CN111144888A - Mobile crowd sensing task allocation method with differential privacy protection function - Google Patents

Abstract

The invention discloses a differential privacy protection mobile crowd sensing task allocation method, which comprises the steps of 1, establishing an auction model according to task information submitted by a requester, request information submitted by a worker and a cloud platform; 2. determining final settlement price groups of the requesters and the workers by combining a differential privacy technology under the framework of perception task distribution based on auction; 3. determining corresponding winning requestors and workers based on the settlement price group; 4. after receiving the data submitted by the winning workers, the cloud platform calculates an aggregate result and sends the aggregate result to the winning requester; 5. the winning requester pays the corresponding reward and the winning worker receives the corresponding reward. The invention can effectively solve the problem of leakage of bid information of the worker and the requester on the premise of fully considering the perception of the data of the worker, thereby protecting the bid privacy of the worker and the requester, realizing better task distribution and simultaneously improving the data quality of the task of the requester.

Description

Mobile crowd sensing task allocation method with differential privacy protection function

Technical Field

The invention relates to the technical field of mobile crowd sensing, in particular to a mobile crowd sensing task allocation method for realizing differential privacy protection.

Background

The mobile crowd sensing system utilizes various mobile devices equipped by the public crowd to complete large-scale sensing tasks, and flexible and extensible sensing coverage is realized at lower deployment cost. At present, the mobile crowd sensing systems are widely applied, including applications of environmental monitoring, traffic monitoring, road condition detection and the like.

For workers, performing a sensory task requires the consumption of certain resources, including computing power, batteries, etc., and to encourage more workers to participate, auctions are a reasonable way of motivating workers to participate and distribute the task. In the framework formed by the cloud platform, the requester and the worker, the cloud platform is used as an auctioneer, the requester is used as a buyer, and the worker is used as a seller to construct a model based on the two-way auction. In order to maximize the utility of the requester, the requester and the requester need to submit real price quote and ask, but the attacker can cause privacy information leakage by reasoning attack, namely deducing the real price quote and ask of the requester according to the result of auction, thereby reducing the enthusiasm of the worker and the requester to participate.

In order to solve the problem of inferring the bidding privacy of participants from the public result, the concept of differential privacy is introduced into the mobile crowd sensing application, but most privacy protection schemes only comprise two parties of a platform and a worker in the case of one-way auction, so that the bid information of the worker is protected. For a Mobile crowd-sourcing perception framework based on the two-way Auction, a differential privacy two-way Auction Scheme is proposed in the document DDPA (distributed privacy Private Double Auction) Scheme for Mobile crown sensing,2018, so as to protect the bidding privacy of two parties in Auction. However, this solution does not take into account the quality of the sensory data provided by the workers. In addition to designing a reasonable incentive mechanism capable of realizing privacy protection to attract more workers to participate in the mobile crowd sensing system, the mechanism also needs to consider the quality problem of the collected worker sensing data to ensure the benefit of a task requester.

Disclosure of Invention

In order to solve the defects of the prior art, the invention provides a differential privacy protection mobile crowd sensing task allocation method, so that the problem of leakage of bid information of workers and requesters can be effectively solved on the premise of fully considering the data sensed by the workers, the bid privacy of the workers and the requesters is protected, better task allocation is realized, and the data quality of the tasks of the requesters is improved.

The technical scheme adopted by the invention for solving the technical problem is as follows:

the invention discloses a differential privacy protection mobile crowd sensing task allocation method which is characterized in that the method is applied to a cloud platform, and m requesters m & ltm & gt ═ r & ltr & gt₁,r₂,...,r_j,...,r_mN worker Ν ═ u₁,u₂,...,u_i,...,u_nIn a mobile crowd sensing environment, where r_jDenotes the jth requestor, u_iRepresenting the ith worker, j is 1,2, …, m, i is 1,2, …, n, and the mobile crowd sensing task allocation method is carried out according to the following steps:

step 1, establishing an auction model and initializing:

step 1.1, the cloud platform is used as an auctioneer to collect task information submitted by m requesters, wherein the jth requester r_jIs recorded as S_j＝＜τ_j,s_j＞，τ_jDenotes the jth requester r_jRequested aware task, s_jIndicating completion of the perception task τ by the jth requester_jAnd s, and_j∈Π^s＝[v_min,v_max]，v_minminimum value, v, representing all requesters' quotes_maxMaximum value, Π, of quotes representing all requesters^sThe value space set representing the quotation, thereby obtaining m task information sets S ═ { S } provided by the requesters₁,S₂,...,S_j,...,S_mAnd its corresponding perceptual task set T ═ τ₁,τ₂,...,τ_j,...,τ_m}；

Step 1.2, the cloud platform issues the collected perception task set T to n workers, wherein the ith worker u_iThe submitted request information is marked as B_i＝＜Γ_i,b_i＞，Γ_iIndicates the ith worker u_iA set of tasks of interest, an

b_iIndicates the ith worker u_iExecuting task set Γ_iAsking price of all tasks in, and b_i∈Π^b＝[c_min,c_max]，c_minMinimum value representing the asking price of all workers, c_maxMaximum value, Π, representing the asking price for all workers^bValue space set representing the price to be charged, thereby obtaining request information set BETA ═ { B) submitted by n workers₁,B₂,...,B_i,...,B_n}；

Step 1.3, define settlement price group set of requester and worker

Wherein the content of the first and second substances,

represents the l-th settlement price group,

indicating the ultimate award to the winning worker,

denotes a reward finally paid by a winning requester, L denotes the number of settlement price groups, and L ═ Π^b×Π^sAnd has: II type^b×Π^s＝{(p_b,p_s)|p_b∈Π^b∧p_s∈Π^s}；

Step 2, according to the task information set S and the request information set BETA submitted by the requester and the worker, the cloud platform determines the settlement price group rho of the winning requester and the winning worker by using an index mechanism_win；

Step 2.1, initializing 1;

step 2.2, the l-th settlement price group rho_lPrice of (1)

Determining that the ask does not exceed the ask among the n workers

Set of workers Ν_lAnd determine set of workers Ν_lA candidate task set T formed by task sets requested by all workers_l；

Step 2.3, from set of workers Ν_lSequentially selecting the workers providing the best technical level until the candidate task set T_lAll tasks in (b) meet the quality of service requirements, resulting in a set of candidate workers W_{B_l}And a set of candidate requestors W_{S_l}；

Step 2.4, the cloud platform from candidate worker set W_{B_l}To a candidate task set T_lThe j (middle) thTask tau_jAnd constitutes the jth task τ_jWorker set C of_{j_l}，j＝{1,2,...,|T_l|}，|T_lI represents a candidate task set T_lThe size of (d);

step 2.5, the l settlement price group rho_lPrice of (1)

Determining a set of candidate requestors W_{S_l}Wherein the requester's bid is not less than

Winning requester set of

And corresponding task set

By the task collection

The set of workers for all tasks in the set of winning workers constitutes a winning set of workers

Step 2.6, judge in winning worker set

And winning requester set

Platform revenue generated

If yes, directly executing the step 2.8; otherwise, the platform profit is negative, and step 2.7 is executed;

step 2.7, delete in turn the request that produces the greatest impact on the negative platform revenue valueThe player and the corresponding worker, thereby updating the winning set of workers

And winning requester set

Until the platform profit is not less than 0, so as to obtain the final updated winning worker set and winning requester set and respectively assign the values to

And

step 2.8, calculate the set of winning workers using equation (1)

And winning requester set

Social welfare Wel_l(B,S,ρ_l)：

Step 2.9, assigning L +1 to L, judging whether L is greater than L, and if so, indicating that social benefits corresponding to L settlement price groups are obtained; otherwise, returning to the step 2.2;

step 2.10, obtaining settlement price group set by using the formula (2)

Selecting the l-th settlement price group rho_lProbability of (p) Pr (p)_l) Thereby obtaining a probability distribution { Pr (ρ) of the settlement price group_l) 1,2, …, L, and according to the probability distribution of the settlement price group { Pr (ρ |)_l) 1,2, …, L, randomly selecting a settlement price group as the final winningSettlement price group rho_win：

In the formula (2), ε represents a privacy budget, and ρ' represents a settlement price group set

Any one of the settlement price groups, Wel ' (B, S, ρ ') represents a value of social welfare calculated when the settlement price group is ρ ', Δ Wel represents sensitivity, and Δ Wel ═ m × v ″_max-c_min；

Step 3, the final winning settlement price group rho_winCorresponding set of winning workers

And winning requester set

As a final set of winners;

step 4, final winning worker set

Wherein the ith winning worker u submits the perception data to the cloud platform_iThe submitted set of sensory data is noted as

d_i,jIndicating the ith winning worker u_iPerforming the jth task τ_jThe perceptual data that is generated at the time of the generation,

represents the size of the final winning worker set, j 1,2_i|，|Γ_iI denotes the ith winning worker u_iAny thing of interestA service set size;

the cloud platform calculates a final winning requester set using equation (3)

The j winning requester r_jAggregated results of tasks of (1)

And sent to the corresponding j-th winning requester r_j：

In the formula (3), θ_i,jIndicating that the ith worker performed the jth task τ_jAnd theta_i,j＝Pr[d_i,j＝D_j]∈[0,1]，D_jRepresenting tasks τ_jTrue perception data of (1), Pr [ d ]_i,j＝D_j]Indicating that the ith worker performed task τ_jData submitted at time d_i,jAnd task tau_jTrue perception data D of time_jThe probability of being equal is determined,

represents a winning requester set size;

step 5, when the final winning requester set

After each winning requester receives the data, the cloud platform according to the final winning settlement price group rho_winCollecting consideration from all winning requestors

And will award

Paying to the set of final winning workers

All winning workers.

The mobile crowd sensing task allocation method with the differential privacy protection function is also characterized in that the step 2.3 is carried out according to the following process:

step 2.3.1, set of slaves Ν_lSelecting workers with technical level satisfying formula (4) as workers capable of letting task tau_jCandidate workers meeting quality of service requirements;

in the formula (4), θ_i,jIth worker performs task τ_jβ_jE (0,1) represents the jth task tau_jAccuracy requirement of, W_{B_l}Representing a set of candidate workers;

step 2.3.2, let q_i,j＝(2θ_i,j-1)²Indicates the ith worker u_iFor jth task τ_jThe quality of service of (a) is,

denotes the jth task τ_jQ ═ q, then_i,j]^n×mQuality of service matrix representing n workers for m tasks, Q ═ Q_j]^m×1Representing the quality of service requirement matrix of m tasks, equation (4) is simplified to equation (5):

in the formula (5), x_iIndicates the ith worker u_iWhether winning, if winning, it is 1, otherwise it is 0, y_jDenotes the jth requester r_jWhether the card wins, if so, the card is 1, otherwise, the card is 0;

step 2.3.3, initialization

Step 2.3.4, set of slaves Ν_lSelecting a worker u satisfying the formula (6)_maxAs candidate workers, and from the set of workers Ν_lMiddle delete worker u_maxPost-joining to the set W_{B_l}In order to update the set of workers Ν_l：

Step 2.3.5, assign value to set of updated workers to Ν_lIs mixing Q with_j-min{Q_j,q_max,jAssign a value to Q_j(ii) a Wherein q is_max,jRepresent worker u_maxFor jth task τ_jQuality of service of (2);

step 2.3.6, judgment

If yes, returning to the step 2.3.4; otherwise, the representation gets a candidate worker set W_{B_l}And a set of candidate workers T_lCorresponding candidate requester set W_{S_l}。

The step 2.7 is carried out according to the following processes:

step 2.7.1, gather from winning requesters

In which the kth requester r is removed_kGet the kth set of requesters

Thereby obtaining

A set of requesters

Representing a winning requester set

Size;

step 2.7.2, obtaining the kth requester set by using the formula (7)

Corresponding kth worker set

Thereby obtaining

A set of workers

Step 2.7.3, determine the requester r that has the greatest impact on the negative platform revenue value using equation (8)_max：

Step 2.7.4, gather from winning requestors

Deletion requester r_maxGet the set of requesters

And corresponding set of workers

And respectively assigned with

And

step 2.7.5, judging whether the platform income is less than 0, if so, returning to step 2.7.1, otherwise, indicating that a winning requester set is obtained

And winning worker set

Compared with the scheme in the prior art, the invention has the beneficial effects that:

1. the invention applies a differential privacy technology under a mobile crowd sensing framework based on bidirectional auction, realizes privacy protection by a mode of probabilistically selecting and settling price groups through an index mechanism, ensures the safety of bid information of workers and requesters, and attracts more requesters and workers to participate.

2. According to the method, the privacy information of the requester and the worker is protected by using the differential privacy, and compared with the method for realizing the safety by using an encryption technology, the calculation and communication overhead of an encryption algorithm is reduced, and the task allocation efficiency is improved; in addition, the distribution method comprehensively considers the service quality of data submitted by workers and improves the total distribution utility.

3. In the task allocation process, the invention selects the workers with high reliability to execute the related tasks, and adopts the data aggregation method to calculate the data aggregation result, thereby improving the quality of the perception task, ensuring that the requirement of data accuracy provided by the platform can be met, ensuring the benefit of the requester, and improving the participation enthusiasm of the requester.

4. In the process of using the index mechanism, the social benefit is used as a utility function for determining the settlement price, so that the result of task allocation can generate better social benefit and has better practicability.

Drawings

FIG. 1 is a framework diagram of the present invention;

FIG. 2 is a flow chart of the main implementation steps of the present invention.

Detailed Description

In this embodiment, a mobile crowd sensing task allocation method with differential privacy protection, as shown in fig. 1, is applied to a cloud platform, where m requesters m ═ r₁,r₂,...,r_j,...,r_mN worker Ν ═ u₁,u₂,...,u_i,...,u_nIn a mobile crowd sensing environment, where r_jDenotes the jth requestor, u_iRepresenting the ith worker, j is 1,2, …, m, i is 1,2, …, n, as shown in fig. 2, the mobile crowd sensing task allocation method is performed as follows:

step 1, establishing an auction model and initializing: according to information submitted by a requester and a worker, the credible cloud platform is regarded as an auctioneer, the requester is regarded as a buyer, the worker is regarded as a seller, and a two-way auction model is constructed;

step 1.1, the cloud platform is used as an auctioneer to collect task information submitted by m requesters, wherein the jth requester r_jIs recorded as S_j＝＜τ_j,s_j＞，τ_jDenotes the jth requester r_jRequested aware task, s_jIndicating completion of the perception task τ by the jth requester_jAnd s, and_j∈Π^s＝[v_min,v_max]，v_min10 denotes the minimum of all requesters' quotes, v_max15 represents the maximum value of the quotes for all requesters, Π^sThe value space set representing the quotation, thereby obtaining m task information sets S ═ { S } provided by the requesters₁,S₂,...,S_j,...,S_mAnd its corresponding perceptual task set T ═ τ₁,τ₂,...,τ_j,...,τ_m}；

Step 1.2, the cloud platform will collectThe sensed task set T is issued to n workers, wherein the ith worker u_iThe submitted request information is marked as B_i＝＜Γ_i,b_i＞，Γ_iIndicates the ith worker u_iA set of tasks of interest, an

b_iIndicates the ith worker u_iExecuting task set Γ_iAsking price of all tasks in, and b_i∈Π^b＝[c_min,c_max]，c_min5 represents the minimum value of the asking price of all workers, c_max10 represents the maximum value of the asking price for all workers, Π^bValue space set representing the price to be charged, thereby obtaining request information set BETA ═ { B) submitted by n workers₁,B₂,...,B_i,...,B_n}；

Step 1.3, define settlement price group set of requester and worker

Wherein the content of the first and second substances,

represents the l-th settlement price group,

indicating the ultimate award to the winning worker,

denotes a reward finally paid by a winning requester, L denotes the number of settlement price groups, and L ═ Π^b×Π^sAnd has: II type^b×Π^s＝{(p_b,p_s)|p_b∈Π^b∧p_s∈Π^s}；

Step 2, according to the task information set S and the request information set BETA submitted by the requester and the worker, the cloud platform determines the settlement price of the winning requester and the winning worker by using an index mechanismGroup rho_win；

Step 2.1, initializing 1;

step 2.2, the price in the l-th settlement price group rho l

Determining that the ask does not exceed the ask among the n workers

Set of workers Ν_lAnd determine set of workers Ν_lA candidate task set T formed by task sets requested by all workers_l；

Step 2.3, from set of workers Ν_lSequentially selecting the workers providing the best technical level until the candidate task set T_lAll tasks in (b) meet the quality of service requirements, resulting in a set of candidate workers W_{B_l}And a set of candidate requestors W_{S_l}；

Step 2.3.1, set of slaves Ν_lSelecting workers with technical level satisfying formula (1) as workers capable of letting task tau_jCandidate workers meeting quality of service requirements;

in the formula (1), θ_i,jIth worker performs task τ_jβ_jE (0,1) represents the jth task tau_jI.e. the jth requester r_jThe probability that the obtained task data is not equal to the real data of the task is not more than β_j，W_{B_l}Representing a set of candidate workers;

step 2.3.2, let q_i,j＝(2θ_i,j-1)²Indicates the ith worker u_iFor jth task τ_jThe quality of service of (a) is,

denotes the jth task τ_jQuality of service ofWhen the amount is required, q is ═ q_i,j]^n×mQuality of service matrix representing n workers for m tasks, Q ═ Q_j]^m×1Representing the quality of service requirement matrix of m tasks, equation (1) is simplified to equation (2):

in the formula (2), x_iIndicates the ith worker u_iWhether winning, if winning, it is 1, otherwise it is 0, y_jDenotes the jth requester r_jWhether the card wins, if so, the card is 1, otherwise, the card is 0;

step 2.3.3, initialization

Step 2.3.4, set of slaves Ν_lSelecting a worker u satisfying the formula (3)_maxAs candidate workers, and from the set of workers Ν_lMiddle delete worker u_maxPost-joining to the set W_{B_l}In order to update the set of workers Ν_l：

Step 2.3.5, assign value to set of updated workers to Ν_lIs mixing Q with_j-min{Q_j,q_max,jAssign a value to Q_j(ii) a Wherein q is_max,jRepresent worker u_maxFor jth task τ_jQuality of service of (2);

step 2.3.6, judgment

If yes, returning to the step 2.3.4; otherwise, the representation gets a candidate worker set W_{B_l}And a set of candidate workers T_lCorresponding candidate requester set W_{S_l}；

Step 2.4, cloud platform from candidate worker set W_{B_l}To a candidate task set T_lTask j (j) in (1)_jAnd constitutes the jth task τ_jWorker set C of_{j_l}，j＝{1,2,...,|T_l|}，|T_lI represents a candidate task set T_lSize;

step 2.5, the l-th settlement price group rho_lPrice of (1)

Determining a set of candidate requestors W_{S_l}Wherein the requester's bid is not less than

Winning requester set of

And corresponding task set

By task aggregation

The set of workers for all tasks in the set of winning workers constitutes a winning set of workers

Step 2.6, judge in winning worker set

And winning requester set

Platform revenue generated

If yes, directly executing the step 2.8; otherwise, the platform profit is negative, and step 2.7 is executed; in order to guarantee budget balance in task allocation, platform profit is ensured to be not less than0；

Step 2.7, sequentially deleting the requesters and corresponding workers having the greatest impact on negative platform revenue values, thereby updating the winning set of workers

And winning requester set

Until the platform profit is not less than 0, so as to obtain the final updated winning worker set and winning requester set and respectively assign the values to

And

step 2.7.1, gather from winning requesters

In which the kth requester r is removed_kGet the kth set of requesters

Thereby obtaining

A set of requesters

Representing a winning requester set

Size;

step 2.7.2, obtaining the kth requester set by using the formula (4)

Corresponding kth worker set

Thereby obtaining

A set of workers

In equation (4), to guarantee the set of requesters

The task of the requester in (1) meets the quality of service requirement, so the set of workers

Is made up of a collection of workers for each task;

step 2.7.3, determine the requester r that has the greatest impact on the negative platform revenue value using equation (5)_max：

Step 2.7.4, gather from winning requestors

Deletion requester r_maxGet the set of requesters

And corresponding set of workers

And respectively assigned with

And

step 2.7.5, judging whether the platform income is less than 0, if so, returning to step 2.7.1, otherwise, indicating that a winning requester set is obtained

And winning worker set

Step 2.8, calculate the set of winning workers using equation (6)

And winning requester set

Social welfare Wel_l(B,S,ρ_l)：

Step 2.9, assigning L +1 to L, judging whether L is greater than L, and if so, indicating that social benefits corresponding to L settlement price groups are obtained; otherwise, returning to the step 2.2;

step 2.10, obtaining settlement price group set by using formula (7)

Selecting the l-th settlement price group rho_lProbability of (p) Pr (p)_l) Thereby obtaining a probability distribution { Pr (ρ) of the settlement price group_l) 1,2, …, L, and according to the probability distribution of the settlement price group { Pr (ρ |)_l) 1,2, …, L, randomly selecting a settlement price group as the final winning settlement price group ρ_win：

In the formula (7), ε represents a privacy budget, and ρ' represents a settlement price group set

Any one of the settlement price groups, Wel ' (B, S, ρ ') represents a value of social welfare calculated when the settlement price group is ρ ', Δ Wel represents sensitivity, and Δ Wel ═ m × v ″_max-c_min(ii) a Privacy protection can be realized by randomly selecting the settlement price group by using an index mechanism, and the higher the social welfare generated at the same time, the higher the probability that the corresponding settlement price group is selected is;

step 3, the final winning settlement price group rho_winCorresponding set of winning workers

And winning requester set

As a final winner set, determining workers corresponding to each task;

step 4, final winning worker set

Wherein the ith winning worker u submits the perception data to the cloud platform_iThe submitted set of sensory data is noted as

d_i,jIndicating the ith winning worker u_iPerforming the jth task τ_jThe perceptual data that is generated at the time of the generation,

represents the size of the final winning worker set, j 1,2_i|，|Γ_iI denotes the ith winning worker u_iThe size of the set of tasks of interest; (ii) a

Cloud platform computing final winning requester set using equation (8)

The j winning requester r_jAggregated results of tasks of (1)

And sent to the corresponding j-th winning requester r_j：

In the formula (8), θ_i,jIndicating that the ith worker performed the jth task τ_jAnd theta_i,j＝Pr[d_i,j＝D_j]∈[0,1]，D_jRepresenting tasks τ_jTrue perception data of (1), Pr [ d ]_i,j＝D_j]Indicating that the ith worker performed task τ_jTime-committed data and task τ_jTrue perception data D of time_jThe probability of being equal is determined,

represents a winning requester set size; theta_i,jThe higher the data quality provided, the greater the weight taken up in the polymerization process;

step 5, when the final winning requester set

After each winning requester receives the data, the cloud platform according to the final winning settlement price group rho_winCollecting consideration from all winning requestors

And will award

Paying to the set of final winning workers

All winning workers.

In summary, the method for mobile crowd sensing task allocation provided by the invention protects the bid information of the requester and the worker by combining an exponential mechanism in differential privacy, and simultaneously selects the data of the worker with high reliability in the task allocation process to calculate the aggregation result by adopting a data aggregation method in consideration of the data quality of the worker, so as to ensure the accuracy of the data.

Claims (3)

1. A mobile crowd sensing task allocation method with differential privacy protection is characterized by being applied to a cloud platform, and m requesters m ═ r₁,r₂,...,r_j,...,r_mN worker Ν ═ u₁,u₂,...,u_i,...,u_nIn a mobile crowd sensing environment, where r_jDenotes the jth requestor, u_iRepresenting the ith worker, j is 1,2, …, m, i is 1,2, …, n, and the mobile crowd sensing task allocation method is carried out according to the following steps:

step 1, establishing an auction model and initializing:

step 1.1, the cloud platform is used as an auctioneer to collect task information submitted by m requesters, wherein the jth requester r_jIs recorded as S_j＝＜τ_j,s_j＞，τ_jDenotes the jth requester r_jRequested aware task, s_jIndicating completion of the perception task τ by the jth requester_jAnd s, and_j∈Π^s＝[v_min,v_max]，v_minminimum value, v, representing all requesters' quotes_maxIndicating all requestersMaximum value of quoted price of |, n^sThe value space set representing the quotation, thereby obtaining m task information sets S ═ { S } provided by the requesters₁,S₂,...,S_j,...,S_mAnd its corresponding perceptual task set T ═ τ₁,τ₂,...,τ_j,...,τ_m}；

Step 1.2, the cloud platform issues the collected perception task set T to n workers, wherein the ith worker u_iThe submitted request information is marked as B_i＝＜Γ_i,b_i＞，Γ_iIndicates the ith worker u_iA set of tasks of interest, an

b_iIndicates the ith worker u_iExecuting task set Γ_iAsking price of all tasks in, and b_i∈Π^b＝[c_min,c_max]，c_minMinimum value representing the asking price of all workers, c_maxMaximum value, Π, representing the asking price for all workers^bValue space set representing the price to be charged, thereby obtaining request information set BETA ═ { B) submitted by n workers₁,B₂,...,B_i,...,B_n}；

Step 1.3, define settlement price group set of requester and worker

Wherein the content of the first and second substances,

represents the l-th settlement price group,

indicating the ultimate award to the winning worker,

indicating the final compensation paid by the winning requester, L TableShowing the number of price groups for settlement, and L ═ Π^b×Π^sAnd has: II type^b×Π^s＝{(p_b,p_s)|p_b∈Π^b∧p_s∈Π^s}；

Step 2, according to the task information set S and the request information set BETA submitted by the requester and the worker, the cloud platform determines the settlement price group rho of the winning requester and the winning worker by using an index mechanism_win；

Step 2.1, initializing 1;

step 2.2, the l-th settlement price group rho_lPrice of (1)

Determining that the ask does not exceed the ask among the n workers

Set of workers Ν_lAnd determine set of workers Ν_lA candidate task set T formed by task sets requested by all workers_l；

Step 2.3, from set of workers Ν_lSequentially selecting the workers providing the best technical level until the candidate task set T_lAll tasks in (b) meet the quality of service requirements, resulting in a set of candidate workers W_{B_l}And a set of candidate requestors W_{S_l}；

Step 2.4, the cloud platform from candidate worker set W_{B_l}To a candidate task set T_lTask j (j) in (1)_jAnd constitutes the jth task τ_jWorker set C of_{j_l}，j＝{1,2,...,|T_l|}，|T_lI represents a candidate task set T_lThe size of (d);

step 2.5, the l settlement price group rho_lPrice of (1)

Determining a set of candidate requestors W_{S_l}Wherein the requester's bid is not less than

Winning requester set of

And corresponding task set

By the task collection

The set of workers for all tasks in the set of winning workers constitutes a winning set of workers

Step 2.6, judge in winning worker set

And winning requester set

Platform revenue generated

If yes, directly executing the step 2.8; otherwise, the platform profit is negative, and step 2.7 is executed;

step 2.7, sequentially deleting the requesters and corresponding workers having the greatest impact on negative platform revenue values, thereby updating the winning set of workers

And winning requester set

Up toThe platform profit is not less than 0, so as to obtain the final updated winning worker set and winning requester set and respectively assign the winning worker set and the winning requester set to

And

step 2.8, calculate the set of winning workers using equation (1)

And winning requester set

Social welfare Wel_l(B,S,ρ_l)：

Step 2.9, assigning L +1 to L, judging whether L is greater than L, and if so, indicating that social benefits corresponding to L settlement price groups are obtained; otherwise, returning to the step 2.2;

step 2.10, obtaining settlement price group set by using the formula (2)

Selecting the l-th settlement price group rho_lProbability of (p) Pr (p)_l) Thereby, a probability distribution { Pr (ρ L) | L ═ 1,2, …, L } of the settlement price group is obtained, and the probability distribution { Pr (ρ L) | L of the settlement price group is obtained from the settlement price group_l) 1,2, …, L, randomly selecting a settlement price group as the final winning settlement price group ρ_win：

In the formula (2), ε represents a privacy budget, and ρ' representsSettlement price group aggregation

Any one of the settlement price groups, Wel ' (B, S, ρ ') represents a value of social welfare calculated when the settlement price group is ρ ', Δ Wel represents sensitivity, and Δ Wel ═ m × v ″_max-c_min；

Step 3, the final winning settlement price group rho_winCorresponding set of winning workers

And winning requester set

As a final set of winners;

step 4, final winning worker set

Wherein the ith winning worker u submits the perception data to the cloud platform_iThe submitted set of sensory data is noted as

d_i,jIndicating the ith winning worker u_iPerforming the jth task τ_jThe perceptual data that is generated at the time of the generation,

represents the size of the final winning worker set, j 1,2_i|，|Γ_iI denotes the ith winning worker u_iThe size of the set of tasks of interest;

the cloud platform calculates a final winning requester set using equation (3)

The j winning requester r_jAggregated results of tasks of (1)

And sent to the corresponding j-th winning requester r_j：

In the formula (3), θ_i,jIndicating that the ith worker performed the jth task τ_jAnd theta_i,j＝Pr[d_i,j＝D_j]∈[0,1]，D_jRepresenting tasks τ_jTrue perception data of (1), Pr [ d ]_i,j＝D_j]Indicating that the ith worker performed task τ_jData submitted at time d_i,jAnd task tau_jTrue perception data D of time_jThe probability of being equal is determined,

represents a winning requester set size;

step 5, when the final winning requester set

After each winning requester receives the data, the cloud platform according to the final winning settlement price group rho_winCollecting consideration from all winning requestors

And will award

Paying to the set of final winning workers

All winning workers.

2. The differential privacy protection mobile crowd sensing task distribution method according to claim 1, wherein the step 2.3 is performed as follows:

step 2.3.1, set of slaves Ν_lSelecting workers with technical level satisfying formula (4) as workers capable of letting task tau_jCandidate workers meeting quality of service requirements;

in the formula (4), θ_i,jIth worker performs task τ_jβ_jE (0,1) represents the jth task tau_jAccuracy requirement of, W_{B_l}Representing a set of candidate workers;

step 2.3.2, let q_i,j＝(2θ_i,j-1)²Indicates the ith worker u_iFor jth task τ_jThe quality of service of (a) is,

denotes the jth task τ_jQ ═ q, then_i,j]^n×mQuality of service matrix representing n workers for m tasks, Q ═ Q_j]^m×1Representing the quality of service requirement matrix of m tasks, equation (4) is simplified to equation (5):

in the formula (5), x_iIndicates the ith worker u_iWhether winning, if winning, it is 1, otherwise it is 0, y_jDenotes the jth requester r_jWhether the card wins, if so, the card is 1, otherwise, the card is 0;

step 2.3.3, initialization

Step 2.3.4, set of slaves Ν_lSelecting a worker u satisfying the formula (6)_maxAs candidate workers, and from the set of workers Ν_lMiddle delete worker u_maxPost-joining to the set W_{B_l}In order to update the set of workers Ν_l：

Step 2.3.5, assign value to set of updated workers to Ν_lIs mixing Q with_j-min{Q_j,q_max,jAssign a value to Q_j(ii) a Wherein q is_max,jRepresent worker u_maxFor jth task τ_jQuality of service of (2);

step 2.3.6, judgment

If yes, returning to the step 2.3.4; otherwise, the representation gets a candidate worker set W_{B_l}And a set of candidate workers T_lCorresponding candidate requester set W_{S_l}。

3. The differential privacy protection mobile crowd sensing task distribution method according to claim 1, wherein the step 2.7 is performed as follows:

step 2.7.1, gather from winning requesters

In which the kth requester r is removed_kGet the kth set of requesters

Thereby obtaining

A set of requesters

Representing a winning requester set

Size;

step 2.7.2, obtaining the kth requester set by using the formula (7)

Corresponding kth worker set

Thereby obtaining

A set of workers

Step 2.7.3, determine the requester r that has the greatest impact on the negative platform revenue value using equation (8)_max：

Step 2.7.4, gather from winning requestors

Request for deletionPerson r is sought_maxGet the set of requesters

And corresponding set of workers

And respectively assigned with

And

step 2.7.5, judging whether the platform income is less than 0, if so, returning to step 2.7.1, otherwise, indicating that a winning requester set is obtained

And winning worker set

Images