Spatial crowdsourcing position privacy protection algorithm based on R tree and Monte Carlo method

Abstract

The invention designs a spatial crowdsourcing position privacy protection algorithm based on an R-tree and a Monte Carlo algorithm, and belongs to the field of Internet of Things in the computer field. Spatial crowdsourcing usually requires spatial crowdsourcing participants to provide accurate position information to a server, but this may cause leakage of position privacy of the participants. Therefore, the space crowdsourcing task allocation technology with the privacy protection capability is a technology which is urgently needed to be researched at present. The existing research provides the same position privacy protection level for each worker, and cannot meet the requirements of different workers. In addition, an existing research assumes that a task submitted by a task publisher is a single-position task, and a complex multi-position task is not considered. Therefore, the invention provides a complex multi-position task allocation algorithm for personalized position privacy protection, a candidate worker set is quickly obtained by using an R-tree, and an accessibility calculation algorithm based on a Monte Carlo algorithm is provided to realize efficient complex multi-position task allocation.

Description

Technical Field

[0001] The invention designs a spatial crowdsourcing location privacy protection algorithm based on R-tree and Monte Carlo algorithm, and belongs to the field of Internet of Things in the field of computer. Background Art

[0002] With the rapid development of wireless networks and the continuous enhancement of the communication and computing capabilities of mobile devices and sensors, a new form of crowdsourcing has emerged: spatial crowdsourcing [1]. Spatial crowdsourcing effectively utilizes the large number of users brought about by this development trend and has very important advantages in smart cities, news reporting, etc. In the spatial crowdsourcing platform, both workers and task publishers will register with the spatial crowdsourcing server to ensure the service quality of the platform and the privacy of users. Compared with traditional wireless sensor networks (WSNs), spatial crowdsourcing has many advantages. By utilizing widely distributed mobile devices for data collection and processing, it saves the additional cost of installing and maintaining new hardware infrastructure, and can also provide a wider coverage than WSNs. Therefore, spatial crowdsourcing has been widely studied and applied in academia and industry. For example, the CarTel system [3] is a spatial crowdsourcing network that uses car sensors to collect and transmit traffic patterns.

[0003] In order to efficiently implement task allocation, workers are usually required to disclose their location information to the server and task publishers. Then, the spatial crowdsourcing server matches tasks with workers according to the needs of the tasks. However, sometimes the server is not completely trustworthy. Based on the location information uploaded by the workers, the server and other spatial crowdsourcing participants can infer sensitive information such as the user's identity and address. Once leaked by an untrustworthy server, the user's privacy and security will be seriously threatened [5]. Attackers can use the above information to launch attacks, such as monitoring, tracking, identity theft, selling sensitive information, etc. This may cause workers to be unwilling to upload their real location information or even quit the spatial crowdsourcing service, affecting the normal operation of the platform. Therefore, protecting location privacy is crucial for spatial crowdsourcing.

[0004] There have been some relevant studies on location privacy protection, but most of the studies focusing on location privacy protection do not set different privacy levels for different workers, resulting in some workers being insufficiently protected and others being overly protected. Secondly, most of the current studies that consider location privacy protection only study single-location task scenarios, and there is no relevant research on spatial crowdsourcing complex multi-location task allocation technology with personalized location privacy protection capabilities.

[0005] This paper proposes a complex multi-location task allocation algorithm with personalized location privacy protection, which can effectively allocate suitable workers to complex multi-location tasks while protecting the location privacy of workers. The algorithm uses R-tree to quickly obtain a set of candidate workers, and proposes a reachability calculation algorithm based on Monte Carlo algorithm to achieve efficient complex multi-location task allocation

[0006] [1]Kazemi L, Shahabi C.Geocrowd: enabling query answering with spatialcrowdsourcing[C] / / Proceedings of the 20th international conference onadvances in geographic information systems.2012: 189-198.

[0007] [2]Wang Y, Yan Z, Feng W, et al.Privacy protection in mobile crowdsensing: a survey[J].World Wide Web, 2020, 23(1): 421-452.

[0008] [3]Hull B, Bychkovsky V, Zhang Y, et al.Cartel: a distributed mobilesensor computing system[C] / / Proceedings of the 4th international conference on Embedded networked sensor systems.2006: 125-138.

[0009] [4]Tong Y, She J, Ding B, et al.Online mobile micro-task allocation inspatial crowdsourcing[C] / / 2016 IEEE 32Nd international conference on dataengineering(ICDE).IEEE, 2016: 49-60.

[0010] [5]Feng W, Yan Z, Zhang H, et al.A survey on security, privacy, and trustin mobile crowdsourcing[J]. IEEE Internet of Things Journal, 2017, 5(4): 2971-2992. Summary of the invention

[0011] The present invention aims to solve the following technical problems:

[0012] There are some security and privacy threats in spatial crowdsourcing. For example, the task uploading and assignment process may leak some sensitive task information to malicious workers. Secondly, workers also need to upload some personal information, which may affect the worker's information privacy. Therefore, if the server cannot be fully trusted, private information should not be disclosed to it. In addition to security and privacy issues, trust in the assignment process must also be ensured. For example, workers may deliberately upload forged information to the server in order to ensure that their real personal information will not be leaked.

[0013] There are some defects in the existing technical solutions. First, most of the research focusing on location privacy protection does not set different privacy levels for different workers, resulting in some workers being insufficiently protected and others being over-protected. Second, most of the current research considering location privacy protection only studies single-location task scenarios, and there is no relevant research on spatial crowdsourcing complex multi-location task allocation technology with personalized location privacy protection capabilities.

[0014] The present invention adopts the following technical solutions to solve the technical problems:

[0015] This invention patent proposes a spatial crowdsourcing location privacy protection algorithm based on R-tree and Monte Carlo algorithm, which includes four steps:

[0016] (1) Location obfuscation. In the location obfuscation stage, workers do not directly upload their real locations. Instead, they use the location obfuscation algorithm to convert their point locations into a circular area and upload it. The obfuscation circle has the worker’s randomly generated new location as the center and the privacy level as the radius, hiding the worker’s real location.

[0017] (2) Construct a set of candidate workers. The confusion circles of all workers obtained by the above process are stored using an R-tree. The sub-positions of each task are surrounded by a minimum enclosing rectangle, and the set of candidate workers is obtained by finding the R-tree nodes that intersect with the minimum enclosing rectangle of the task.

[0018] (3) Sorting the set of candidate workers. The spatial crowdsourcing server ranks each candidate worker according to the proximity between the worker and the task. However, since the location is obfuscated and the privacy level is different, it is difficult to estimate the distance between the task and the worker. The present invention uses a Monte Carlo algorithm for ranking and sorting.

[0019] (4) Task allocation. The spatial crowdsourcing server sends the task to the candidate worker, who then determines whether the task is within his service scope. If so, he accepts the task. Otherwise, the spatial crowdsourcing server sends the task to the next candidate worker and the process continues.

[0020] Compared with the prior art, the present invention adopts the above technical solution and has the following beneficial effects:

[0021] (1) Compared with the traditional spatial crowdsourcing location privacy protection algorithm, the algorithm proposed in this patent can set different privacy levels for different workers.

[0022] (2) The algorithm proposed in this patent has good results in complex multi-location task scenarios of spatial crowdsourcing. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is the task and worker model diagram.

[0024] Figure 2 It is a diagram of algorithm stages.

[0025] Figure 3 It is a position confusion map. DETAILED DESCRIPTION

[0026] The present invention is further described in detail below in conjunction with the accompanying drawings.

[0027] The tasks and worker models considered in this invention are as follows: Figure 1 As shown. Each task t = <st>are all complex multi-location tasks, which require workers to go to all locations and perform corresponding actions. Each location and the specified action is called a subtask st, and the location of each subtask is recorded as l st Each task contains a set of subtasks ST and is assigned to a worker w to complete. After collecting the data submitted by the workers, the server sends it to the task publisher and charges a corresponding fee. During task processing, the present invention assumes that each worker only performs a single task and all workers can accurately perform their assigned tasks, so each task needs to be assigned to one and only one worker.

[0028] The distance between worker w and task t is denoted by d(w, t). For each task t, where each subtask position l st Both with worker position l w There is a certain distance, and the minimum value in this set of distances is called the distance d(w, t) between worker w and task t, that is, d(w, t) = min({d(w, st)|st∈ST}), where min means finding the minimum value in the set.

[0029] Subtask set traversal distance uses d tra (t) indicates that the subtask set traversal distance refers to the shortest distance taken to pass through all subtask positions in the task. When calculating this distance, each point can be passed repeatedly. tra (w, t), then the starting point is l w , that is, from the starting point l w Start traversing all subtask positions in the task ST The shortest distance.

[0030] Each worker w = <l w , l′ w , d w , r w > Have smart mobile devices with sensors such as mobile phones, and can move to designated locations at will to complete subtasks in sequence. Each worker has his own position w , and they will be willing to reach the task distance d w Once a worker believes that he cannot move to any subtask position within this distance, that is, d(w, t)>d w , they will reject the task, otherwise the worker will perform the task and submit the collected data to the server, and the server will pay the corresponding salary to the worker.

[0031] The present invention focuses on the allocation of online scenarios, that is, workers will upload their locations to the server in advance. After receiving the task information published by the task publisher, the server will immediately allocate a worker who can reach and complete the task. At this time, the worker and the task constitute a worker-task pair a (a =<w,l> , and d w ≥d(w, t)), where d(w, t) represents the distance between the worker and the task. After the allocation is completed, the set of all worker-task pairs is represented by A(W, T).

[0032] Each task t contains multiple subtasks, and the locations of these subtasks are surrounded by the minimum bounding rectangle (MBR). Each worker w has a circle c(l w , d w ), using R-tree index. Nodes in the R-tree that do not intersect with the minimum bounding rectangle of the task indicate that the worker is unreachable for the task. These workers should be filtered out through pruning, and the remaining workers should be added to the candidate worker set for the task.

[0033] like Figure 1 As shown, the dotted circle represents c(l w , d w ), the solid rectangle m represents the minimum bounding rectangle of the task, the subtask positions of the same color belong to the same complex multi-position task, and the gray employees represent the positions of the employees after confusion. Figure 1 It can be seen that c 1 With m 1 、m 2 There is an intersection, c 2 With m 2 、m 3 There is an intersection, c 3 With m 2 、m 3 There is an intersection. So t 1 The set of candidate workers is w 1 , t 2 The set of candidate workers is w 1 、w 2 With w 3 , t 3 The set of candidate workers is w 2 With w 3 Since online scenario tasks arrive in a certain order, assume that the tasks arrive in t 1 ,t 2 ,t 3 The order of arrival can be 1 Assign to w 1 , t 2 Assign to w 2 , t 3 Assign to w 3 , so that the number of allocations is maximized.

[0034] In order to allocate tasks as efficiently as possible while protecting the personalized location privacy of workers, this paper proposes a spatial crowdsourcing location privacy protection algorithm based on R-tree and Monte Carlo method. The algorithm is divided into four stages: (1) location obfuscation; (2) constructing a set of candidate workers; (3) sorting the set of candidate workers; and (4) task allocation.

[0035] (1) Position confusion. Figure 3 , obfuscate the location of each worker, convert the point location into a circular area, and upload the area to the server. The basic steps of location obfuscation are as follows:

[0036] (1) Each worker determines the privacy radius r according to his or her actual wishes w At this time, the worker is assigned to his real position. w is the center of the circle, r w The circle with radius c(l w , r w ).

[0037] (2) Randomly select k points in this circle and calculate the average value of these points as the new confusion position l according to the following formula w′ . w′ is the center of the circle, r w radius, and obtain the new confusion area c(l w′ , r w ), and the confusion area c(l w′ , r w ) must contain the worker's real position l w .

[0038]

[0039] In the formula, x i ,y i Respectively represent l w The horizontal and vertical coordinates of .

[0040] (3) Workers will w′ With r w This action ensures that the server and other spatial crowdsourcing participants who obtain information from the server can only see the obfuscated location uploaded by the workers, and only the workers themselves know their real location, thereby avoiding the leakage of the workers' location privacy.

[0041] (2) Construct a set of candidate workers. For each worker, a circular area (c(l w , d w )) is indexed by using an R-tree. Then, we wait for the tasks to arrive one by one, and use the minimum enclosing rectangle to enclose the positions of each subtask in the task. We prune the leaf nodes of the R-tree that intersect with the minimum enclosing rectangle. The workers stored in these leaf nodes constitute the candidate worker set W for the task.

[0042] (3) Sort the candidate worker set. The Monte Carlo algorithm is a method that estimates the probability through a large number of simulation experiments. w′ , r w ) randomly selects K sampling points, K is the number of Monte Carlo sampling, each sampling point is denoted as l, and calculates how many points satisfy d w ≥d(l, t), the worker reachability probability is calculated using the following formula:

[0043]

[0044] In the formula, p(d(w, t)<d w ) represents the probability that a worker is reachable, K is the number of Monte Carlo sampling, N(l|d w ≥d(l, t)) indicates the number of points where the distance from the sampling points to the task is less than the worker's desired distance.

[0045] (4) Task allocation. The spatial crowdsourcing server sends task information to the worker with the highest ranking in the candidate worker set and removes it from the candidate worker set for the task. After receiving this information, the worker calculates based on his or her actual location to determine whether the distance to the task is within his or her desired range, i.e., to determine d w ≥d(w, t), if so, the worker accepts the task and will not appear in the candidate worker set for other subsequent tasks; otherwise, the worker rejects the task, and the server continues to send task location information to the candidate worker with the smallest distance in the candidate worker set and repeats the above process.< / st>

Claims

1. A spatial crowdsourcing location privacy protection algorithm based on R-tree and Monte Carlo method, characterized by The following steps: (1) Position confusion: The worker’s point position is converted into a circular area. (2) Constructing a set of candidate workers: The spatial crowdsourcing server selects a set of candidate workers that can complete the task. (3) Sorting the candidate worker set: The distance between workers and tasks is evaluated according to the Monte Carlo algorithm, and the candidate worker set is sorted. (4) Task assignment: Try to assign tasks to candidate workers in an orderly manner.

2. The position obfuscation process as claimed in claim 1, comprising the following: Each worker determines the privacy radius r according to his actual wishes w At this time, the worker is assigned to his real position. w is the center of the circle, r w The circle with radius c(l w , r w ). Randomly select k points in this circle and take the average of these points as the new confusion position l w′ . w′ is the center of the circle, r w radius, and obtain the new confusion area c(l w′ , r w ), and the confusion area c(l w′ , r w ) must contain the worker's real position l w . Workers will l w′ With r w Send to the space crowdsourcing server.

3. The process of constructing a candidate worker set as claimed in claim 1, comprising the following: For each worker, a circular area (c(l w , d w )) is indexed by using an R-tree. Then, we wait for the tasks to arrive one by one, and use the minimum enclosing rectangle to enclose the positions of each subtask in the task. We prune the leaf nodes of the R-tree that intersect with the minimum enclosing rectangle. The workers stored in these leaf nodes constitute the candidate worker set W for the task.

4. The sorted candidate worker set as claimed in claim 1, comprising the following: In the worker confusion position circle c(l w′ , r w ) randomly selects K sampling points, K is the number of Monte Carlo sampling, each sampling point is denoted as l, and calculates how many points satisfy d w ≥d(l, t), the worker reachability probability is calculated using the following formula: In the formula, p(d(w, t)<d w ) represents the probability that a worker is reachable, K is the number of Monte Carlo sampling, N(l|d w ≥d(l, t)) indicates the number of points where the distance from the sampling points to the task is less than the worker's desired distance.

5. The task allocation process as claimed in claim 1, comprising the following contents: The spatial crowdsourcing server sends the task information to the worker with the highest ranking in the candidate worker set and removes it from the candidate worker set for the task. After receiving this information, the worker calculates based on his or her actual location to determine whether the distance to the task is within his or her desired range, i.e., to determine d w ≥d(w, t), if so, the worker accepts the task and will not appear in the candidate worker set for other subsequent tasks; otherwise, the worker rejects the task, and the server continues to send task location information to the candidate worker with the smallest distance in the candidate worker set and repeats the above process.

Images