Fine-grained and high-quality privacy protection data aggregation method

Abstract

The invention discloses a fine-grained and high-quality privacy protection data aggregation method. A trust mechanism, a cloud service platform, a fog node, a task requester and a task executor are included. The trust mechanism is responsible for completing generation, distribution and management of a secret key and generation of a pseudonym of a task executor; the cloud service platform is responsible for receiving tasks issued by a task requester, uniformly collecting data and distributing the tasks and information; the fog node is responsible for encrypting privacy information of task executors, distributing selection tasks of an adaptive task executor group and collecting and uploading sensing data; the task requester is responsible for making a task request to the cloud service platform and receiving data collected by a task; the task executor is responsible for executing the task and submitting sensing data to the fog node; according to the method, high-quality data aggregation can be realized, and meanwhile, the location privacy and the data privacy of the user are protected. And the contribution of each user to a single task is considered, so that fine-grained personalized rewards are realized.

Description

Technical Field

[0001] The present invention relates to a privacy-preserving data aggregation method, and in particular to a fine-grained and high-quality privacy-preserving data aggregation method. Background Art

[0002] In mobile crowdsensing systems, truth discovery is often used in data aggregation. In practice, estimating true information from large amounts of data is crucial. Researchers typically use truth discovery to infer accurate aggregation results based on the quality of perception data aggregation. While truth discovery has been widely explored to improve aggregation accuracy, many security and privacy issues remain to be addressed. Existing solutions either fail to guarantee the privacy of each task performer or fail to consider the individual privacy needs of each task performer. They are unable to accurately refine each task performer's contribution to data quality, resulting in poor quality perception data aggregation results. Summary of the Invention

[0003] In response to the above-mentioned problems in the existing technology, the present invention provides a fine-grained and high-quality privacy-preserving data aggregation method, which takes into account the location and data privacy of the task executors while considering the contributions of the task executors, thereby improving the quality of data aggregation.

[0004] To achieve the above-mentioned object, the present invention provides the following technical solution: comprising a trust institution, a cloud service platform, fog nodes, a task requester, and a task executor; the trust institution, fog nodes, and task requester are all connected to the cloud service platform via a network, and the task executor is connected to the fog nodes via a network;

[0005] The trust agency is responsible for the generation, distribution, and management of secret keys and the generation of pseudonyms for task executors;

[0006] The cloud service platform is responsible for receiving tasks issued by task requesters, unified data collection, and distribution of tasks and information;

[0007] Fog nodes are responsible for encrypting the privacy information of task performers, selecting and assigning tasks to appropriate task performer groups, and collecting and uploading sensory data. Therefore, more than 60% of data processing will occur in fog nodes, reducing the data upload load.

[0008] The task requester is responsible for submitting task requests to the cloud service platform and receiving the data collected by the task;

[0009] The task executor is responsible for executing the task and submitting the perception data to the fog node;

[0010] The data aggregation method includes two steps: initialization phase and iteration phase;

[0011] In the initial stage, the pseudonym and secret key are generated by the trusted authority;

[0012] In the iteration phase, by continuously iterating the weights and true values ​​until the true values ​​converge, the weight of the task executor for each task is obtained under the pseudonym of the task executor, so as to aggregate high-quality data.

[0013] Furthermore, the specific steps of the initialization phase are:

[0014] The trust agency is responsible for the generation and distribution of keys. After generating the keys, the trust agency sends the public key to the task executor for encryption and sends the private key to the cloud service platform for decryption. In addition, when the task executor registers, the trust agency is responsible for assigning a pseudonym to each registered task executor.

[0015] Furthermore, the specific steps of the iteration phase are:

[0016] Define the task set as Task, Task = {t1, t2, ..., t i}, specify each task t i They are all region-related and managed by a fog node; the cloud service platform generates the initial weight of each task After that, the initial weight is sent to the task executors managed by the fog node;

[0017] S1: The fog node sends the task t i Send to task executor u k , task executor u k After completing the perception data After collecting, calculate the task executor u k The distance between the perception data and the initial true value Then the task executor u k Choose a random value r k And save it separately, Encrypted using the following formula: After encryption is completed, the task executor will and Kana pse k Upload them to the fog node together;

[0018] S2: The fog node receives the task t i After encrypting the distance data of U′ task performers, the data are aggregated according to the homomorphism of Paillier encryption:

[0019]

[0020] The aggregated data Upload to the cloud service platform;

[0021] S3: The cloud service platform obtains aggregated encrypted data Finally, the aggregated data is decrypted using the private key sk:

[0022]

[0023] The decrypted data is then returned to the task executor through the fog node;

[0024] S4: Assuming that all task executors perform calculations honestly, then task executor u j Getting the data Then update the weight:

[0025]

[0026] At this time the weight Represents the task executor u j For task t i After knowing the weight, the task executor first calculates his own perception data contribution Then it is encrypted with the public key pk And the encrypted data and the encrypted weight Upload them to the fog node together;

[0027] Similarly, when executing task t i After all U′ task executors submit their data, the sum can be calculated The summed data and the encrypted weight Submit them to the cloud service platform together;

[0028] S5: After receiving the data, the cloud service platform first decrypts the encrypted data using the private key sk:

[0029]

[0030] Then update the true value:

[0031]

[0032] After completing the truth value update, the cloud service platform sends the updated truth value to the task executors it manages through the fog nodes;

[0033] S6: Iterate the above weight update and true value update multiple times until the true value converges, then the pseudonym pse of the task executor can be obtained. k Under this circumstance, the weight of the task executor for each task is set to 0 for the tasks in which the task executor does not participate. The weight set of the task executor can be obtained as follows:

[0034]

[0035] Then, the task executor will use the pseudonym pse k and weight set It is handed over to a trust agency. The trust agency knows the reward for each task and rewards the task executors according to different weights, thus achieving fine-grained rewards and encouraging subsequent task executors to submit fine-grained data, promoting data aggregation results, and obtaining high-quality data.

[0036] Furthermore, the trust authority generates, distributes, and manages secret keys and generates pseudonyms for task performers during the initialization phase as follows:

[0037] First, the task executor u k Send an application to the trust agency, and the trust agency assigns a pseudonym pse to each task executor k ,Then the pseudonym is returned to the task executor;

[0038] Secondly, the trust agency selects a security parameter k and two large prime numbers p and q, which satisfy the condition |p|=|q|=k. Then the trust agency calculates n=pq and λ=(p-1)(q-1). After these parameters are prepared, the trust agency sends the public key pk=(n, h n ) is passed to each task executor, allowing the task executor to encrypt the data. Then, the trust organization sends the private key sk = λ = (p-1)(q-1) to the cloud service platform, allowing the cloud service platform to decrypt the data.

[0039] Furthermore, in step S1, the task executor u k The collected sensory data are divided into continuous data and discrete data;

[0040] The distance calculation formula between continuous data and the initial true value is:

[0041]

[0042] Discrete data can be represented as vectors The selected parameter is 1 and the others are 0. The distance calculation formula between discrete data and the initial true value is:

[0043]

[0044] Compared to existing technologies, this invention offers the following advantages: It utilizes pseudonym generation and public key encryption to protect the location and data privacy of task performers; it also utilizes truth-value discovery to calculate each task performer's contribution to a single task. This allows for precise selection of valid data for each task during data aggregation, compared to uniform weight calculation, improving data aggregation quality. Specifically, based on the calculated weights, precise rewards can be assigned to user performers, ensuring that their efforts are commensurately rewarded. This increases the enthusiasm of high-quality task performers for performing perception tasks, further promoting data aggregation results and improving data aggregation quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a framework diagram of the present invention;

[0046] Figure 2 Flowchart of the present invention;

[0047] Figure 3 This is a comparison chart of the weight under dimension 1 and the weight under unified dimension in an embodiment of the present invention;

[0048] Figure 4 This is a comparison chart of unified dimension rewards and separate dimension rewards in an embodiment of the present invention. DETAILED DESCRIPTION

[0049] The present invention will be further described below with reference to the accompanying drawings.

[0050] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0051] like Figure 1 As shown, the present invention provides a technical solution:

[0052] It includes the trust agency TA, the cloud service platform CSP, the fog node FN, the task requester and the task executor; the trust agency, the fog node and the task requester are all connected to the cloud service platform through the network, and the task executor is connected to the fog node through the network;

[0053] The trust agency is responsible for the generation, distribution, and management of secret keys and the generation of pseudonyms for task executors;

[0054] The cloud service platform is responsible for receiving tasks issued by task requesters, unified data collection, and the distribution of some tasks and information;

[0055] The fog nodes are responsible for encrypting the privacy information of task executors, selecting and assigning tasks to the appropriate task executor group, and collecting and uploading sensory data. Therefore, more than 60% of data processing will occur in the fog nodes FN, reducing the data upload load.

[0056] The task requester is responsible for submitting task requests to the CSP and receiving the data collected by the task;

[0057] The task executor is responsible for executing the task and submitting the perception data to the fog node.

[0058] The data aggregation method includes two steps: initialization phase and iteration phase.

[0059] Example: Figure 2 As shown:

[0060] In the initial stage, the pseudonym and secret key are generated by the trusted authority;

[0061] We assume that the trusted authority (TA) is fully trustworthy. The trusted authority (TA) is responsible for key generation and distribution, and will not collude with others to leak the private information of task performers. After generating the keys, the TA sends the public key to the task performer for encryption and the private key to the CSP for decryption. Furthermore, to protect the privacy of task performers, the TA is responsible for assigning pseudonyms to each registered task performer during registration.

[0062] First, the task executor u k Send an application to TA, TA assigns a pseudonym pse to each task executor k , and then returns the pseudonym to the task executor to protect the privacy of the task executor.

[0063] Next, the TA selects a security parameter k and two large prime numbers p and q. At this point, the condition |p|=|q|=k is satisfied, and then the TA calculates n=pq and λ=(p-1)(q-1). After these parameters are prepared, the TA uses the public key pk=(n, h n ) is passed to each task executor, who then encrypts the data. TA then sends the private key sk = λ = (p-1)(q-1) to the CSP, which then decrypts the data.

[0064] In the iteration phase, by continuously iterating the weights and true values ​​until the true values ​​converge, we can obtain the weight of each task for the task executor under the pseudonym of the task executor, and aggregate high-quality data.

[0065] Assume that there are 5 tasks in the current task set, and define the task set as Task, Task = {t1, t i ,...,t5}, ​​specify each taskt i is region-related and is managed by a fog node FN. Each node is defined to manage 24 task executors. CSP generates the initial weight of each task. After that, the initial weight is sent to the task executors it manages through the fog node FN.

[0066] Step 1: The fog node FN sends the task t i Send to task executor u k , task executor u k After completing the perception data After collecting, calculate the task executor u k The distance between the perception data and the initial true value It should be noted here that since the data collected by the task executor is divided into continuous data and discrete data, the calculation of distance for different data is different. The distance calculation formula between continuous data and the initial true value is:

[0067]

[0068] For discrete data, it can be represented as a vector The selected parameter is 1 and the others are 0. The distance calculation formula between discrete data and the initial true value is:

[0069]

[0070] Then the task executor u k Choose a random value r k And save it separately, Encrypted using the following formula:

[0071]

[0072] After encryption is completed, the task executor will and Kana pse k Upload them together to the fog node FN.

[0073] Step 2: The fog node FN receives the task t i After encrypting the distance data of U′ task performers, the data are aggregated according to the homomorphism of Paillier encryption:

[0074]

[0075] The aggregated data Upload to CSP.

[0076] Step 3: CSP obtains the aggregated encrypted data Then, the aggregated data is decrypted using the private key sk, i.e. λ = (p-1)(q-1):

[0077]

[0078] The decrypted data is then returned to the task executor through the fog node FN.

[0079] Step 4: Assuming that all task executors perform calculations honestly, then task executor u j Getting the data Then update the weight:

[0080]

[0081] At this time the weight Represents the task executor u j For task t i The weight contributed. For example, during the iteration process, the weight of task executor u1 for task t1 can be calculated. It is 2.67.

[0082] After knowing the weight, the task executor first calculates his own perception data contribution Then it is encrypted with the public key pk And the encrypted data and the encrypted weight Upload them together to the fog node FN.

[0083] Similarly, when executing task t i After all U′ task executors submit their data, we can calculate the sum The summed data and the encrypted weight Submit to CSP together.

[0084] Step 5: After receiving the data, CSP first decrypts the encrypted data using the private key sk:

[0085]

[0086] Then update the true value:

[0087]

[0088] After completing the truth value update, CSP sends the updated truth value to the task executors it manages through the fog node FN.

[0089] S6: Iterate the above weight update and true value update multiple times until the true value converges, then the pseudonym pse of the task executor can be obtained. kUnder this circumstance, the weight of the task executor for each task is set to 0 for the tasks that the task executor does not participate in. The weight set of the task executor can be obtained as follows:

[0090]

[0091] Therefore, by calculation we can get the task executor weight under task t1, such as Figure 3 As shown;

[0092] In order to facilitate comparison with existing work (unified weight calculation), we also calculate the weight under unified task calculation. In the calculation process, we regard a task as a dimension. Figure 3 As shown, among the 24 task performers, the weights calculated for Task Performers 1-4 and Task Performers 12-17 under Dimension 1 are essentially the same as those calculated under the unified dimensionality. For Task Performers 5, Task Performers 8, and Task Performers 10-11, there are slight differences between the weights calculated under Dimension 1 and the unified dimensionality, but the differences are not significant. However, for the remaining task performers, the differences between the weights calculated under Dimension 1 and the unified dimensionality are significant. Among these task performers, Task Performers 7 and 19 are particularly notable. The weights calculated under Dimension 1 are two to three times greater than those under the unified dimensionality. This means that if the weights under the unified dimensionality are used, a significant amount of valid data will be lost, resulting in lower data quality. For Task Performers 20-24, the weights calculated under the unified dimensionality are higher than those under Dimension 1. If the weights under the unified dimensionality are used, a lot of useless data will be added during data aggregation, reducing data quality. Through comparison, we confirmed that in most cases the weights calculated using the unified dimension are close to those calculated using a single dimension. However, for the data of some task executors, single-dimensional calculation can more accurately obtain the contribution weights of the task executors. In this way, during the data aggregation process, we can aggregate as much valid data as possible, filter out invalid or low-quality data, and improve the quality of aggregated data.

[0093] Then, the task executor will use the pseudonym pse k and weight set The task is handed over to the TA, who knows the reward for each task and rewards the task performers according to different weights. The corresponding reward values ​​for the five tasks in the plan are 60, 70, 80, 90 and 100 respectively. We compare the rewards obtained by the task performers under the unified dimension (existing work) with the rewards obtained by the task performers under the separate dimension calculations, as shown in the figure below. Figure 4 As shown, from Figure 4The data shows that for Taskers 8, 11, and 13-17, the difference between the two schemes is very small, with the difference within 1. However, for other Taskers, the difference in rewards between the two schemes is significant. This can be divided into two scenarios. The first scenario involves Taskers receiving rewards under the unified dimension being higher than those calculated using the separate dimensions. For example, Taskers 20-24 receive rewards from the unified dimension that are significantly higher than the total rewards calculated using the separate dimensions. This means that these Taskers receive rewards that are significantly higher than they should be. This leads them to believe that their current perception behavior is correct and to maintain this behavior in subsequent perception tasks. However, the perception data obtained from this behavior is actually of low quality. This strategy creates a vicious cycle that undermines data perception quality. For example, Taskers 7 receive rewards from Taskers under the unified dimension that are lower than those calculated using the separate dimensions. In the unified dimension calculation, they should have received a reward of 14.49, but the total reward calculated using the separate dimensions was 20.78, a difference of 6.29. For such a task performer, their high-quality data contribution should have resulted in a reward of 20.78, but in the unified dimension calculation, they only received a reward of 14.49. For task performers who contribute high-quality perception data but do not receive the rewards they deserve, their enthusiasm for participating in perception tasks may be reduced, and they may even refuse to participate in subsequent perception tasks, leading to task failure or reduced data quality. Therefore, personalized and fine-grained perception strategies are beneficial for accurately calculating the contributions of task performers, increasing their enthusiasm and positive attitude towards participating in perception tasks, and ultimately improving the quality of aggregated perception data.

[0094] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

[0095] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any minor modifications, equivalent replacements, and improvements made to the above embodiments based on the technical essence of the present invention shall be included in the scope of protection of the technical solution of the present invention.

Claims

1. A fine-grained and high-quality privacy-preserving data aggregation method, characterized by: Including trust institutions, cloud service platforms, fog nodes, task requesters and task executors; Trust agencies, fog nodes, and task requesters are all connected to the cloud service platform through the network, and task executors are connected to fog nodes through the network; The trust agency is responsible for the generation, distribution, and management of secret keys and the generation of pseudonyms for task executors; The cloud service platform is responsible for receiving tasks issued by task requesters, unified data collection, and distribution of tasks and information; The fog node is responsible for encrypting the privacy information of task executors, selecting and assigning tasks to the appropriate task executor group, and collecting and uploading perception data; Therefore, more than 60% of data processing will occur in fog nodes, reducing the data upload load; The task requester is responsible for submitting task requests to the cloud service platform and receiving the data collected by the task; The task executor is responsible for executing the task and submitting the perception data to the fog node; The data aggregation method includes two steps: initialization phase and iteration phase; In the initial stage, the pseudonym and secret key are generated by the trusted authority; In the iteration phase, by continuously iterating the weights and true values ​​until the true values ​​converge, the weight of the task executor for each task is obtained under the pseudonym of the task executor, so as to aggregate high-quality data.

2. A fine-grained and high-quality privacy-preserving data aggregation method according to claim 1, characterized in that: The specific steps of the initialization stage are: The trust agency is responsible for the generation and distribution of keys. After generating the keys, the trust agency sends the public key to the task executor for encryption and sends the private key to the cloud service platform for decryption. In addition, when the task executor registers, the trust agency is responsible for assigning a pseudonym to each registered task executor.

3. The fine-grained and high-quality privacy-preserving data aggregation method according to claim 1, characterized in that: The specific steps of the iterative stage are: Define the task set as Task, Task = {t1, t2, ..., t i }, specify each task t i They are all region-related and managed by a fog node; The cloud service platform generates the initial weight of each task After that, the initial weight is sent to the task executors managed by the fog node; S1: The fog node sends the task t i Send to task executor u k , task executor u k After completing the perception data After collecting, calculate the task executor u k The distance between the perception data and the initial true value Then the task executor u k Choose a random value r k And save it separately, Encrypted using the following formula: After encryption is completed, the task executor will and Kana pse k Upload them to the fog node together; S2: The fog node receives the task t i After encrypting the distance data of U′ task performers, the data are aggregated according to the homomorphism of Paillier encryption: The aggregated data Upload to the cloud service platform; S3: The cloud service platform obtains aggregated encrypted data Finally, the aggregated data is decrypted using the private key sk: The decrypted data is then returned to the task executor through the fog node; S4: Assuming that all task executors perform calculations honestly, then task executor u j Getting the data Then update the weight: At this time the weight Represents the task executor u j For task t i After knowing the weight, the task executor first calculates his own perception data contribution Then it is encrypted with the public key pk And the encrypted data and the encrypted weight Upload them to the fog node together; Similarly, when executing task t i After all U′ task executors submit their data, the sum can be calculated The summed data and the encrypted weight Submit them to the cloud service platform together; S5: After receiving the data, the cloud service platform first decrypts the encrypted data using the private key sk: Then update the true value: After completing the truth value update, the cloud service platform sends the updated truth value to the task executors it manages through the fog nodes; S6: Iterate the above weight update and true value update multiple times until the true value converges, then the pseudonym pse of the task executor can be obtained. k Under this circumstance, the weight of the task executor for each task is set to 0 for the tasks that the task executor does not participate in. The weight set of the task executor can be obtained as follows: Then, the task executor will use the pseudonym pse k and weight set It is handed over to a trust agency. The trust agency knows the reward for each task and rewards the task executors according to different weights, thus achieving fine-grained rewards and encouraging subsequent task executors to submit fine-grained data, promoting data aggregation results, and obtaining high-quality data.

4. The fine-grained and high-quality privacy-preserving data aggregation method according to claim 1, characterized in that: The method for generating, distributing, and managing secret keys and generating pseudonyms for task performers during the initialization phase of the trust authority is as follows: First, the task executor u k Send an application to the trust agency, and the trust agency assigns a pseudonym pse to each task executor k ,Then the pseudonym is returned to the task executor; Secondly, the trust agency selects a security parameter k and two large prime numbers p and q, which satisfy the condition |p|=|q|=k. Then the trust agency calculates n=pq and λ=(p-1)(q-1). After these parameters are prepared, the trust agency sends the public key pk=(n, h n ) is passed to each task executor, allowing the task executor to encrypt the data. Then, the trust organization sends the private key sk = λ = (p-1)(q-1) to the cloud service platform, allowing the cloud service platform to decrypt the data.

5. The fine-grained and high-quality privacy-preserving data aggregation method according to claim 3, characterized in that: In step S1, the task executor u k The collected sensory data are divided into continuous data and discrete data; The distance calculation formula between continuous data and the initial true value is: Discrete data can be represented as vectors The selected parameter is 1 and the others are 0. The distance calculation formula between discrete data and the initial true value is:

Images