Privacy-protected edge node vehicle networking data aggregation method, system and application

Abstract

The invention provides a privacy-protected edge node vehicle networking data aggregation method, system and application, vehicles with similar attributes are clustered, an aggregated vehicle is selected from each cluster, the vehicle is responsible for collecting vehicle data in the cluster, the aggregated vehicle transmits the data to an edge node, and the edge node transmits the data to the edge node. After the edge nodes primarily gather the gathered vehicle data in the coverage range, zero sum noise is added in the gathering process, and finally the gathered data of the edge nodes are sent to the cloud server for final gathering; in consideration of unstable connection caused by relatively high vehicle mobility in a road condition, a natural neighbor set based on stability between connected vehicles is designed to be combined with a density clustering algorithm, vehicles in a road form are clustered, and cluster head vehicles are selected according to connection stability between the vehicles and a foundation; the reliability and stability of transmission are greatly improved, noise is added to each data matrix in the form of zero sum noise, and the privacy of a user can be well protected.

Description

Technical Field

[0001] The present invention belongs to the field of Internet of Vehicles privacy protection and information security technology, and specifically relates to a privacy-protected edge node Internet of Vehicles data aggregation method, system and application. Background Art

[0002] With the development of intelligent transportation systems and the popularization of Internet of Vehicles technology, information exchange between vehicles has become increasingly frequent. In the Internet of Vehicles environment, vehicles communicate and exchange data to achieve more intelligent and efficient traffic management and vehicle operation. Data aggregation, as an important link in the Internet of Vehicles, plays a key role in collecting, integrating and analyzing vehicle data. For example, vehicles transmit information such as the location, speed, and driving route of the vehicle. By aggregating and analyzing these data, real-time traffic status information can be provided to traffic management departments, thereby optimizing traffic signal control, route planning and other decisions, and improving traffic operation efficiency and safety. The Chinese patent with publication number CN105657642A provides a clustered non-encrypted privacy-preserving data fusion scheme, which focuses on reducing computational complexity and energy consumption while ensuring data privacy protection. Through reasonable network clustering, non-encrypted camouflage data filling, and efficient intra-cluster and inter-cluster data fusion methods, the energy efficiency and scalability of the network are improved to meet the application needs of large-scale wireless sensor networks. The Chinese patent with publication number CN113868695A provides a trusted privacy protection scheme based on blockchain in crowd-sensing data aggregation. The scheme focuses on ensuring the correctness of data aggregation results while using blockchain characteristics to ensure the traceability of the data aggregation process, and effectively protects data privacy through noise addition mechanism and grouping strategy. Compared with the encryption mechanism-based method, it focuses on improving the efficiency of privacy protection execution, reducing the cost of calculation and communication, and achieving a better balance between credibility, privacy and efficiency to meet the needs of crowd-sensing network applications. Neither of the above two patents takes into account the characteristics of the Internet of Vehicles world. On today's increasingly crowded roads, the number of vehicles is increasing, which means that if the data of all vehicles is directly transmitted to the cloud server, it will inevitably cause serious communication delays and network congestion problems. Considering the complexity and quantity of vehicle data in reality, if each vehicle uploads its data directly to the cloud server, it will not only put pressure on the communication network, but also may cause high latency in data transmission, and even affect the stable operation of the entire transportation system due to data congestion. Therefore, it is crucial to design a reasonable data aggregation mechanism.

[0003] In addition, the current social awareness of privacy protection is gradually increasing, and vehicle data needs to upload information such as vehicle location and speed. This information is directly related to the sensitive information of the owner. Therefore, the protection of vehicle information needs to be fully considered during the data aggregation process. The data availability is also closely related to the accuracy of the aggregation results. Therefore, while protecting vehicle information, it is also necessary to pay attention to the availability of data. Summary of the invention

[0004] In order to solve the problems existing in the prior art, the present invention provides a privacy-preserving edge node vehicle-connected data aggregation method, in which vehicles with similar attributes are clustered. In order to reduce the amount of transmitted data, an aggregation vehicle is selected in each cluster, which is responsible for collecting vehicle data in the cluster. The aggregation vehicle transmits the data to the edge node. After the edge node performs preliminary aggregation of the aggregated vehicle data within the coverage area, zero-sum noise is added in the aggregation process, and finally the aggregated data of the edge node is sent to the cloud server for final aggregation.

[0005] In order to achieve the above objectives, in a first aspect, the present invention provides a privacy-preserving edge node vehicle-connected data aggregation method, comprising the following steps: Cluster the vehicles using density peak clustering combined with natural neighbor sets, and divide the vehicles into multiple clusters; Send the sensing data of each cluster to the roadside unit; Noising the sensing data with zero-sum noise and then sending the data; The noisy perception data is received, and the noisy perception data is subjected to zero-sum noise elimination and aggregation processing to obtain a vehicle-connected data aggregation result.

[0006] Furthermore, the density peak clustering is based on the assumption that the cluster center is surrounded by other data points with lower density in the cluster and the distance between cluster centers is relatively far. The vehicle density is defined based on weights and the density of data points is defined based on natural neighbors. The distance between cluster centers is described by relative distance, and the relative distance of a general point is the distance to the point with a greater density than it and closest to it.

[0007] Furthermore, the natural neighbor set is specifically: R - Given a set of vehicles in the neighborhood, use vehicle mobility to measure the similarity between vehicles and search R -neighborhood to a stable search state and obtain a set of natural neighbors for a vehicle.

[0008] Furthermore, sending the sensing data matrix of each cluster to the roadside unit includes: The cluster head vehicle is determined in each divided cluster. The cluster head vehicle acts as an intermediate bridge to collect vehicle data within the cluster and send the collected data to the roadside unit.

[0009] Furthermore, combined with V2V link stability Link stability with vehicles and roadside units Determine the cluster head vehicle;

[0010] in, When it is 0, it means the vehicles are in the same direction, and 1 means they are in the opposite direction. Indicates vehicle and The absolute value of the speed difference, Indicates the vehicle For vehicles The impact of use Function pair standardization:

[0011] in, is the distance from vehicle i to the nearest roadside unit. If the center position of the vehicle group where vehicle i is located is closest to a certain roadside unit, then this roadside unit is the nearest roadside unit for this group of vehicles. is the average distance from the group of vehicles to the roadside unit, is the standard deviation of the distance from the group of vehicles to the roadside unit; the link stability is:

[0012] in, and is the weight parameter, adjusting and The weight selection among the cluster head vehicles has the largest The fraction of vehicles acting as cluster heads.

[0013] Furthermore, when the sensed data of each cluster is sent to the roadside unit: the vehicles in the cluster send the sensed data to the cluster head vehicle, the cluster head vehicle preliminarily aggregates the data of the vehicles in the cluster in the form of a matrix, and sends the aggregated results to the roadside unit.

[0014] Furthermore, when the perception data is noisyed with zero-sum noise and then sent: the roadside unit receives the perception data, finds that the maximum number of vehicles in each cluster is M, and determines whether the number of rows of the matrix is ​​M when the roadside unit receives the perception matrix. If not, fill the matrix with 0 to make it reach M rows, and then superimpose the received perception matrices to obtain the sum of all data within the perception range of the roadside unit; each roadside unit randomly generates m noise matrices, the sum of the m matrices is 0, and distributes the matrices in sequence. Each roadside unit will receive m noise matrices, and the m noise matrices are added to obtain the final noise matrix, and the final noise matrix is ​​added to the perception data matrix to obtain the noisy perception data matrix, and it is sent to the cloud server.

[0015] Furthermore, when receiving the noisy perception data: the cloud server obtains the perception data matrix sent by the roadside unit, accumulates all the perception data matrices, eliminates zero-sum noise, obtains the element sum of the perception data matrix, averages the element sum, calculates the data item sum of the perception data matrix and divides it by the number of vehicles to obtain the final aggregation result.

[0016] In a second aspect, the present invention provides an edge node vehicle connection system, including a cloud server, multiple vehicles and a roadside unit; data processing is performed based on the edge node vehicle connection data aggregation method with privacy protection as described above.

[0017] Thirdly, the above-mentioned privacy-preserving edge node vehicle-connected data aggregation method is used to detect vehicle anomalies and identify vehicle behaviors that are inconsistent with normal practices.

[0018] Compared with the prior art, the present invention has at least the following beneficial effects: the present invention proposes a vehicle-connected data aggregation method based on edge nodes for privacy protection, designs a clustering method that considers vehicle characteristics through a density clustering algorithm, divides vehicles into several clusters, selects the optimal head vehicle in each cluster according to a stability score function, is responsible for receiving the data of vehicles in the cluster, and sends it to the edge node; constructs zero-sum noise at the edge node, and each edge node adds the noise matrix to the original data matrix to obtain the final data matrix, and then sends it to the cloud server, and the cloud server adds the received data matrices to eliminate the zero-sum noise, and obtains the average value of the final perception data; the present invention considers that the high mobility of vehicles in road conditions may lead to unstable connection, designs a natural neighbor set based on the stability of the connection between vehicles combined with a density clustering algorithm, clusters the vehicles in the road, and selects the cluster head vehicle according to the connection stability between the vehicle and the basic setting, which greatly improves the reliability and stability of transmission; the present invention considers that the data accuracy cannot be lost while protecting the user's privacy, and uses zero-sum noise to add noise to each data matrix, which can well protect the user's privacy.

[0019] Furthermore, using natural neighbor sets to calculate local density can adaptively and without parameters obtain different numbers of natural neighbors based on the characteristics of each point. Normal points will not regard noise points and abnormal points as natural neighbors. Vehicles have the characteristics of high mobility, and dividing vehicles with similar speed attributes into a group can improve the stability of the cluster. The reliability of clustering can be better improved through natural neighbors.

[0020] Furthermore, zero-sum noise is used to add noise to the data matrix to protect the privacy of vehicle users without losing data accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 Flow chart of the method for aggregating edge node vehicle-connected data based on privacy protection. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are 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 creative work are within the scope of protection of the present invention.

[0023] The Internet of Vehicles data aggregation is to aggregate the data of vehicles traveling on the road and analyze the aggregated data, so as to improve the efficiency of transportation. In view of the fact that people are paying more and more attention to personal privacy and have extremely high requirements for data accuracy, the present invention proposes a privacy-protected edge node Internet of Vehicles data aggregation method, which includes the following steps: (1) Clustering based on density peak clustering combined with natural neighbor clustering; In the Internet of Vehicles environment, the distribution of vehicles may be irregular. Density clustering can adapt to this complex situation and effectively identify vehicle clusters of different shapes. In 2014, Rodriguez et al. proposed a fast search and clustering method for finding density peaks in the paper "clustering by fastsearch and find of density peaks", referred to as the density peak clustering (DPC) algorithm. It is based on two assumptions: 1) The cluster center is surrounded by other data points with lower density in the cluster; 2) The distance between cluster centers is relatively far. From this, two important concepts can be obtained: local density and relative distance. Using natural neighbor sets to calculate local density can adaptively and without parameters obtain different numbers of natural neighbors according to the characteristics of each point, so as to calculate the local density of each data point. As follows: Define inter-vehicle mobility (MR) as:

[0024] Among them, the first half is the speed influence, and the second half is the distance influence. and Weight parameters, and For vehicles and The absolute value of the speed, is the maximum value of the velocity vector in the vehicle. For vehicles and The angle of speed, For vehicles and The distance between is the maximum distance between vehicles.

[0025] The vehicle density is defined based on the weights, and the density of the data points is defined based on the natural neighbors as follows:

[0026] in, For vehicles The natural neighbor set of For vehicles The number of natural neighbors of is the total number of vehicles in this cluster, For vehicles and The distance between.

[0027] Relative distance The calculation principle of relative distance is that the relative distance of a general point is the distance to the point with greater density than it and closest to it, and the distance to the point with the largest density is the maximum distance from it to other points.

[0028]

[0029] The natural nearest neighbor (TN) is a new concept of nearest neighbor. It is a scale-free nearest neighbor, which is also the biggest difference between it and K-nearest neighbor and ε-nearest neighbor. The basic idea of ​​the natural nearest neighbor is that the data points in the dense area of ​​the data set have more natural neighbors, the data points in the sparse area have fewer natural neighbors, and the most outlier data points in the data set have only a few or no nearest neighbors. The characteristic of the natural nearest neighbor is that the calculation process does not require any parameters, and the data points obtain accurate neighbors according to the attribute characteristics of the data set itself. The number of natural neighbors varies due to the density of the data. Since noise points and abnormal points have no neighbors, normal points will not regard noise points and abnormal points as natural neighbors. The natural neighbor set has the following definition: R -Neighborhood ( R-neighbor ) Given a set of vehicles ,use MR Weighing vehicles and The similarities between Indicates return No. r Neighbors. yes A subset of R-neighbor The definition is as follows:

[0030] Stable search state, search The process of the -neighborhood stops if and only if the following conditions are met:

[0031] The search round , when the stable search state is reached, is the natural eigenvalue . Then the natural neighbors are:

[0032] in, For vehicles The set of natural neighbors.

[0033] Density peak clustering based on natural neighbor sets can better discover vehicles with the same attributes and cluster them. As vehicles are highly mobile during driving, it is crucial to improve the quality of vehicle-to-vehicle communication (V2V). Define vehicle mobility (MR), integrate vehicle mobility into density peak clustering, and divide vehicles into multiple clusters denoted as , n is the number of clusters.

[0034] (2) Select the cluster head vehicle and send the data to the roadside unit; In step (1), based on density peak clustering, the best cluster head vehicle is found in each cluster. The cluster head vehicle acts as an intermediate bridge, collecting vehicle data within the cluster and sending the collected data to the roadside unit. Therefore, the communication quality between the cluster head vehicle and other vehicles must be high, and the communication quality with the roadside unit must also be guaranteed. Based on these two conditions, the communication quality of V2V and the communication quality between vehicles and roadside units are defined, and the combination of the two forms the final standard for judging the cluster head vehicle.

[0035] The perception data of the i-th vehicle in the cluster is recorded as , kTo sense the data dimension, the vehicle sends the sensed data to the cluster head vehicle, which aggregates the data sensed by each vehicle in the form of a two-dimensional matrix. , where the rows represent the data perceived by the vehicle and the columns represent each attribute value perceived. Each cluster head vehicle sends the collected two-dimensional matrix to the roadside unit.

[0036] (3) Generate zero-sum noise matrix and add zero-sum noise The roadside unit collects the data matrix sent by the cluster head vehicles within the coverage area. To facilitate subsequent operations, each matrix is ​​first padded to a j*k size matrix, where j is the maximum number of vehicles in each cluster and k is the dimension of the vehicle perception data. Then the received matrices are superimposed, and finally a j*k size matrix is ​​formed in each roadside unit. In order to protect the privacy of vehicle users without losing data accuracy, zero-sum noise is used to add noise to the data matrix. Suppose there are m roadside units on the road, and each roadside unit generates m random matrices, and the generated random matrices are added to a zero matrix. And the roadside units of the corresponding order retain the random matrices of the corresponding order, and send the remaining random matrices to the roadside units with the same sequence number according to the sequence number. Then the final data matrix of each roadside unit is the original data matrix plus the retained random matrix plus the received random matrix. The roadside unit sends the final matrix to the cloud server for aggregation.

[0037] (4) Eliminate zero-sum noise and aggregate data on cloud servers; After the cloud server adds the received data matrices, it eliminates the impact of zero-sum noise and obtains the sum of the original data matrices. The data aggregation method is to find the average value of the sensed data on the road, find the sum of the data items in the matrix and divide it by the number of vehicles to get the final result.

[0038] The technical solution of the present invention is further described below with reference to the accompanying drawings and examples: Figure 1 As shown, the edge node vehicle-connected data aggregation flow chart based on privacy protection of the present invention, the steps of the method are as follows: (1) Calculate the natural neighbor set based on vehicle mobility; Vehicles have the characteristic of high mobility. Grouping vehicles with similar speed attributes into a group can improve the stability of the cluster. The reliability of clustering can be better improved through natural neighbors. According to the above definition, find the natural neighbor set of each vehicle.

[0039] (2) Cluster vehicles using density peak clustering based on natural neighbor sets; Obtain the natural neighbor sets of all vehicles in (1) and cluster the vehicles according to the density peak clustering algorithm.

[0040] (3) Select the vehicle head in the cluster and aggregate the vehicle data within the cluster; Obtain the clustering result in (2) and inform each vehicle of the clustering result. To maintain communication stability, when selecting the cluster head vehicle, it is necessary to consider the communication between vehicles and the communication between vehicles and the infrastructure. Then there is the following definition: Link stability, indicating vehicle The vehicles in the cluster Link Quality:

[0041] in The value of is {0,1}, where 0 means the vehicles are moving in the same direction, and 1 means they are moving in the opposite direction. Indicates vehicle and The absolute value of the speed difference, Indicates the vehicle For vehicles impact.

[0042] Link stability, indicating vehicle Link quality with roadside units, using Function pair standardization.

[0043]

[0044] in is the distance from the vehicle to the nearest roadside unit. If the center position of the group of vehicles is closest to the roadside unit, then the roadside unit is the nearest roadside unit for the group of vehicles. is the average distance from the group of vehicles to the roadside unit, is the standard deviation of the distance from this group of vehicles to the roadside unit.

[0045] Link stability,In order to evaluate the selection of cluster head vehicles, link stability is proposed:

[0046] in, and is the weight parameter, adjusting and The weight selection among the cluster head vehicles has the largest The fraction of vehicles acting as cluster heads.

[0047] Maintaining a The unique vehicle identifier of each vehicle is stored in ),Score( ), and a set of single-hop nodes. The vehicle performs a breadth-first search traversal of the graph based on the clustering results and selects the vehicle with the highest score as the cluster head vehicle. After the cluster head vehicle is selected, the vehicles in the cluster send the sensed data to the cluster head vehicle, which performs a preliminary clustering of the data of the vehicles in the cluster in the form of a matrix and sends the clustered results to the roadside unit.

[0048] (4) Preliminary aggregation of roadside units. Collect cluster head vehicle data to form an original data matrix; Due to different vehicle densities, the number of vehicles in the cluster will be inconsistent, which will further lead to different numbers of matrix rows sent by the cluster head vehicle to the roadside unit. First, find the maximum number of vehicles in each cluster is M. When the roadside unit receives the perception matrix, determine whether the number of rows in the matrix is ​​M. If not, fill the matrix with 0 to make it reach M rows. Then superimpose the received perception matrices to obtain the sum of all data within the perception range of the roadside unit.

[0049] (5) Generate a noise matrix, add it to the original data matrix to form the final matrix, and send the matrix to the cloud server; In order to improve data accuracy while protecting user privacy, the present invention uses zero-sum noise. Assume that there are m roadside units at this time, and each roadside unit randomly generates m noise matrices, and the sum of the m matrices is 0. After generation, the matrices are distributed in order. For example, the first roadside unit retains the first noise matrix, and the remaining m-1 noise matrices are distributed in order according to the sequence number of the roadside unit. In the end, each roadside unit will receive m noise matrices. Add these m noise matrices to obtain the final noise matrix. Finally, add the final noise matrix to the data matrix to obtain the final data matrix of the roadside unit, and send it to the cloud server.

[0050] (6) The cloud server collects the obtained matrices and adds them together to eliminate noise; In the Internet of Vehicles environment, data anomaly detection is crucial to ensure the security of vehicle communications and improve traffic management efficiency. Anomaly detection can be performed by summing multiple data matrices and calculating the average value. By analyzing the average value and fluctuation level of the data matrix, unconventional vehicle behaviors can be identified. The present invention aggregates the perception data of all vehicles and uses the average value to determine whether anomalies occur during vehicle driving.

[0051] First, the cloud server obtains the data matrix sent by the roadside unit, accumulates all the data matrices, eliminates zero-sum noise, obtains the element sum of the matrix, and calculates the average value.

[0052] Example 2: Based on the data aggregation method, the present invention provides an edge node vehicle connection system, including a cloud server, multiple vehicles and a roadside unit; data aggregation is performed based on the edge node vehicle connection data aggregation method with privacy protection as described above.

[0053] In addition, the present invention can also use the above-mentioned privacy-protected edge node vehicle-connected data aggregation method to perform anomaly detection on vehicles and identify vehicle behaviors that do not conform to normal practices.

[0054] In summary, the present invention provides a privacy-protected edge node vehicle-connected data aggregation method, which clusters vehicles with similar attributes. In order to reduce the amount of transmitted data, a clustering vehicle is selected in each cluster, which is responsible for collecting vehicle data in the cluster. The clustering vehicle transmits the data to the edge node. After the edge node performs preliminary aggregation on the aggregated vehicle data within the coverage area, zero-sum noise is added in the aggregation process, and finally the aggregated data of the edge node is sent to the cloud server for final aggregation; on the one hand, the present invention greatly improves the stability of the communication link by designing a clustering algorithm and an index for selecting clustered vehicles. On the other hand, the present invention fully considers the data availability and the size of the data volume, constructs zero-sum noise, and fully ensures the availability of the data. At the same time, when designing the intermediate aggregation process, the data delay is effectively reduced. The present invention considers that the high mobility of vehicles in the road situation will lead to unstable connection. The natural neighbor set combined with the density clustering algorithm based on the stability of the connection between vehicles is designed to cluster the vehicles in the road, and then selects the cluster head vehicle according to the connection stability between the vehicle and the basic setting, which greatly improves the reliability and stability of the transmission. The present invention considers protecting user privacy while not losing data accuracy, and uses zero-sum noise to add noise to each data matrix, which can well protect user privacy.

Claims

1. A privacy-preserving edge node vehicle-connected data aggregation method, characterized in that: The following steps are involved: Cluster the vehicles using density peak clustering combined with natural neighbor sets, and divide the vehicles into multiple clusters; Send the sensing data of each cluster to the roadside unit; Noising the sensing data with zero-sum noise and then sending the noise; The noisy perception data is received, and the noisy perception data is subjected to zero-sum noise elimination and aggregation processing to obtain a vehicle-connected data aggregation result.

2. The privacy-preserving edge node vehicle-connected data aggregation method according to claim 1, characterized in that: The density peak clustering is based on the assumption that the cluster center is surrounded by other data points with lower density in the cluster and the distance between cluster centers is relatively far. The vehicle density is defined based on weights and the density of data points is defined based on natural neighbors. The distance between cluster centers is described by relative distance. Generally, the relative distance of a point is the distance to the point with greater density than it and closest to it.

3. The privacy-preserving edge node vehicle-connected data aggregation method according to claim 2, characterized in that: The natural neighbor set is: R - Given a set of vehicles in the neighborhood, use vehicle mobility to measure the similarity between vehicles and search R -neighborhood to a stable search state and obtain a set of natural neighbors for a vehicle.

4. The privacy-preserving edge node vehicle-connected data aggregation method according to claim 1, characterized in that: Sending the sensing data matrix of each cluster to the roadside unit includes: The cluster head vehicle is determined in each divided cluster. The cluster head vehicle acts as an intermediate bridge to collect vehicle data within the cluster and send the collected data to the roadside unit.

5. The privacy-preserving edge node vehicle-connected data aggregation method according to claim 1, characterized in that: Combined with V2V link stability Link stability with vehicles and roadside units Determine the cluster head vehicle; in, When it is 0, it means the vehicles are in the same direction, and 1 means they are in the opposite direction. Indicates vehicle and The absolute value of the speed difference, Indicates the vehicle For vehicles The impact of use Function pair standardization: in, is the distance from vehicle i to the nearest roadside unit. If the center position of the vehicle group where vehicle i is located is closest to a certain roadside unit, then this roadside unit is the nearest roadside unit for this group of vehicles. is the average distance from the group of vehicles to the roadside unit, is the standard deviation of the distance from the group of vehicles to the roadside unit; the link stability is: in, and is the weight parameter, adjusting and The weight selection among the cluster head vehicles has the largest The fraction of vehicles is regarded as cluster head.

6. The privacy-preserving edge node vehicle-connected data aggregation method according to claim 1, characterized in that: When the sensed data of each cluster is sent to the roadside unit: the vehicles in the cluster send the sensed data to the cluster head vehicle, the cluster head vehicle preliminarily aggregates the data of the vehicles in the cluster in the form of a matrix, and sends the aggregated results to the roadside unit.

7. The privacy-preserving edge node vehicle-connected data aggregation method according to claim 1, characterized in that: When the perception data is noisyed with zero-sum noise and then sent: the roadside unit receives the perception data, finds that the maximum number of vehicles in each cluster is M, and determines whether the number of rows of the matrix is ​​M when the roadside unit receives the perception matrix. If not, fill the matrix with 0 to make it reach M rows, and then superimpose the received perception matrix to obtain the sum of all data within the perception range of the roadside unit; each roadside unit randomly generates m noise matrices, the sum of the m matrices is 0, and distributes the matrices in sequence. Each roadside unit will receive m noise matrices, and the m noise matrices are added to obtain a final noise matrix. The final noise matrix is ​​added to the perception data matrix to obtain a noisy perception data matrix, and it is sent to the cloud server.

8. The privacy-preserving edge node vehicle-connected data aggregation method according to claim 1, characterized in that: When receiving the noisy perception data: the cloud server obtains the perception data matrix sent by the roadside unit, accumulates all the perception data matrices, eliminates zero-sum noise, obtains the element sum of the perception data matrix, averages the element sum, calculates the data item sum of the perception data matrix and divides it by the number of vehicles to obtain the final aggregation result.

9. An edge node vehicle networking system, characterized in that: It includes a cloud server, multiple vehicles and roadside units; and performs data processing based on the privacy-protected edge node vehicle-connected data aggregation method described in any one of claims 1 to 8.

10. Application of the privacy-preserving edge node vehicle-connected data aggregation method according to any one of claims 1 to 8, characterized in that: Perform anomaly detection on vehicles to identify vehicle behaviors that are inconsistent with normal practices.

Images