Mobile crowd sensing task allocation method based on confusion privacy protection

Abstract

A mobile crowd sensing task allocation method based on confusion privacy protection, an MCS platform discloses a set of sensing tasks; a user maintains a location application times statistics table and a total application times counter locally, determines the utility of a task issued by the platform, randomly samples the task set based on the utility using a differential privacy exponential mechanism, obtains a sampling probability of each task, and determines an initial application task; the user disturbs the real distance of the initial application task using a Laplace mechanism to obtain a disturbed distance, and submits pre-application information to the platform; the platform aggregates all the pre-application information of the users, counts and discloses the number of applicants of each task; the user determines the expected income of the application task, selects the task with the maximum expected income, and performs distance disturbance using the same Laplace mechanism as in the pre-application stage to obtain a disturbed distance, and the MCS platform completes task allocation and determines the payment amount according to the disturbed information submitted by the user.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of data processing systems or methods specially adapted for administrative, commercial, financial, management, supervision or prediction purposes, and particularly relates to a mobile crowd sensing task allocation method based on confusion privacy protection. BACKGROUND

[0002] With the popularity of 5G / 6G wireless communication technology and mobile intelligent terminals such as smart phones and wearable devices, mobile crowd sensing (MCS) has developed into a core paradigm of data processing specially adapted for administrative, commercial, financial, management, supervision or prediction purposes. It aggregates the sensing capabilities of a large number of user terminals to provide low-cost, wide-coverage data support for multiple fields. In the administrative field, it can be used for urban governance, public service data collection in remote areas, and policy implementation tracking. In the commercial field, it can serve the sinking market research, precise marketing touch, and consumer trend monitoring. In the financial field, it can support remote credit demand investigation, cross-regional risk monitoring, and financial service coverage evaluation. In the management field, it can help optimize resource scheduling and manage cross-regional service efficiency. In the supervision field, it is suitable for environmental compliance verification and cross-regional supervision data collection. In the prediction field, it can achieve market demand prediction and public service supply and demand trend analysis based on regional data, becoming the key basis for data-driven decision-making in various fields.

[0003] In the MCS system, task allocation is the core link to ensure service quality and data value transformation. The MCS platform publishes sensing tasks with specific geographic locations, and users apply for tasks by submitting their distance information. The platform allocates tasks to users based on the nearest distance principle. Although this mechanism improves task allocation efficiency, it exposes the following key problems when adapting to the application needs of administrative, commercial, financial and other fields:

[0004] First, location privacy leakage leads to multiple scene risks. Users need to submit location-related distance information to the platform for task application. The platform and related data users such as commercial institutions and administrative departments can accurately infer the active area and permanent location (such as residence, workplace, and frequently visited consumption places) of the user by analyzing the user's historical task application data combined with community maps, traffic routes, and consumption behavior records. This privacy leakage not only exposes the user's personal sensitive information, but also may be used for commercial harassment, financial credit discrimination, and excessive administrative intervention, which seriously infringes on the user's legal rights and hinders the promotion and application of MCS technology in high-privacy-sensitive fields. Although differential privacy technology has been widely used in MCS privacy protection, most existing solutions only design indistinguishability constraints based on distance dimension, and do not consider that the user's preference for close-range tasks will lead to spatial concentration of historical application data. The platform can still break through the privacy protection boundary through statistical analysis, which cannot meet the strict requirements of privacy protection in administrative, commercial, and financial fields. Second, task completion failure in sparse areas affects data service coverage. Users have a natural preference for close-range tasks based on the rational choice of reducing execution costs, which leads to a very low task application rate in sparse areas such as suburbs and remote monitoring points, and may result in no one applying. This problem is particularly prominent in administrative supervision, such as environmental monitoring in remote areas and data collection for rural public services, market consumption behavior research in sinking markets, and credit demand investigation in remote areas, directly leading to incomplete data collection and insufficient decision-making basis in related fields, which seriously affects the service coverage and usability of the MCS system and limits its deep application in a wider range of scenarios.

[0005] Therefore, the existing MCS task allocation scheme has obvious limitations in adapting to specific data processing scenarios such as administration, business, finance, and management. In terms of privacy protection, only the distance factor is considered, and the protection strength is not dynamically adjusted in combination with the number of task applications, which cannot effectively resist inference attacks based on historical application data. At the same time, there is a lack of task guiding mechanism in sparse areas, leading to a high failure rate of tasks in sparse areas. Some solutions introduce excessive noise to strengthen privacy, leading to distorted allocation decisions, while other solutions sacrifice privacy protection strength to improve efficiency, making it difficult to balance privacy protection and allocation efficiency. SUMMARY

[0006] The technical problem to be solved by the present application is to overcome the shortcomings of the prior art and provide a mobile crowd sensing task allocation method based on confusion privacy protection, which significantly improves the accuracy of privacy protection and the efficiency of task allocation.

[0007] The technical solution adopted to solve the above technical problems is: a mobile crowd sensing task allocation method based on confusion privacy protection, comprising the following steps:

[0008] Step 1. MCS platform publicly releases a set of sensing tasks to be executed , and provides the geographic location information of each task, while building a global location application times statistics table ;

[0009] Step 2. User maintains a location application times statistics table L and a total application times counter C on the local terminal, which are used to record the historical application times of the user in different location cells. Each time a new task is applied for, the application times and the total application times of the corresponding location are updated;

[0010] Step 3. User determines the utility of the task sent by the MCS platform according to the following formula ,

[0011]

[0012] wherein, is the true distance between the user and the task , and is the application times proportion of the location where the task is located;

[0013] Step 4. User randomly samples the task set T based on the utility using the differential privacy exponential mechanism, to obtain the sampling probability of each task , and determine the initial application task ;

[0014] Step 5. User perturbs the true distance of the initial application task using the Laplace mechanism, to obtain the perturbed distance , so as to avoid the leakage of location information by the user's submitted true distance;

[0015] Step 6. User submits the pre-application information to the platform, is the unique identifier of the user, is the willingness to adjust, , represents the secondary allocation when the initial task is not accepted, represents the non-acceptance;

[0016] Step 7. MCS platform aggregates the pre-application information of all users, and counts and publicly releases the number of applicants of each task ;

[0017] Step 8. User determines the number of applicants of the task ​the real distance , the expected revenue of the application task is determined according to the following formula ,

[0018]

[0019] wherein, is a revenue scale factor, ;

[0020] Step 9. The user traverses all the application numbers satisfying the task , is the initial application number of the user for the application task , the task with the maximum expected revenue is screened out , and the distance disturbance is performed on the task using the same Laplace mechanism as in the pre-application stage to obtain the disturbed distance , and the adjusted application information is submitted to the MCS platform ;

[0021] Step 10. The MCS platform completes task allocation and determines the payment amount according to the disturbance information submitted by the user

[0022] Step 10.1. The MCS platform constructs a user set S recording the allocation state of the user, and the initial state is all unallocated; for each task , the users applying for the task are collected to form a user set B, the user with the minimum submitted disturbance distance is selected as the winner in the user set B, and the user set S is updated;

[0023] Step 10.2. The MCS platform counts the remaining task set that has not been applied for, sorts the remaining tasks according to the application frequency from small to large based on the historical application records of the user recorded in the global position application frequency statistical table ;

[0024] Step 10.3. For each remaining task , the users who have not been allocated and accept the adjustment are screened out to obtain a user set , and the comprehensive index of the user in the user set is determined according to the following formula ,

[0025]

[0026] wherein, is a weight coefficient, is a set of position unit cells covered by the MCS platform, is the remaining task ​The cell where the user is located;

[0027] Step 10.4. Select the user with the smallest comprehensive index as the winner, assign the task , and update the user set S;

[0028] Step 10.5. For each winning user, the MCS platform determines the payment amount based on the submitted perturbed distance.

[0029] As a preferred technical solution, in the step 3, the task The proportion of the number of applications for the location where the task is located is:

[0030] .

[0031] As a preferred technical solution, in the step 4, the sampling probability of each task is:

[0032] In the formula, is the privacy protection budget.

[0033] As a preferred technical solution, in the step 5, the perturbed distance is:

[0034]

[0035] In the formula, is the Laplace distribution noise.

[0036] As a preferred technical solution, in the step 10.5. For each winning user, the MCS platform determines the payment amount based on the submitted perturbed distance , .

[0037] The beneficial effects of the present application are as follows:

[0038] The task application confusion privacy proposed by the present application avoids the platform to make privacy inference only through the distance dimension by dynamically constraining the probability indistinguishability of users selecting different tasks. Not only the user's preference for close distance tasks is considered, but also the historical information of the number of task applications is combined, so that the privacy protection is more accurate. Effectively prevent the platform from inferring the user's permanent location through historical application data, significantly improve the privacy protection strength, and overcome the defects of traditional differential privacy scheme based only on distance.

[0039] The user of the application locally maintains a task application record, avoids leakage of privacy information in network transmission, adopts a utility function combining distance and application times proportion, meets the user's preference for close distance tasks, provides a basis for subsequent guiding the user to apply for sparse area tasks, randomly selects tasks through a differential privacy index mechanism, ensures that the task selection probability meets the confusion privacy, and further protects the user's location privacy by perturbing the real distance through Laplace distance perturbation.

[0040] The application is based on a user application secondary adjustment mechanism based on expected income, guides the user to actively apply for sparse area tasks, significantly improves the application rate and completion rate of sparse area tasks, and optimizes the task allocation efficiency. The mechanism of determining the payment amount based on the perturbed distance realizes the task full allocation and the user non-negative income, guarantees the long-term sustainability of the system, and reduces the system operation cost through global optimization. BRIEF DESCRIPTION OF DRAWINGS

[0041] Figure 1 is a flowchart of a mobile crowd sensing task allocation method based on confusion privacy protection of the application. DETAILED DESCRIPTION

[0042] The application will be further described in detail below in combination with the drawings and examples, but the application is not limited to the following embodiments.

[0043] In Figure 1 , a mobile crowd sensing task allocation method based on confusion privacy protection of the embodiment includes the following steps:

[0044] Step 1. The MCS platform publicly releases the sensing tasks to be executed

[0045] The MCS platform publicly releases the sensing task set to be executed , and provides the geographic location information of each task, at the same time, builds a global location application times statistical table ;

[0046] Step 2. The user maintains a location application times statistical table L and a total application times counter C on the local terminal, which is used to record the historical application times of the user in different location cells. Each time a new task is applied for, the application times and the total application times of the corresponding location are updated;

[0047] Step 3. The user determines the utility of the task issued by the MCS platform according to the following formula ,

[0048]

[0049] In the formula, is the utility of the user and the task The real distance of the task is the task The proportion of the number of applications of the location where the task is located

[0050] The task The proportion of the number of applications of the location where the task is located

[0051] .

[0052] Step 4. The user determines the initial application task based on the utility The sampling probability of each task in the task set T is determined by using the differential privacy exponential mechanism, and the initial application task is determined;

[0053] The sampling probability is:

[0054]

[0055] In the formula, is the privacy protection budget;

[0056] Step 5. The user uses the Laplace mechanism to disturb the real distance of the initial application task , and obtains the disturbed distance , , The Laplace distribution noise avoids the leakage of location information by the user submitting the real distance;

[0057] Step 6. The user submits the pre-application information to the platform, is the unique identification of the user, is the willingness to adjust, , represents the secondary allocation when the initial task is not accepted, represents not accepting;

[0058] Step 7. The MCS platform collects all the pre-application information of the user, and counts and publishes the number of applicants of each task ;

[0059] Step 8. The user determines the expected income of the application task based on the number of applicants published by the MCS platform and the real distance of the task to the task ,

[0060]

[0061] wherein, is the benefit scale factor, ;

[0062] Step 9. The user traverses all the applicant numbers satisfying the task , is the applicant number of the initial application task of the user, filters out the task with the largest expected benefit, and performs distance disturbance on the task using the same Laplace mechanism as in the pre-application stage to obtain the disturbed distance , submits the adjusted application information to the MCS platform ;

[0063] Step 10. The MCS platform completes task allocation and determines the payment amount according to the disturbance information submitted by the user

[0064] Step 10.1. The MCS platform constructs a user set S recording the allocation state of the user, and the initial state is all unallocated; for each task , collects the users applying for the task to form a user set B, selects the user with the smallest submitted disturbance distance in the user set B as the winner, and updates the user set S;

[0065] Step 10.2. The MCS platform counts the remaining task set that has not been applied for, sorts the remaining tasks according to the application frequency from small to large based on the historical application records of the users recorded in the global position application frequency statistics table ;

[0066] Step 10.3. For each remaining task , filter out the users who have not been allocated and accept the adjustment to obtain a user set , determine the comprehensive index of the users in the user set according to the following formula ,

[0067]

[0068] wherein, is the weight coefficient, is the set of position cell covered by the MCS platform, is the cell where the remaining task is located;

[0069] Step 10.4. Select the user with the smallest comprehensive index as the winner, allocate the task , and update the user set S;

[0070] Step 10.5. For each winning user, the MCS platform determines the payout amount based on the perturbation distance submitted by it , .

Claims

1. A method for mobile crowd sensing task allocation based on obfuscation privacy protection, characterized in that, Comprising the following steps: Step 1. The MCS platform publicly releases a set of perception tasks to be performed and provides geographic location information for each task, while building a global location application times table ; Step 2. The user maintains a location application times statistics table L and a total application times counter C at the local terminal, which are used to record the historical application times of the user in different location cells. Each time a new task is applied for, the application times and the total application times of the corresponding location are updated; Step 3. The user determines the utility of the tasks issued by the MCS platform according to the following formula ,​ In the formula, the real distance between the user and the task , the application times proportion of the location where the task is located; Step 4. User based on utility Randomly sample the task set T using the differential privacy exponential mechanism to obtain the sampling probability of each task Determine the initial application task ; Step 5. User adopts Laplace mechanism to perturb the real distance of initial application task to obtain the perturbed distance , avoiding the real distance submitted by the user to leak the location information . Step 6. The user submits pre-application information to the platform , is a unique identification for the user, is a dispensing intention, , represents a secondary allocation when the initial task is not accepted, represents not accepted; Step 7. The MCS platform aggregates all the pre-application information of users, counts and discloses the number of applicants of each task ;​ Step 8. User based on MCS platform publicized the number of applicants with the real distance to the task , the expected return of the application task ,​​ In the formula, is a yield scale factor, ; Step 9. The user traverses all the tasks of the applicant number satisfying , , the initial application task of the user , filters out the task with the largest expected income , and uses the same Laplace mechanism as in the pre-application stage to disturb the distance of the task , to obtain the disturbed distance , and submits the adjusted application information to the MCS platform ; Step 10. The MCS platform completes task allocation and determines the payment amount according to the disturbance information submitted by the user Step 10.

1. The MCS platform constructs a user set S recording the allocation state of users, and the initial state is all unallocated; for each task , collects the users applying for the task to form a user set B, selects the user with the minimum submitted perturbation distance in the user set B as the winner, and updates the user set S; Step 10.

2. MCS platform statistics unapplied remaining task set Based on the global location application frequency statistics table Recorded user history application records, sort the remaining tasks by application frequency from small to large; Step 10.

3. For each remaining task , screen out the users who are not assigned and accept the adjustment, to obtain a user set , determine the user comprehensive index in the user set , according to the following formula , In the formula, is a weight coefficient, is a set of location cells covered by the MCS platform, is a remaining task the cell where the UE is located; Step 10.

4. Select the user with the smallest integrated index as the winner, assign the task , and update the set of users S; Step 10.

5. For each winning user, the MCS platform determines the payment amount based on the disturbance distance submitted by it.

2. The method of claim 1, wherein, In step 3, the task The proportion of the number of applications in the location Is: 。 3. The method of claim 1, wherein, In step 4, the sampling probability of each task is: ​ In the formula, is a privacy protection budget.

4. The method of claim 1, wherein, The step 5, the distance after the disturbance is: In the formula, is a Laplace-distributed noise.

5. The method of claim 1, wherein, The step 10.

5. For each winning user, the MCS platform determines the payout amount based on the perturbation distance submitted by it , .

Images