A method for discovering safe truth values ​​of vehicle crowd perception based on homomorphic encryption and differential privacy

Abstract

The present invention discloses a method for discovering the safe truth value of vehicle crowd intelligence perception based on homomorphic encryption and differential privacy, and the method includes: selecting vehicles participating in crowd intelligence perception tasks based on trust; the vehicles homomorphically encrypt their perception data; the edge server uses a differential privacy algorithm to process the ciphertext of the perception data of each vehicle, and sends it to the cloud server after decryption by a reputation center; the cloud server calculates the truth value of differential privacy, and the reputation center iteratively updates the weight and truth value of each vehicle; the reputation center adjusts the trust of the vehicle according to the vehicle weight, and updates the vehicle weight based on the trust; the edge server and the reputation center collaborate to update and calculate the weighted data ciphertext, and finally obtain the weighted perception data ciphertext; the reputation center decrypts and restores the weighted perception data ciphertext to obtain the safe truth value. The present invention reduces the complexity of truth value calculation and improves the accuracy of the truth value.

Description

Technical Field

[0001] The present invention belongs to the technical field of digital information transmission, and specifically relates to a vehicle crowd-sensing security truth value discovery method based on homomorphic encryption and differential privacy. Background Art

[0002] Crowd sensing originates from the concept of participatory sensing, involving mobile social networks and smart mobile devices, through which data can be explicitly or implicitly collected from users. Previous crowd sensing systems mainly use mobile devices (such as smartphones) to sense and share information around them. However, the limited sensing coverage and restricted resources of mobile devices greatly limit their sensing scale and duration. In the Internet of Vehicles scenario, vehicles have obvious advantages over smartphones in crowd sensing due to the increasing number of smart sensing devices on vehicles and the high-speed movement of vehicles that can quickly cover larger areas. Vehicle crowd sensing can provide a flexible, time-saving and cost-saving way to perform crowd sensing tasks.

[0003] Vehicle crowd perception provides a simple way to complete tasks, but in practice, there are large differences in the capabilities and work quality of vehicles. Some vehicles perceive data carefully, while others may make up or tamper with data because their goal is to answer tasks quickly to get more money. To solve this problem, the crowd perception platform can adopt a redundancy strategy to assign the same task to different vehicles to complete, and then integrate the data perceived by these vehicles to obtain the real data for each perception task.

[0004] But this faces the problem of noisy or conflicting data. Truth discovery is a data mining paradigm that was created to resolve conflicts in multi-source noisy data. It can mine accurate results, namely true values, from multiple conflicting data. The intuitive way to resolve conflicts is to take the majority of the classified data as the true value. However, majority voting assumes that all vehicles are equally reliable, which may not be true in some cases. In order to ensure the authority of the results, the results must be closer to the information provided by reliable vehicles. Obviously, the reliability of vehicles is very important for truth discovery. However, due to the openness of the crowd sensing platform and the indispensable privacy protection policy, it is generally impossible to infer the capabilities of crowd sensing vehicles from their past performance, such as the overall accuracy of their answers. Therefore, the challenge is that the reliability of vehicles is usually unknown in advance. Summary of the invention

[0005] The present invention provides a vehicle crowd-sensing security truth value discovery method based on homomorphic encryption and differential privacy, which reduces the complexity of truth value calculation and improves the accuracy of truth value.

[0006] In order to achieve the above technical objectives, the present invention adopts the following technical solutions:

[0007] A method for discovering the safety truth value of vehicle crowd intelligence perception based on homomorphic encryption and differential privacy, comprising:

[0008] The reputation center selects vehicles to participate in the crowd sensing task based on the trustworthiness of the vehicles;

[0009] The vehicles participating in the task perform homomorphic encryption on their perception data and differential noise respectively, and upload the obtained perception data ciphertext and differential noise ciphertext to the edge server; the differential noise is obtained by using a differential privacy mechanism on the perception data;

[0010] The edge server processes the perception data ciphertext of each vehicle, the differential noise ciphertext, and the noise ciphertext uploaded by the data requester to obtain the noisy perception data ciphertext, and sends it to the reputation center for decryption to obtain the noisy perception data, and then sends it to the cloud server;

[0011] The cloud server initializes the true value based on the noise-added perception data of each vehicle, and iteratively updates the weight and true value of each vehicle based on the distance between the noise-added perception data of each vehicle and the current true value; the cloud server sends the weight obtained from the final iteration to the reputation center;

[0012] The reputation center adjusts the trust of each vehicle according to the weight of all vehicles, and updates the weight of each vehicle based on the adjusted trust; the reputation center homomorphically encrypts the updated weight of each vehicle and sends it to the edge server;

[0013] The edge server adds disturbances to the ciphertext of the perception data and weight ciphertext of each vehicle, and then the reputation center encrypts and decrypts and fuses the perception data with the weights. The edge server then cancels the disturbance and aggregates all vehicles, and finally sends the obtained weighted perception data ciphertext to the reputation center.

[0014] The reputation center decrypts the weighted perception data ciphertext to obtain the plaintext, and obtains the security truth value based on the weights and restoration of all vehicles.

[0015] Furthermore, when selecting vehicles to participate in the task, the reputation center uses the following method to initialize the trust of each vehicle for all vehicles applying for the task: the initial trust of known trusted vehicles is set to 1, and the initial trust of other unknown vehicles is set to 0.5; then, from all vehicles applying for the task, N vehicles with high initial trust are selected to participate in the group perception task.

[0016] Furthermore, the edge server processes the perception data ciphertext of each vehicle, the differential noise ciphertext, and the noise ciphertext uploaded by the data requester to obtain the noisy perception data ciphertext It is expressed as:

[0017]

[0018] In the formula, For vehicle V i The perception data obtained by executing the crowd perception task m, E pk () represents the use of public key pk for encryption operations, For vehicle V i Perception data The ciphertext of the perceived data obtained by homomorphic encryption; c i For vehicle V i The differential noise provided, E pk (c i ) is the vehicle V i The differential noise c i The noisy ciphertext obtained by homomorphic encryption; s is the noise provided by the data requester, E pk (s) is the noise ciphertext obtained by the data requester through homomorphic encryption of the noise s; For vehicle V i The noisy perception data of the crowd intelligence perception task m is to be processed by the reputation center. Decrypted.

[0019] Furthermore, the cloud server initializes the true value based on the noise perception data of each vehicle, and specifically uses the average value to initialize the true value:

[0020]

[0021] In the formula, The noisy perception data decrypted by the reputation center, specifically the vehicle V i The noisy perception data of the group perception task m; N is the number of vehicles participating in the group perception task, X m is a true value;

[0022] The weight of each vehicle is iteratively updated according to the distance between the noisy perception data of each vehicle and the current true value, specifically:

[0023]

[0024] In the formula, w i For vehicle V i The weight of express and X m The Euclidean distance of

[0025] Update the true value based on the current weight of each vehicle and the noisy perception data, specifically:

[0026]

[0027] Return and repeat the above vehicle weight update and true value update until the difference between the true values ​​obtained from the two most recent updates is less than the threshold.

[0028] Furthermore, the reputation center adjusts the trust of each vehicle according to the weight of all vehicles, specifically:

[0029] Furthermore, the reputation center adjusts the trust of each vehicle according to the weight of all vehicles, specifically:

[0030] Take any known trusted vehicle V among all vehicles participating in the crowd sensing task r , with its weight As a standard weight, adjust each vehicle V i Trust i :

[0031]

[0032] in, Indicates vehicle V i The weight w i and standard weights The distance of i and The closer, The smaller it is; θ is the magnitude of the change in trust; ξ represents the distance threshold;

[0033] Update the weight of each vehicle based on the adjusted trust, specifically:

[0034] w i =w i t i .

[0035] Furthermore, the edge server adds disturbances to the ciphertext of the perception data and the weight ciphertext of each vehicle, expressed as:

[0036]

[0037] In the formula, and is the perturbation data randomly generated by the edge server, E pk () represents the use of public key pk for encryption operations, For vehicle V i Execute crowd intelligence perception task T m The sensory data obtained, is the ciphertext of the sensing data; w i For vehicle V i The weight, E pk (w i ) is the vehicle V i The weighted ciphertext of and The vehicle V i The perturbed perceptual data ciphertext and weighted ciphertext.

[0038] Furthermore, the encryption and decryption and the fusion of the perception data and weights by the reputation center specifically include:

[0039] (1) Use the private key to log in to the vehicle V i Perturbation of the ciphertext of the sensory data and weighted ciphertext Decrypt and obtain the perturbed perception data and weight

[0040] (2) The perturbed perception data of each vehicle and weight First perform the multiplication operation, then use the public key to encrypt the product to obtain the perturbed weighted data ciphertext

[0041] Furthermore, the edge server cancels the disturbance and performs aggregation processing on all vehicles, specifically including:

[0042] (1) Based on known disturbance data and The weighted data of each vehicle disturbance is encrypted using the Paillier additive homomorphic encryption algorithm. Cancel the disturbance and get the weighted perception data ciphertext Right now:

[0043]

[0044] (2) Aggregate the weighted perception data ciphertext of all N vehicles:

[0045]

[0046] In the formula, The weighted perception data ciphertext aggregated from all N vehicles.

[0047] Furthermore, the reputation center uses the private key sk to encrypt the aggregated weighted perception data Decryption is performed to obtain the plaintext D, and finally the true value is calculated based on the weights of all vehicles. It is expressed as:

[0048]

[0049] A vehicle crowd-sensing security truth value discovery system based on homomorphic encryption and differential privacy, comprising: a data request terminal, a vehicle, an edge server, a cloud server and a reputation center;

[0050] The data request terminal is used to send data requests to the edge server and publish vehicle crowd intelligence perception tasks in the reputation center;

[0051] The vehicle, edge server, cloud server and reputation center work together to implement any of the above-mentioned vehicle crowd intelligence perception security truth value discovery methods based on homomorphic encryption and differential privacy, and return the obtained security truth value to the data requesting terminal.

[0052] Compared with the prior art, the present invention has the following beneficial effects:

[0053] (1) Privacy protection of vehicle perception data based on homomorphic encryption. Different from the traditional homomorphic encryption scheme in which truth value discovery requires multiple encryption and decryption operations, the present invention proposes a privacy protection method based on homomorphic encryption, which reduces the complexity of truth value calculation. After submitting the data of the perception task, the mobile vehicle can exit the truth value discovery calculation process, which is conducive to reducing the computing load of the mobile vehicle and solving the problem of the mobile vehicle exiting midway or driving out of the edge server range.

[0054] (2) True value and weight privacy protection based on differential privacy. Different from the traditional differential privacy method in which the true value is calculated by adding noise data, the present invention proposes that the final true value uses real perception data, which can improve the accuracy of the true value.

[0055] (3) Trust-based truth value calculation technology. Different from the traditional solution that uses weights to evaluate user reliability, the invention proposes to evaluate user reliability by user credibility. Mobile vehicles that meet the corresponding trust requirements are selected based on the trust, where the trust represents the reliability of each vehicle. According to the trust of the mobile vehicle, a truth value discovery algorithm is designed to generate high-precision truth values ​​for perception tasks, which can further improve the accuracy of the truth value. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a schematic diagram of the overall architecture of the method of the present invention;

[0057] Figure 2 Schematic diagram of the network model of the method described in the present invention. DETAILED DESCRIPTION

[0058] The following is a detailed description of an embodiment of the present invention. This embodiment is based on the technical solution of the present invention, and provides a detailed implementation method and a specific operation process to further explain the technical solution of the present invention.

[0059] Example 1

[0060] This embodiment 1 provides a method for discovering the true value of vehicle crowd intelligence perception security based on homomorphic encryption and differential privacy, referring to Figure 1 As shown, the following steps are included:

[0061] Step 1: System initialization.

[0062] The reputation center generates keys pk and sk, and publishes the public key pk to the data requester and the vehicle, retaining the private key sk.

[0063] Step 2: Release the task.

[0064] The data requester publishes M crowd-sensing tasks T = {T1, T2, …, T M} to the edge server. The vehicle sends a request to do the task to the reputation center.

[0065] Step 3: Vehicle selection.

[0066] The reputation center obtains the trustworthiness of some vehicles based on the historical interaction records with the vehicles. Assuming that the reputation center knows that a small number of vehicles are trustworthy, the trustworthiness of the trustworthy vehicles Set to 1, the trust of most vehicles is unknown, and the initial trust value Set to 0.5. The set of trusted vehicles is , the set of vehicles with unknown trust is The reputation center sorts the vehicles according to their trustworthiness and selects N vehicles with high trustworthiness V = {V1, V2, …, V N} Participate in crowd perception tasks.

[0067] Step 4: Encrypt vehicle perception data with homomorphic encryption.

[0068] The reputation center notifies the selected N vehicles V = {V1, V2, ..., V N}To do the group intelligence perception task.

[0069] The vehicle uses the public key pk to respectively sense the data and differential noise c i Perform homomorphic encryption and encrypt the obtained perception data and the differential noise ciphertext E pk (c i ) to the edge server.

[0070] Among them, the differential noise c i By sensing data Using the differential privacy mechanism, we get l i,m is the vehicle V i In Task T m The sensitivity of the vehicle V iIn Task T m The privacy budget on

[0071] Definition 1(∈ i,j -Differential Privacy): A mechanism Satisfy∈ i,j -Differential privacy, if and only if for Any two inputs x in the domain i,j and x' i,j , and for Any output x in * ,have:

[0072]

[0073] Among them, ∈ i,j Is user V i In Task T j Privacy budget on ∈ i,j The smaller the value, the better the user U i The more confident the attacker is that * After that, it cannot distinguish x i,j and x' i,j The privacy budget represents the degree of privacy provided by the mechanism and controls how much noise should be added to the dataset. In general, larger perturbation noise will lead to stronger privacy guarantees but worse utility of the dataset.

[0074] Definition 2(l i,j -sensitivity): if |x' i,j -x i,j |≤l i,j , then x' i,j and x i,j Is to satisfy l i,j -adjacent data of sensitivity. Among them, l i,j Is user V i In Task T j The value range of the perception data on .

[0075] Theorem 1 (Laplace mechanism): If user V i Perform perception task T j , by adding noise η i,j , that is, x' i,j =x i,j +η i,j , the perception data reaches ∈ i,j -Differential privacy. Wherein, the noise η i,j It follows the Laplace distribution, and the probability density of this distribution is

[0076] The data requester generates a large noise s, performs homomorphic encryption on the noise, and obtains the large noise ciphertext E pk (s) are uploaded to the edge server.

[0077] The blockchain is deployed on the edge server, and the vehicle’s perception data is encrypted using the blockchain’s smart contract. Differential Noise Ciphertext E pk (c i ), large noise ciphertext E pk (s) to obtain the encrypted data of each vehicle, which is recorded as the noisy perception data ciphertext and will Send to reputation center.

[0078]

[0079] The reputation center uses the private key sk to decrypt Get the noisy perception data and will Send to the cloud server.

[0080] Step 5: Calculation of the truth value of differential privacy.

[0081] The cloud server receives the noise perception data of each vehicle i∈{1,…,N},m∈{1,…,M},initialize the truth value X m The average value of all vehicle noise perception data:

[0082]

[0083] Iteratively update the weight of each vehicle based on the distance between the noisy perception data of each vehicle and the current true value:

[0084]

[0085] In the formula, express and X m The Euclidean distance of

[0086] Among them, the distance is It can be seen that when the weights are updated, the large noises cancel each other out, eliminating the impact of the large noise sent by the data requester.

[0087] The reputation center calculates the weights of all vehicles as:

[0088]

[0089] The reputation center iteratively updates and calculates the true value X according to the truth value calculation formula. m :

[0090]

[0091] Then, the above vehicle weight update and true value update are repeated until the difference between the true values ​​obtained from the two most recent updates is less than the threshold.

[0092] Finally, the cloud center calculates the vehicle weight w i Sent to the reputation center.

[0093] Step 6: Update weights based on credibility.

[0094] The reputation center receives the vehicle weight w returned by the cloud center i If the returning vehicle In the trusted set In the case of a trusted vehicle Weight As standard, with vehicle V i The weight w i For comparison. If w i and If the distance is small, increase the vehicle V i Trust i If w i and If the distance is large, reduce the vehicle V i Trust i .

[0095]

[0096] in, Indicates w i and The distance, that is w i and The closer they are, the The smaller the value, the smaller the trust value. θ is the magnitude of the trust change, and ξ is the distance threshold.

[0097] After adjusting the vehicle trust, the reputation center adjusts the vehicle weight according to the updated vehicle trust:

[0098] w i =w i t i (7)

[0099] The reputation center uses the public key pk to update the weight w i Perform homomorphic encryption and set the encryption weight E pk (w i ) to the edge server.

[0100] Step 7: Update the weighted data ciphertext.

[0101] The edge server randomly generates perturbation data and And encrypt the perception data of each vehicle Add the disturbance to get According to the disturbance data For each participating vehicle’s weighted ciphertext E pk (w i ) Add the disturbance to get Finally, the disturbed perception data ciphertext and weighted ciphertext Send to reputation management center.

[0102] Reputation Management Center's ciphertext for perturbations and Decrypt it and get and The reputation management center performs multiplication on the decrypted data and encrypts it to obtain the disturbed weighted data ciphertext And send it to the edge server.

[0103] The edge server uses the known disturbance data and Through the Paillier addition homomorphic encryption algorithm Cancel the disturbance and get the weighted perception data ciphertext The edge server aggregates the weighted perception data ciphertext of all vehicles for the task Finally, the edge server will weight the data ciphertext Send to reputation management center.

[0104] Step 8: Ground truth calculation.

[0105] The reputation center decrypts the data ciphertext with the private key sk Get the plaintext D, The reputation center calculates the weights of all vehicles as Then, the reputation center calculates the security value based on the truth value calculation formula.

[0106]

[0107] Finally, the reputation center will get the security truth value Sent to the data requester.

[0108] Example 2

[0109] This embodiment 2 provides a vehicle crowd intelligence perception security truth value discovery system based on homomorphic encryption and differential privacy, such as Figure 2 As shown, it includes: a data request terminal, a vehicle, an edge server, a cloud server and a reputation center;

[0110] The data request terminal is used to send data requests to the edge server and publish vehicle crowd intelligence perception tasks in the reputation center;

[0111] The vehicle is equipped with a mobile device that has the ability to perform various sensing tasks assigned by the reputation center and submit the sensing data to the edge server.

[0112] The edge server is responsible for collecting the vehicle's perception data ciphertext and differential noise ciphertext.

[0113] The cloud server performs truth value calculation, truth value update and weight update on a noisy basis.

[0114] The reputation center is responsible for generating Paillier homomorphic encryption keys and distributing public keys to data requesters and vehicles. It is also responsible for calculating vehicle trust and recruiting vehicles based on trust.

[0115] The vehicle, edge server, cloud server and reputation center work together to implement the vehicle crowd intelligence perception security truth value discovery method based on homomorphic encryption and differential privacy described in Example 1, and return the obtained security truth value to the data request terminal.

[0116] The above embodiments are preferred embodiments of the present application. Ordinary technicians in this field can also make various changes or improvements on this basis. Without departing from the overall concept of the present application, these changes or improvements should fall within the scope of protection required by the present application.

Claims

1. A method for discovering the security truth value of vehicle crowd intelligence perception based on homomorphic encryption and differential privacy, characterized in that: include: The reputation center selects vehicles to participate in the crowd sensing task based on the trustworthiness of the vehicles; The vehicles participating in the task perform homomorphic encryption on their perception data and differential noise respectively, and upload the obtained perception data ciphertext and differential noise ciphertext to the edge server; the differential noise is obtained by using a differential privacy mechanism on the perception data; The edge server processes the perception data ciphertext of each vehicle, the differential noise ciphertext, and the noise ciphertext uploaded by the data requester to obtain the noisy perception data ciphertext, and sends it to the reputation center for decryption to obtain the noisy perception data, and then sends it to the cloud server; The cloud server initializes the true value based on the noise-added perception data of each vehicle, and iteratively updates the weight and true value of each vehicle according to the distance between the noise-added perception data of each vehicle and the current true value; The cloud server sends the weight obtained from the final iteration to the reputation center; The reputation center adjusts the trust of each vehicle according to the weight of all vehicles, and updates the weight of each vehicle based on the adjusted trust; the reputation center homomorphically encrypts the updated weight of each vehicle and sends it to the edge server; The edge server adds disturbances to the ciphertext of the perception data and weight ciphertext of each vehicle, and then the reputation center encrypts and decrypts and fuses the perception data with the weights. The edge server then cancels the disturbance and aggregates all vehicles, and finally sends the obtained weighted perception data ciphertext to the reputation center. The reputation center decrypts the weighted perception data ciphertext to obtain the plaintext, and obtains the security truth value based on the weights and restoration of all vehicles.

2. The method for discovering the safety truth value of vehicle crowd intelligence perception based on homomorphic encryption and differential privacy according to claim 1 is characterized in that: When selecting vehicles to participate in the task, the reputation center uses the following method to initialize the trust of each vehicle for all vehicles applying for the task: the initial trust of known trusted vehicles is set to 1, and the initial trust of other unknown vehicles is set to 0.5; then, from all the vehicles applying for the task, N vehicles with high initial trust are selected to participate in the crowd sensing task.

3. The method for discovering the safety truth value of vehicle crowd intelligence perception based on homomorphic encryption and differential privacy according to claim 1 is characterized in that: The edge server processes the perception data ciphertext of each vehicle, the differential noise ciphertext, and the noise ciphertext uploaded by the data requester to obtain the noisy perception data ciphertext. It is expressed as: In the formula, For vehicle V i The perception data obtained by executing the crowd perception task m, E pk ( ) represents the use of public key pk for encryption operations. For vehicle V i Perception data The ciphertext of the perceived data obtained by homomorphic encryption; c i For vehicle V i The differential noise provided, E pk (c i ) is the vehicle V i Its differential noise c i The noisy ciphertext obtained by homomorphic encryption; s is the noise provided by the data requester, E pk (s) is the noise ciphertext obtained by the data requester through homomorphic encryption of the noise s; For vehicle V i The noisy perception data of the crowd intelligence perception task m is to be processed by the reputation center. Decrypted.

4. The method for discovering the safety truth value of vehicle crowd intelligence perception based on homomorphic encryption and differential privacy according to claim 1 is characterized in that: The cloud server initializes the true value based on the noise perception data of each vehicle, specifically using the average value to initialize the true value: In the formula, The noisy perception data decrypted by the reputation center, specifically the vehicle V i The noisy perception data of the group perception task m; N is the number of vehicles participating in the group perception task, X m is a true value; The weight of each vehicle is iteratively updated according to the distance between the noisy perception data of each vehicle and the current true value, specifically: In the formula, w i For vehicle V i The weight of express and X m The Euclidean distance of Update the true value based on the current weight of each vehicle and the noisy perception data, specifically: Return and repeat the above vehicle weight update and true value update until the difference between the true values ​​obtained from the two most recent updates is less than the threshold.

5. The method for discovering the safety truth value of vehicle crowd intelligence perception based on homomorphic encryption and differential privacy according to claim 1 is characterized in that: The reputation center adjusts the trust of each vehicle based on the weight of all vehicles, specifically: Take any known trusted vehicle V among all vehicles participating in the crowd sensing task r , with its weight As a standard weight, adjust each vehicle V i Trust i : in, Indicates vehicle V i The weight w i and standard weights The distance of i and The closer, The smaller it is; θ is the magnitude of the change in trust; ξ represents the distance threshold; Update the weight of each vehicle based on the adjusted trust, specifically: w i =w i t i 。 6. The method for discovering the safety truth value of vehicle crowd intelligence perception based on homomorphic encryption and differential privacy according to claim 1 is characterized in that: The edge server adds perturbations to the ciphertext of the perception data and weight ciphertext of each vehicle, expressed as: In the formula, and is the perturbation data randomly generated by the edge server, E pk () represents the use of public key pk for encryption operations, For vehicle V i The perception data obtained by executing the crowd perception task m, is the ciphertext of the sensing data; w i For vehicle V i The weight, E pk (w i ) is the vehicle V i The weighted ciphertext of and The vehicle V i The perturbed perceptual data ciphertext and weighted ciphertext.

7. The method for discovering the safety truth value of vehicle crowd intelligence perception based on homomorphic encryption and differential privacy according to claim 6 is characterized in that: The encryption and decryption and the fusion of the perception data and weights by the reputation center specifically include: (1) Use the private key to log in to the vehicle V i Perturbation of the ciphertext of the sensory data and weighted ciphertext Decrypt and obtain the perturbed perception data and weight (2) The perturbed perception data of each vehicle and weight First perform the multiplication operation, then use the public key to encrypt the product to obtain the perturbed weighted data ciphertext 8. The method for discovering the safety truth value of vehicle crowd intelligence perception based on homomorphic encryption and differential privacy according to claim 7 is characterized in that: The edge server cancels disturbances and performs aggregation processing on all vehicles, specifically including: (1) Based on known disturbance data and The weighted data ciphertext of each vehicle disturbance is encrypted using the Paillier additive homomorphic encryption algorithm. Cancel the disturbance and get the weighted perceptual data ciphertext Right now: (2) Aggregate the weighted perception data ciphertext of all N vehicles: In the formula, The weighted perception data ciphertext aggregated from all N vehicles.

9. The method for discovering the safety truth value of vehicle crowd intelligence perception based on homomorphic encryption and differential privacy according to claim 8 is characterized in that: The reputation center uses the private key sk to encrypt the weighted perception data obtained by aggregation Decryption is performed to obtain the plaintext D, and finally the true value is calculated based on the weights of all vehicles. It is expressed as:

10. A vehicle crowd intelligence perception safety truth value discovery system based on homomorphic encryption and differential privacy, characterized in that: include: Data request terminals, vehicles, edge servers, cloud servers, and reputation centers; The data request terminal is used to send data requests to the edge server and publish vehicle crowd intelligence perception tasks in the reputation center; The vehicle, edge server, cloud server and reputation center work together to implement the vehicle crowd perception security truth value discovery method based on homomorphic encryption and differential privacy as described in any one of claims 1-9, and return the obtained security truth value to the data request terminal.

Images