, selecting the best collaborative worker group for each subtask according to the workers' task completion benefits includes solving the optimal candidate worker set for each subtask using a genetic-tabu search algorithm with knapsack constraints according to the workers' task completion benefits.

[0018] Furthermore, the capability benefits are encrypted by assigning workers false identities and encryption keys, including generating a shared key based on a group key exchange protocol, each worker encrypts his or her own historical task information using the shared key and the AES encryption algorithm, and further encrypts the encrypted historical task information using the SHA256 algorithm to obtain a double-encrypted historical task information ciphertext, and finally uses a homomorphic encryption key to match a false identity for each worker, and uses the worker's false identity to encrypt and send the worker's benefit information.

[0019] The second aspect is a worker selection system based on privacy protection in a multi-task scenario, including:

[0020] The data acquisition module is configured to obtain worker information and platform authentication keys; calculate the collaboration benefits between workers based on the worker information and the platform authentication keys; obtain task information, and decompose the task information to obtain subtasks;

[0021] The capability benefit calculation module is configured to calculate the capability benefit of the worker for completing the subtask, and encrypt the capability benefit by giving the worker a false identity and an encryption key;

[0022] A completion benefit calculation module is configured to determine a candidate worker group for each subtask according to the capability benefit and the collaboration benefit, and calculate the task completion benefit of the worker for each subtask;

[0023] The selection module is configured to select the best group of collaborative workers for each subtask according to the workers' task completion benefits under the subtask budget constraint.

[0024] In a third aspect, the present invention provides a computer-readable storage medium storing a plurality of instructions, wherein the instructions are suitable for being loaded by a processor of a terminal device and executing the described method.

[0025] In a fourth aspect, the present invention provides a terminal device, comprising a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; and the computer-readable storage medium is used to store multiple instructions, wherein the instructions are suitable for being loaded by the processor and executing the described method.

[0026] In summary, the present invention has the following beneficial technical effects:

[0027] 1. In the context of multiple workers collaborating to complete multiple tasks, the present invention proposes a collaborative worker selection scheme. This scheme not only considers the impact of worker collaboration benefits and ability benefits on task completion, but also focuses on the impact of worker privacy protection and task decomposability on task completion, thereby achieving a more accurate evaluation of worker task completion benefits.

[0028] 2. The present invention proposes a two-stage privacy protection scheme for workers' task competition information and perceived data privacy. This scheme can not only protect workers' task competition information and perceived data privacy at the same time, but also ensure the accuracy of task allocation and the quality of uploaded data. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 4 is a schematic diagram of a worker selection method based on privacy protection in a multi-task scenario according to Embodiment 1 of the present invention.

[0030] Figure 2 This is a comparison chart of data quality of workers in the collaborative worker selection scheme of Example 1 of the present invention and other worker selection schemes.

[0031] Figure 3 This is a comparison chart of the average number of workers recruited for tasks of the collaborative worker selection scheme of Example 1 of the present invention and other worker selection schemes.

[0032] Figure 4 It is a comparison chart of the MSE of the worker privacy protection method according to Example 1 of the present invention and the data of the prior art method.

[0033] Figure 5 It is a comparison chart of the Hamming distance between the worker privacy protection method of Example 1 of the present invention and the prior art method. DETAILED DESCRIPTION

[0034] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0035] Example 1

[0036] Reference Figure 1 , a worker selection method based on privacy protection in a multi-task scenario of this embodiment includes:

[0037] Obtain worker information and platform authentication keys;

[0038] Calculate the collaboration benefits between workers based on worker information and platform authentication keys;

[0039] Obtain task information and decompose the task information into subtasks;

[0040] Calculate the worker's ability benefit for completing the subtask, and encrypt the ability benefit by giving the worker a false identity and encryption key;

[0041] Determine the candidate worker group for each subtask based on the capability benefit and collaboration benefit, and calculate the worker's task completion benefit for each subtask;

[0042] Under the subtask budget constraint, the best group of collaborative workers is selected for each subtask based on the workers' task completion benefits.

[0043] Specifically, the following steps are included:

[0044] In step 1, the worker registers with the trusted authority (TA), and the trusted authority negotiates with the worker to generate an identity authentication key.

[0045] Step 2: The task publisher (TP) publishes the task to the platform through the edge device. The platform receives multiple task publishers ( ) form a set of decomposable tasks, denoted as The platform delivers data to a group of collaborative workers through edge servers. Publish the group's tasks.

[0046] The decomposability of tasks refers to the Can be divided into Parts (i.e. subtasks), represented as ,in, ,in, The bid for the xth subtask of the ith task, represents the position of the xth subtask of the ith task, Represents the type of the xth subtask of the ith task. Workers can choose to complete subtasks from the same or different tasks instead of completing the entire task.

[0047] Step 3: The group of collaborative workers A shared key is obtained through the group key exchange protocol. Then, the worker uses the shared key and double encryption mechanism to encrypt his / her historical task information, and then uses the identity authentication key and signature algorithm to operate the ciphertext of the historical task information to generate identity authentication information. The ciphertext of the historical task information and the identity authentication information are sent to the TA.

[0048] The specific operation is as follows: First, the group of collaborative workers uses the ECDH-based group key exchange protocol to generate a shared key . Then, each worker uses The worker uses the AES encryption algorithm to encrypt his / her local historical task information, and then uses the SHA256 algorithm to further encrypt the encrypted historical task information to obtain the double-encrypted historical task information ciphertext. Finally, the worker uses the authentication key and signature algorithm to generate a signature message for the historical task information ciphertext, and sends the signature message and the double-encrypted historical task information ciphertext to the TA.

[0049] Step 4: TA verifies the identity of the worker. For workers who have passed identity verification, TA calculates the preference similarity benefit and social trust benefit of collaborating with other workers to complete the task. Then, TA generates a pair of homomorphic encryption keys and uses the key to encrypt the worker's preference similarity benefit and social trust benefit. At the same time, TA generates a fake identity for each worker and sends the corresponding worker's collaboration benefit to the platform through the fake identity.

[0050] The specific operation is as follows: First, TA verifies the identity of the worker based on the signature information and historical task information ciphertext sent by the worker. After verification, TA uses the historical task information ciphertext sent by the worker to calculate the preference similarity between the worker and other workers, and calculates the preference similarity benefit of the worker and other verified workers in completing the task in collaboration according to formula (2). At the same time, TA uses the social trust evaluation model based on graph neural network to calculate the social trust value between the verified worker and other verified workers, and calculates the social trust benefit of the worker and other workers in collaborating to complete the task according to formula (1). Finally, TA generates a pair of homomorphic encryption keys , a fake identity is matched for each worker, and the corresponding encrypted preference similarity benefit and social trust benefit are sent to the platform using the worker’s fake identity.

[0051] Among them, TA calculates workers With workers Collaborate on subtasks The preference similarity benefit and social trust benefit of are:

[0052] (1)

[0053] (2)

[0054] (3)

[0055] in, Indicates workers and The social trust benefits and Respectively represent workers and The social information value between workers and The social trust value between Indicates workers and The preference similarity benefit, jacc (wi,wj) Indicates workers and The preference similarity between them.

[0056] Step 5: TA will use the fake identity and Shared with workers. Then, workers use Encrypting the distance benefit, experience benefit, and privacy protection weakening benefit of one’s own local computation. Finally, workers Use a fake identity to submit benefit ciphertext and bid information to the platform through the edge server.

[0057] The specific operation is as follows: First, TA will and fake identity is shared with each worker. The workers with fake identities calculate the distance benefit, experience benefit, and privacy protection weakening benefit of completing each subtask according to equations (4), (5), and (6). Then, use The worker uses the encryption algorithm to encrypt his distance benefit, experience benefit and privacy protection weakening benefit. Finally, the worker uses the fake identity to send the ciphertext of distance benefit, experience benefit and privacy protection weakening benefit to the platform.

[0058] Among them, workers Complete subtasks The capacity benefit calculation formula is as follows:

[0059] (4)

[0060] (5)

[0061] (6)

[0062] in, , and Respectively represent workers Execute subtasks The distance benefit, experience benefit and privacy protection weakening benefit brought by it, Indicates workers To subtask distance.

[0063] Step 6: After receiving the collaboration information, capability information, and bid information, the platform counts the candidate worker groups for each subtask based on the bid information. Subsequently, the task completion benefits of the workers in each subtask candidate worker set are calculated, and the ciphertext of the task completion benefit value is sent to the TA for decryption. Finally, under the constraint of the subtask budget, the platform selects the best collaborative worker group for each subtask based on the worker's task completion benefit value and notifies the selected workers.

[0064] Among them, workers Complete subtasks The quality of service is:

[0065] (7)

[0066] (8)

[0067] (9)

[0068] in, , Represents a subtask The set of candidate workers, Indicates that workers are , ) Collaborate to complete subtasks The collaborative benefits Indicates workers Complete subtasks capacity benefits.

[0069] As a further implementation method, the genetic-tabu search algorithm with knapsack constraints proposed in this embodiment solves the worker recruitment problem in the scenario of this article, wherein the genetic algorithm is embedded in the taboo search algorithm as a result correction mechanism to further prevent the taboo search algorithm from falling into a local optimum. The specific implementation steps of the worker selection process are as follows:

[0070] First, the platform obtains the set of candidate workers for each subtask based on the worker's bid. Then, the genetic-tabu search algorithm with knapsack constraints is used to solve the optimal set of candidate workers for each subtask. Finally, the platform notifies the selected worker of the matching subtask and the group of collaborating workers for the subtask.

[0071] According to the worker's task completion efficiency, the genetic-tabu search algorithm with knapsack constraints is used to solve the optimal set of candidate workers for each subtask. The expression is:

[0072]

[0073] in, Represents a subtask The total benefit of all workers in the winning collaborative worker set, formula (10) refers to the subtask The total bid of the winning worker cannot exceed the subtask The budget of subtask is expressed in equation (11): Each winning worker with subtask The distance should be less than a threshold .

[0074] Step 7: Workers use the edge server to train the perception data into a local Diffusion model and use CKKS pk Encrypt local models and privacy budgets. Finally, workers aggregate their local models and privacy budgets through a secondary aggregation mechanism, and the final leader worker submits the aggregated data to the platform.

[0075] The specific operation is as follows: first, the worker will perform the Diffusion forward diffusion process locally on the collected perception data, and add noise that satisfies local differential privacy in the process of adding noise in the forward diffusion, and train a differential-Diffusion model; then, the local differential-Diffusion model and the privacy budget are encrypted using the TA shared key and the CKKS homomorphic encryption algorithm; then, the winning workers of the same subtask sequentially perform the first model aggregation process, and the leader worker finally obtains the aggregated model of the subtask. Then, the leader workers of all subtasks of the same task perform another data aggregation process, and the general leader worker collects and submits all subtask aggregation models.

[0076] Step 8: The platform sends the model ciphertext and privacy budget ciphertext of each subtask to the TA for decryption to obtain plaintext, and then sends it to the corresponding TR. Then, the TR performs Diffusion to recover the data in the model.

[0077] The entire federation-diffusion algorithm execution process is as follows: First, the task publisher generates an initial subtask model for each subtask Then, the workers After collecting the perception data, we use the task perception data to train a local difference-Diffusion model locally. ,in Indicates workers The matching task set. The training formula is shown below.

[0078] (13)

[0079] (14)

[0080] (15)

[0081] (16)

[0082] (17)

[0083] (18)

[0084] in, represents the forward diffusion formula; Represents the data of the e-th time step; is a coefficient that controls the noise level and usually increases with time; The identity matrix indicates that the noise in each dimension is independent and has the same variance; Indicates that the added noise is the noise that satisfies localized differential privacy; Indicates sensitivity, usually set to 1; Indicates the cumulative noise gradient parameter; represents the training loss function of forward diffusion, Indicates workers Local Diffusion Model, represents the noise of the model prediction at the e-th time step.

[0085] Next, the winning workers of the same subtask aggregate their local models in turn and hand over the aggregated model to the leader worker of the subtask. The first model aggregation formula is shown below.

[0086] (19)

[0087] in, Represents a subtask The aggregation model.

[0088] Then, the leader workers of all subtasks in the same task will select one of them as the general leader worker, and summarize the aggregated data of each group and hand it over to the general leader worker, who will submit the final data to the platform. After receiving the aggregated model of each group of workers, the platform will average the received subtask model parameters, as shown in the following formula.

[0089] (20)

[0090] in, Represents a subtask The set of winning workers, Represents a subtask The final aggregate model.

[0091] Then, the model ciphertext is sent to the TA for decryption. After decryption, the TA sends the model back to the platform, and the platform updates the global model of each subtask accordingly. Finally, after the task is completed, the platform will send the global model of each task to the corresponding task publisher, and the task publisher will perform the reverse diffusion process of the Diffusion model to denoise and restore the data stored in the model. The formula for the reverse diffusion process is shown below.

[0092] (twenty one)

[0093] (twenty two)

[0094] in, represents a parameterized mean function for denoising, Represents the covariance matrix used for prediction.

[0095] Example 2

[0096] This embodiment provides a worker selection method based on privacy protection in a multi-task scenario, including:

[0097] Step 1, Registration: The worker registers with TA, and TA negotiates with the worker to generate a unique identity authentication key.

[0098] Step 2, task upload: multiple task publishers upload tasks to the platform through the edge server, and the platform publishes the group of tasks to a group of collaborative workers.

[0099] Step 3: Key exchange and historical task information upload: Workers establish a shared key through the group key exchange protocol and use The double encryption mechanism is used to encrypt the historical task information. Then, the worker generates the identity authentication information using the identity authentication key and signature algorithm, and sends the encrypted historical task information and identity authentication information to the TA.

[0100] Step 4, Collaboration Information Upload: After TA verifies the identity of the worker, it calculates the preference similarity benefit and social trust benefit between the workers. Then, TA generates a pair of homomorphic encryption keys , encrypting the worker’s benefit data. At the same time, TA assigns a fake identity to each worker and sends the encrypted benefits to the platform through the fake identity. Finally, TA sends the fake identity and Share with workers.

[0101] Step 5, capability and bid information upload: The worker locally calculates and encrypts its distance benefit, experience benefit, and privacy protection weakening benefit for each subtask through the edge server, and then uses The benefit is encrypted using the encryption algorithm. Finally, the benefit ciphertext and the bid information of each subtask are uploaded to the platform using a fake identity.

[0102] Step 6, worker selection: After receiving the collaboration information, capability information, and bid information, the platform counts the candidate worker groups for each subtask based on the bid information. Subsequently, the task completion benefits of the workers in each subtask candidate worker set are calculated, and the task completion benefit value ciphertext is sent to the TA for decryption. Finally, under the constraint of the subtask budget, the platform selects the best collaborative worker group for each subtask based on the worker's task completion benefit value and notifies the selected workers.

[0103] Step 7, perception data submission: Workers train the perception data into a local Diffusion model through the edge server and use Encrypt the local model and privacy budget. Subsequently, the winning worker of the same task aggregates the local model and privacy budget of the workers through a two-step aggregation mechanism, and the leader worker submits the aggregated local model and privacy budget to the platform.

[0104] Step 8, data recovery: The platform sends the aggregated model ciphertext and privacy budget of each subtask to TA for decryption to obtain plaintext. Then, the platform sends the subtask model and privacy budget to the corresponding TR. Then, TR performs Diffusion to recover the data in the model.

[0105] The simulation experiment is as follows:

[0106] The collaborative worker selection method with strong privacy protection and considering task decomposability is experimented with real datasets.

[0107] (1) Task allocation stage

[0108] Among them, PDP-MTMCW represents the method of the present invention, and the other algorithms are comparative algorithms. The experimental results are shown in Figure 2 and Figure 3 As shown in the experiment, it can be seen that the total QoD of the workers selected by the task allocation method proposed in the present invention is better than the total QoD of the workers selected by the comparison algorithm.

[0109] (2) Data submission stage

[0110] Among them, FL-DPDiffusion represents the method of the present invention, and the other algorithms are comparative algorithms. The experimental results are shown in Figure 4 and Figure 5 From the experiment, it can be seen that the data quality of the uploaded perception data of the privacy protection method proposed in the present invention is better than the quality of the perception data uploaded by the comparison algorithm.

[0111] Example 3

[0112] This embodiment provides a worker selection system based on privacy protection in a multi-task scenario, including:

[0113] The data acquisition module is configured to obtain worker information and platform authentication keys; calculate the collaboration benefits between workers based on the worker information and the platform authentication keys; obtain task information, and decompose the task information to obtain subtasks;

[0114] The capability benefit calculation module is configured to calculate the capability benefit of the worker for completing the subtask, and encrypt the capability benefit by giving the worker a false identity and an encryption key;

[0115] A completion benefit calculation module is configured to determine a candidate worker group for each subtask according to the capability benefit and the collaboration benefit, and calculate the task completion benefit of the worker for each subtask;

[0116] The selection module is configured to select the best group of collaborative workers for each subtask according to the workers' task completion benefits under the subtask budget constraint.

[0117] A computer-readable storage medium stores a plurality of instructions, wherein the instructions are suitable for being loaded and executed by a processor of a terminal device for a worker selection method based on privacy protection in a multi-tasking scenario.

[0118] A terminal device includes a processor and a computer-readable storage medium, wherein the processor is used to implement various instructions; the computer-readable storage medium is used to store multiple instructions, wherein the instructions are suitable for being loaded by the processor and executing a worker selection method based on privacy protection in a multi-tasking scenario.

[0119] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A worker selection method based on privacy protection in a multi-task scenario, characterized in that: include: The worker registers with the trusted authority TA, and the trusted authority negotiates with the worker to generate an identity authentication key; The platform receives tasks issued by multiple task publishers to form a set of decomposable tasks; The platform publishes the set of tasks to a group of collaborative workers through the edge server; workers choose to complete subtasks from the same or different tasks instead of completing the entire task; The group of collaborative workers obtain a shared key through the group key exchange protocol; then, the workers use the shared key and double encryption mechanism to encrypt their own historical task information, and then use the identity authentication key and signature algorithm to operate the ciphertext of the historical task information to generate identity authentication information; the ciphertext of the historical task information and the identity authentication information are sent to the TA; TA verifies the identity of the worker; for the authenticated worker, TA calculates the collaborative benefit of the worker collaborating with other workers to complete the task; Then, TA generates a pair of homomorphic encryption keys and uses the homomorphic encryption keys to encrypt the workers' collaborative benefits. At the same time, TA generates a fake identity for each worker and sends the corresponding worker's collaborative benefits to the platform through the fake identity. The TA shares the fake identity and homomorphic encryption key with the worker; then, the worker uses the homomorphic encryption key to encrypt the capacity benefits of his local calculation; finally, the worker uses the fake identity to submit the capacity benefit ciphertext and bid information to the platform through the edge server; After receiving the collaboration information, capability information, and bid information, the platform counts the candidate worker groups for each subtask based on the bid information; Then, the task completion benefits of the workers in each subtask candidate worker set are calculated, and the ciphertext of the task completion benefit value is sent to the TA for decryption; Finally, under the constraint of the subtask budget, the platform selects the best group of collaborative workers for each subtask based on the workers' task completion benefit value and notifies the selected workers.

2. According to claim 1, a worker selection method based on privacy protection in a multi-task scenario is characterized in that: The collaboration benefit includes preference similarity benefit and social trust benefit, wherein the social trust benefit between workers is calculated based on the social information value and social trust value between workers and other workers; the preference similarity benefit between workers is calculated by the preference similarity value between workers and other workers.

3. According to the privacy protection-based worker selection method in a multi-task scenario of claim 2, it is characterized in that: The platform receives tasks issued by multiple task publishers to form a group of decomposable tasks, including dividing task information into several subtasks according to the location, bid and type of the task.

4. The method for selecting workers based on privacy protection in a multi-task scenario according to claim 3, characterized in that: The calculating of the worker's ability benefit for completing the subtask includes calculating the worker's distance benefit, experience benefit and privacy protection weakening benefit for completing the subtask based on the worker information.

5. The method for selecting workers based on privacy protection in a multi-task scenario according to claim 4, characterized in that: The calculation of the worker's task completion benefit for each subtask includes calculating the worker's task completion benefit for each subtask based on the collaboration benefit and capability benefit of the workers in the candidate worker group as the service quality of the worker completing the subtask.

6. A worker selection method based on privacy protection in a multi-task scenario according to claim 5, characterized in that: The method selects the best cooperative worker group for each subtask according to the worker's task completion efficiency, including using a genetic-tabu search algorithm with knapsack constraints to solve the optimal candidate worker set for each subtask.

7. A worker selection method based on privacy protection in a multi-task scenario according to claim 6, characterized in that: Each worker encrypts his or her own historical task information using a shared key and the AES encryption algorithm, and further encrypts the encrypted historical task information using the SHA256 algorithm to obtain a double-encrypted ciphertext of the historical task information.

8. A worker selection system based on privacy protection in a multi-task scenario, the system comprising workers, a platform and a trusted authority TA, characterized in that: The system is used to execute a worker selection method based on privacy protection in a multi-task scenario as described in any one of claims 1-7.

9. A computer-readable storage medium storing a plurality of instructions, characterized in that: The instructions are suitable for being loaded by a processor of a terminal device and executing a worker selection method based on privacy protection in a multi-tasking scenario as described in any one of claims 1-7.

10. A terminal device, comprising a processor and a computer-readable storage medium, wherein the processor is used to implement each instruction; and the computer-readable storage medium is used to store multiple instructions, characterized in that: The instructions are suitable for being loaded by a processor and executing a worker selection method based on privacy protection in a multi-tasking scenario as described in any one of claims 1-7.