Efficient and credible bilateral privacy protection truth discovery method

Abstract

The invention discloses an efficient and credible bilateral privacy protection truth discovery method, and aims to realize accurate truth discovery on the premise of protecting privacy of workers and data requesters. According to the method, a privacy protection mechanism based on matrix encryption is adopted, all links of true phase discovery are penetrated, and it is ensured that sensitive information of workers and data requesters is not leaked while high-quality true phase values are obtained. Besides, the property of the matrix trace is ingeniously utilized, matching of workers and tasks, trust-based worker recruitment, truth value calculation and trust updating under privacy protection are achieved, the accuracy of truth finding is effectively improved, and unnecessary worker participation cost is reduced. On the whole, a safe and efficient solution is provided for data submission and reliable data aggregation in the mobile crowd sensing system.

Description

Technical Field

[0001] The present invention belongs to the field of data collection and truth discovery in mobile crowd-sensing, and in particular to bilateral privacy protection between participants and data requesters in the truth discovery process. Background Art

[0002] In recent years, with the rapid development of communication technology and perception technology, mobile devices equipped with sensors such as global positioning systems, cameras and compasses, such as smartphones, health bracelets and tablets, have become increasingly popular, which has triggered the rise of a new perception paradigm - Mobile Crowd Sensing (MCS). MCS uses the perception, communication and computing capabilities of widely distributed mobile devices to implement data collection and processing. Such mobile devices can break through the limitations of time and space with the help of communication technologies such as 6G networks, WIFI, and Bluetooth, and transmit their perceived data or processed results to the MCS server. MCS is a fusion of crowdsourcing perception and mobility concepts. The system is open to all mobile devices and brings together the power of the crowd, that is, it allows tasks that are usually difficult for a single computer or a group of people to be outsourced to a group of uncertain mobile devices to complete separately, and aggregates the perception results submitted by each device. Traditional perception methods require the deployment of static perception devices at specific locations to achieve data collection, but the MCS system is different.

[0003] Generally speaking, MCS is a location-aware crowdsourcing technology, that is, it requires participants to move to a specific location to complete the corresponding task. In addition, the tasks in MCS can be perception tasks related to a specific location within a specific time period. For example, monitoring air quality, monitoring urban and rural noise, reporting traffic flow, updating the number of natural disasters, etc. MCS can provide a cost-effective method for the collection and aggregation of perception data. For example, for a real-time traffic congestion map construction scenario that a data requester is interested in, the MCS system collects traffic data submitted by smart vehicles with perception and communication capabilities, integrates the information, and generates a traffic congestion map to provide it to the data requester. This is more cost-effective than traditional solutions because MCS does not require additional effort to install and maintain traffic infrastructure at specific task locations.

[0004] MCS can complete data collection in a cheaper and more flexible way, but it still faces many severe tests in terms of data quality reliability and privacy protection. For the MCS system, data quality is a very important core evaluation indicator. The MCS system uses data provided by a large number of participating devices to perceive the target task. The data quality directly determines the accuracy of the target task completion and the reliability of the perception results. Due to the differences in the collection equipment and collection environment of mobile participants, the data source obtained by the MCS system has inevitable deviations. On this basis, due to the large number of mobile participants, dishonest behaviors such as data falsification are inevitable. If the impact of such malicious human error behaviors on data quality cannot be reduced, the system will inevitably misjudge the results of the perception task. For example, the prediction deviation of meteorological data will lead to inaccurate and untimely predictions of severe weather, affecting the daily life of residents and even causing casualties. Therefore, the participation of high-quality participants is crucial to obtaining high-quality aggregation results.

[0005] When performing MCS task perception, the system needs to collect a large amount of perception data information from mobile participants for data aggregation. Once these unprocessed raw data are separated from the collection device, their content and control are no longer dominated by the participants. The adversary may conduct long-term information tracking of participants based on the data submitted by the participants, such as location, health information, and occupational information, and may even threaten the lives and property of users. In addition, the perception results of the participants not only contain the sensitive information of the participants but also may contain sensitive information related to the task published by the data requester, such as store sales, museum exhibition materials, and park environmental data. The leakage of participants' sensitive data and the leakage of sensitive task information published by data requesters will seriously reduce the enthusiasm of participation, making it difficult for the MCS system to operate healthily and continuously. Considering the privacy protection of participants' sensitive data, the data quality of MCS faces more severe challenges. Most of the solutions for privacy protection of data submitted by participants need to sacrifice data availability, which avoids the system from collecting and analyzing participants' sensitive information, but at the same time creates obstacles for timely and accurate evaluation of participants' trust, and it is difficult for the MCS system to evaluate data quality results. Therefore, there is an urgent need for an efficient and reliable truth discovery method based on bilateral privacy protection. Summary of the invention

[0006] The present invention discloses an efficient and reliable truth discovery method with bilateral privacy protection. The purpose is to obtain high-quality truth values ​​under the premise of protecting the privacy of workers and data requesters. In addition to the differences in the perception devices themselves, which cause different workers to have deviations in the perception results of the same task, the false data submitted by malicious workers will further affect the quality of the aggregated results of the final task. In addition, considering the needs of workers and data requesters for privacy protection, some schemes protect the privacy of both parties by adding noise to reduce computing and communication overhead, but the addition of noise affects the availability of data and the quality of perception results. In order to protect the privacy of workers and data requesters without affecting the quality of data, the present invention introduces a privacy protection method based on matrix encryption into the entire truth discovery process, and excludes the data submitted by malicious workers when calculating the truth value.

[0007] The technical solution of the invention is as follows:

[0008] In mobile crowd sensing, we use widely distributed mobile workers to complete large-scale location-based mobile crowd sensing tasks. In order to protect the location and perception data privacy on the worker side and the task privacy on the data requester side, the present invention adopts a random matrix multiplication encryption method to protect their privacy. Based on the encrypted data of both parties, the platform uses the characteristics of the lower triangular matrix trace to carry out privacy-preserving worker-task matching, adaptive worker recruitment, truth discovery under privacy protection, and trust update. In addition, throughout the truth discovery process, the platform gradually updates and identifies the real trust status of the workers, excludes the data of malicious workers, and thus improves the quality of task aggregation results. The specific steps of this efficient and reliable bilateral privacy-preserving truth discovery method are as follows:

[0009] (1) Step 1: At the beginning, the data requester, worker, and platform send registration requests to the trusted authority respectively; once the legitimacy of these entities is verified, the trusted authority will and Send to workers, key and Sent to the data requester, and keys K1, K2, K3, K4, K5 are sent to the platform; among them, K5=M1×L3×M2, K1, K2, K3, K4, M1, M2 are 6 reversible matrices with rows and columns of (l×w+1); L1, L2, L3 are random lower triangular matrices with 3 main diagonal elements all being 1;

[0010] (2) Step 2: To facilitate the representation of the locations of workers and tasks, the entire network is divided into l×w squares; worker w i According to the square position where it is located, a (l×w+1)-dimensional vector w is constructedi,0 ={(1-l i,0 ),(1-l i,1 ),…,(1-l i,lw-1 ),0}, and construct a random lower triangular matrix W with the vector as the main diagonal, with rows and columns of (l×w+1) i,0 , the expression is shown in (1); where l i,k =1(or0) means w i Is it located in square G? k , r i,k is a random number; to protect the worker's location information from being stolen, the worker uses the key received from the trusted authority and To W i,0 Encryption And the encrypted information E i [W i,0 ] to the platform;

[0011]

[0012] Similarly, for each task t that a data requester is interested in, j Construct a random lower triangular matrix T at the location j , the main diagonal element is t j = {l j,0 ,l j,1 ,…,l lw-1 ,r2}, where l j,k =1(or0) indicates whether the task is located in grid G k , r2 is a random number; next, using the key received from the trusted authority and To T j Encryption And transmit the encrypted data to the platform;

[0013] (3) Step 3: The platform first receives encrypted messages from workers and data requesters. and Re-encryption is performed to reduce communication and computing overhead. The re-encryption operation is as follows:

[0014]

[0015] Next, the platform uses the encrypted messages between workers and data requesters to match workers with tasks and determine the many-to-many relationship between workers and tasks. Here, we use the trace function of the lower triangular matrix for matching, which is expressed as follows:

[0016]

[0017]

[0018] If α i,j = 0, indicating that worker w i Can complete the task j ;

[0019] (4) Step 4: Based on the information about workers and tasks received and the assessed trustworthiness of workers, the platform recruits a group of suitable workers to complete the task. The specific recruitment method is as follows:

[0020] ① Recruiting highly trusted workers: First, a greedy approach is used to select workers with high trust and a large number of completed tasks from the identified highly trusted workers set each time, until all highly trusted workers are recruited, or the newly added marginal contribution is 0, or the number of recruited workers reaches the threshold;

[0021] ② Recruiting workers with unknown credibility: Next, a greedy method is used to select workers with high credibility, a large number of newly completed tasks, high verification rewards, and a small number of recruitments from workers with unknown credibility each time, until all workers with unknown credibility are recruited or the newly added edge contribution is 0 or the number of recruited workers reaches the threshold;

[0022] Finally, the platform notifies the recruits to submit their perception data;

[0023] (5) Step 5: Similarly, to prevent the leakage of their perception data, the workers use a method similar to position encryption to construct the perception data into a matrix W i,2 And encrypted as E i [W i.2 ] to the platform; in order to prevent the truth value of the task from being stolen by other entities, the data requester sends the noise matrix T a,0 Encrypted to E a [T a,0 ] is transmitted to the platform, so that the truth value of the task can only be obtained by the data requester;

[0024] (6) Step 6: In order to ensure the accuracy of the truth value obtained, for tasks completed by highly trusted workers, the platform only uses the data submitted by highly trusted workers to construct the truth value; for tasks not completed by highly trusted workers, the truth value is constructed using the data submitted by workers with unknown trust. Specifically, the platform first re-encrypts the encryption matrix sent by the worker and the data requester, and then calculates the weighted perturbation truth value of the trust and perception data based on the characteristics of the lower triangular matrix trace using the encrypted information on the worker and the data requester. For tasks completed by trusted workers, the perturbed truth value and the perturbed truth value of the task completed by no trusted worker As shown in formula (5) and (6) respectively:

[0025]

[0026] in, Indicates that task t can be completed at time p j The set of workers R p is the set of workers recruited at time p; For workers i The trust level at time p-1, Γ Δ =0.9 is the trust value threshold of highly trusted workers, is the trust value threshold of malicious workers; s k is the noise given by the data requester; d i,k For workers i Perception Grid G k data;

[0027] Next, the platform weights the perturbed trust-aware data with the truth value or and the sum of the trust values ​​of the corresponding trusted workers or the sum of the trust values ​​of workers whose trust is unknown Sent to the data requester;

[0028] (7) The data requester can obtain the original truth value based on the sum of the received perturbation truth value and the trust value and the noise value generated by itself as follows:

[0029]

[0030] Compared with previous methods, the truth discovery method based on bilateral privacy protection of the present invention can improve data quality and reduce the cost of recruiting workers.

[0031] Beneficial Effects

[0032] The present invention discloses an efficient and trusted bilateral privacy-preserving truth discovery method, which adopts the ETBP-TD (Efficient and Trusted Bilateral Privacy-Preserving Truth Discovery) strategy. The invented method cleverly utilizes the properties of matrix traces, and can achieve matching of workers and tasks, worker recruitment, truth value calculation, and trust update operations while protecting the privacy of both parties. In addition, the invented method can identify the trust status of workers and eliminate malicious workers, thereby improving the quality of truth discovery. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 A system flow chart of the method proposed by the present invention is shown;

[0034] Figure 2 The evaluation errors of five different methods with different numbers of workers are shown;

[0035] Figure 3 The task completion quality of five different methods with different numbers of workers is shown. DETAILED DESCRIPTION

[0036] In order to facilitate the understanding of the present invention, the present invention will be described in more comprehensive and detailed manner below in conjunction with the accompanying drawings and preferred embodiments of the specification, but the protection scope of the present invention is not limited to the following specific embodiments.

[0037] Unless otherwise defined, all the professional terms used below have the same meanings as those generally understood by those skilled in the art. The professional terms used herein are only for the purpose of describing specific embodiments and are not intended to limit the scope of protection of the present invention.

[0038] Unless otherwise specified, various raw materials, reagents, instruments and equipment used in the present invention can be purchased from the market or prepared by existing methods.

[0039] Example:

[0040] In large-scale scenarios based on location and perception, including environmental monitoring (such as noise map construction, air pollution regulation, water quality assessment, disaster monitoring), traffic flow management, and medical big data acquisition, the data perceived by widely distributed mobile workers plays a huge role. However, if the data requester directly submits the task information of interest to the platform, the platform may infer the privacy of the data requester based on the task information. Similarly, if the worker directly submits his location information and perceived data, the platform will still infer the worker's sensitive data through the existing information, thereby reducing the worker's willingness to participate. Therefore, protecting privacy is crucial to improving the participation enthusiasm of both parties and maintaining the healthy operation of the entire system. In addition, in an open network environment, the presence of malicious workers may affect the quality of the aggregated results of the task. Therefore, it is crucial to identify the real trust status of the workers and exclude malicious workers to improve the quality of task completion for the healthy development of the system.

[0041] Figure 1The system flow chart of the method proposed in the present invention is shown. Workers and data requesters use matrix encryption to encrypt their sensitive information and upload it to the platform. The platform first re-encrypts the received encrypted message to reduce the computational complexity. Next, the platform uses the properties of matrix traces to match workers with tasks, recruit workers, calculate truth values, and update trust without access to the original data, thereby accurately identifying the trust of workers and obtaining high-quality truth values ​​while protecting the privacy of both parties.

[0042] Figure 2 The evaluation errors of the system when using the ETBP-TD method proposed in this invention and the comparative methods α-optimal, EPTD, ∈-first, and random methods under different numbers of workers are given. Among them, α-optimal is the optimal solution under all known conditions; EPTD means that the platform recruits all workers to complete the task and uses the CRH method for truth value calculation; ∈-first means randomly recruiting workers in the first ∈T cycles, and then using a greedy method to recruit workers; the random method is to randomly recruit workers. Figure 2 It can be seen that as the number of workers increases, the evaluation errors of the truth value of these methods will decrease, and the method proposed by the present invention is closest to the optimal value. This is because the present invention can gradually identify the true trust status of workers, and can eliminate malicious workers and exclude false data submitted by malicious workers during worker recruitment and truth value calculation, thereby reducing the evaluation error. In general, ETBP-TD can reduce the evaluation error by 60.96%-73.77%, 21.30%-26.85%, and 27.50%-36.69% compared to EPTD, ∈-first, and random methods, respectively.

[0043] Figure 3 The task completion quality of the system using the ETBP-TD method proposed by the present invention and the comparative methods α-optimal, EPTD, ∈-first, and random methods under different numbers of workers is given. It can be seen that as the number of workers increases, the quality of the workers' task completion increases, and our method is closest to the optimal value. Our method can reach 95.30% of α-optimal. In general, ETBP-TD can improve the task completion quality by 2.79%-49.48% compared with EPTD, ∈-first, and random methods.

Claims

1. An efficient and reliable bilateral privacy-preserving truth discovery method, characterized by the following steps: (1) Step 1: At the beginning, the data requester, worker, and platform send registration requests to the trusted authority respectively; once the legitimacy of these entities is verified, the trusted authority will and Send to workers, key and Sent to the data requester, and keys K1, K2, K3, K4, K5 are sent to the platform; among them, K5=M1×L3×M2, K1, K2, K3, K4, M1, M2 are 6 reversible matrices with rows and columns of (l×w+1); L1, L2, L3 are random lower triangular matrices with 3 main diagonal elements all being 1; (2) Step 2: To facilitate the representation of the locations of workers and tasks, the entire network is divided into l×w squares; worker w i According to the square position, a (l×w+1)-dimensional vector w is constructed i,0 ={(1-l i,0 ),(1-l i,1 ),…,(1-l i,lw-1 ),0}, and construct a random lower triangular matrix W with the vector as the main diagonal, with rows and columns of (l×w+1) i,0 , the expression is shown in (1); where l i,k =1(or 0) means w i Is it located in square G? k , r i,k is a random number; to protect the worker's location information from being stolen, the worker uses the key received from the trusted authority and To W i,0 EncryptionE i [W i,0 ]= And the encrypted information E i [W i,0 ] to the platform; Similarly, for each task t that a data requester is interested in, j Construct a random lower triangular matrix T at the location j , the main diagonal element is t j = {l j,0 ,l j,1 ,…,l lw-1 ,r2}, where l j,k =1 (or 0) indicates whether the task is located in square G k , r2 is a random number; next, using the key received from the trusted authority and To T j Encryption And transmit the encrypted data to the platform; (3) Step 3: The platform first receives encrypted messages from workers and data requesters. and Re-encryption is performed to reduce communication and computing overhead. The re-encryption operation is as follows: Next, the platform uses the encrypted messages between workers and data requesters to match workers with tasks and determine the many-to-many relationship between workers and tasks. Here, we use the trace function of the lower triangular matrix for matching, which is expressed as follows: If α i,j = 0, indicating that worker w i Can complete the task j ; (4) Step 4: Based on the information about workers and tasks received and the assessed trustworthiness of workers, the platform recruits a group of suitable workers to complete the task. The specific recruitment method is as follows: ① Recruiting highly trusted workers: First, a greedy approach is used to select workers with high trust and a large number of completed tasks from the identified highly trusted workers set each time, until all highly trusted workers are recruited, or the newly added marginal contribution is 0, or the number of recruited workers reaches the threshold; ② Recruiting workers with unknown credibility: Next, a greedy method is used to select workers with high credibility, a large number of newly completed tasks, high verification rewards, and a small number of recruitments from workers with unknown credibility each time, until all workers with unknown credibility are recruited or the newly added edge contribution is 0 or the number of recruited workers reaches the threshold; Finally, the platform notifies the recruits to submit their perception data; (5) Step 5: Similarly, to prevent the leakage of their perception data, the workers use a method similar to position encryption to construct the perception data into a matrix W i,2 And encrypted as E i [W i.2 ] to the platform; in order to prevent the truth value of the task from being stolen by other entities, the data requester sends the noise matrix T a,0 Encrypted to E a [T a,0 ] is transmitted to the platform, so that the truth value of the task can only be obtained by the data requester; (6) Step 6: In order to ensure the accuracy of the truth value obtained, for tasks completed by highly trusted workers, the platform only uses the data submitted by highly trusted workers to construct the truth value; for tasks not completed by highly trusted workers, the truth value is constructed using the data submitted by workers with unknown trust. Specifically, the platform first re-encrypts the encryption matrix sent by the worker and the data requester, and then calculates the weighted perturbation truth value of the trust and perception data based on the characteristics of the lower triangular matrix trace using the encrypted information on the worker and the data requester. For tasks completed by trusted workers, the perturbed truth value and the perturbed truth value of the task completed by no trusted worker As shown in formula (5) and (6) respectively: in, Indicates that task t can be completed at time p j The set of workers R p is the set of workers recruited at time p; For workers i The trust level at time p-1, Γ Δ =0.9 is the trust value threshold of highly trusted workers, is the trust value threshold of malicious workers; s k is the noise given by the data requester; d i,k For workers i Perception Grid G k data; Next, the platform weights the perturbed trust-aware data with the truth value or and the sum of the trust values ​​of the corresponding trusted workers or the sum of the trust values ​​of workers whose trust is unknown Sent to the data requester; (7) The data requester can obtain the original truth value based on the sum of the received perturbation truth value and the trust value and the noise value generated by itself as follows:

Images