Multi-modal data space-time correlation anonymization processing method in vehicle-road cooperation scene

Abstract

The invention discloses a multi-modal data space-time association anonymization processing method in a vehicle-road cooperation scene, and relates to the technical field of data security. The method comprises the following steps: constructing a GIS road map of a traffic road according to road information by utilizing a GIS technology, dividing the traffic road into different anonymous areas according to a road coefficient, constructing a space-time association graph in each anonymous area according to a multi-modal data set of different vehicles in the same anonymous area, and constructing a space-time association graph in each anonymous area according to an edge coefficient between different nodes and corresponding association strength thereof. Obtaining association states among different nodes, constructing an association evaluation model of the nodes, carrying out anonymization processing on the multi-modal data set to obtain an anonymized data set, and judging whether the anonymized data set meets an anonymization standard or not; the overall efficiency of data anonymization can be improved, and the data security can be better ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of data security, and specifically to a method for spatio-temporal correlation anonymization processing of multi-modal data in a vehicle-road collaborative scenario. Background Art

[0002] In the vehicle-road collaborative scenario, through information interaction and collaboration among vehicles, road facilities, and the cloud, more efficient and safe traffic operation can be achieved. During this process, a large amount of multi-modal data will be generated. This data contains rich sensitive information. If not effectively protected, it may lead to the leakage of vehicle and user privacy; Most current data anonymization methods are for processing single-type data, and it is difficult to handle the complex spatio-temporal correlation relationships among multi-modal data in the vehicle-road collaborative scenario, resulting in low data security. Moreover, the existing technology's judgment mechanism for data that needs to be anonymized is often not precise enough, leading to low overall efficiency of data anonymization. In view of the deficiencies of the existing technology, the present invention provides a method for spatio-temporal correlation anonymization processing of multi-modal data in a vehicle-road collaborative scenario. Summary of the Invention

[0003] The purpose of the present invention is to provide a method for spatio-temporal correlation anonymization processing of multi-modal data in a vehicle-road collaborative scenario.

[0004] The purpose of the present invention can be achieved through the following technical solutions: A method for spatio-temporal correlation anonymization processing of multi-modal data in a vehicle-road collaborative scenario, including the following steps: Step S1: Obtain the road information of the traffic road, use GIS technology to construct a GIS road map of the traffic road according to the road information, obtain the road coefficient of the traffic road, and divide the traffic road into different anonymous regions according to the road coefficient in the GIS road map; Step S2: Obtain the multi-modal data sets of different vehicles in the same anonymous region, construct a spatio-temporal correlation graph within each anonymous region, obtain the edge coefficients and their corresponding correlation intensities between different nodes in the spatio-temporal correlation graph, and obtain the correlation status between different nodes according to the correlation intensity; Step S3: Construct an association evaluation model for nodes according to different multi-modal data sets and their corresponding correlation intensities, perform anonymization processing on the multi-modal data sets of vehicles to obtain corresponding anonymized data sets, and use the association evaluation model to judge whether the obtained anonymized data sets meet the anonymization standards.

[0005] Further, the process of obtaining the road information of the traffic road and using GIS technology to construct a GIS road map of the traffic road according to the road information includes: Collect road information of all traffic roads in a single area, including the route distribution and specification parameters of all traffic roads in the single area, the geographical locations of traffic intersections in all traffic roads, and the deployment locations of roadside devices on both sides of all traffic roads; Use GIS technology to construct a GIS road map of all traffic roads in the single area based on the collected road information.

[0006] Further, the process of obtaining the road coefficient of a traffic road and dividing the traffic road into different anonymous areas according to the road coefficient in the GIS road map includes: In the GIS road map, with each traffic intersection as the center, all traffic roads within its preset fixed range are used as its first anonymous area; Obtain the road coefficient P of the other traffic roads in the GIS road map except for the first anonymous area di , where i = 1, 2,..., n, i is the road number of each traffic road, and n is the total number of traffic roads in the GIS road map; ; is a preset weight value, C ai is the average daily traffic volume of the corresponding traffic road, C bi is the total number of roadside devices on both sides of the corresponding traffic road, excluding the roadside devices within the first anonymous area; According to the road coefficient of each traffic road, divide the other parts except for the first anonymous area into several second anonymous areas. The larger the road coefficient, the more the total number of second anonymous areas divided by the corresponding traffic road; Upload the divided anonymous areas to the GIS road map for synchronization, including the first anonymous area and the second anonymous area, and bind the roadside devices on both sides of the traffic road to their anonymous areas.

[0007] Further, the process of obtaining the multi-modal data set of different vehicles in the same anonymous area and constructing the spatio-temporal association graph in each anonymous area includes: Set the construction period. When a construction period is reached, incorporate the multi-modal data of all vehicles that appear in a single anonymous area within a single construction period into the same multi-modal data set; The multi-modal data includes the GPS trajectory information and V2X event information of the vehicle, the roadside monitoring information of the roadside device. The GPS trajectory information includes a timestamp and a trajectory point. The roadside monitoring information includes a timestamp and a monitoring image. The V2X event information includes a timestamp and an event content; Generate corresponding trajectory nodes, event nodes, and monitoring nodes for GPS trajectory information, V2X event information, and roadside monitoring information respectively, and construct a spatio-temporal association graph of the single anonymized area within the single construction cycle based on each node.

[0008] Further, the process of obtaining the edge coefficients and their corresponding association strengths between different nodes in the spatio-temporal association graph and obtaining the association status between different nodes according to the association strength includes: The edge coefficients include time similarity, space similarity, and semantic similarity. Obtain the time similarity S between two nodes in a single spatio-temporal association graph t , space similarity S v , and semantic similarity S w ; ; t 1 , t 2 are the timestamps corresponding to the two nodes, and t 0 is a preset time parameter; ; (x 1 , y 1 , z 1 ), (x 2 , y 2 , z 2 ) are the coordinates corresponding to the two nodes, and d 0 is a preset space parameter; ; A and B are the text contents corresponding to the two nodes; Obtain the association strength G between the two nodes according to the time similarity, space similarity, and semantic similarity between the two nodes; ; If the two nodes are in the first anonymized area, adopt preset weight values , and the sum of the three is 1. If the two nodes are in the second anonymized area, adopt preset weight values , and the sum of the three is 2; Set an association threshold, obtain the association strengths between the nodes in the single spatio-temporal association graph, and compare them with the association threshold respectively to obtain the association status between the nodes, including strong association status and weak association status. Upload the association strengths and their association status between different nodes as the edges between the corresponding nodes to the spatio-temporal association graph for synchronization.

[0009] Further, the process of constructing an association evaluation model for nodes according to different multimodal data sets and their corresponding association strengths includes: Generate an association evaluation set based on the multi-modal data sets in different anonymous regions during different construction cycles and the association strengths between various nodes in their spatio-temporal association graph, and divide the association evaluation set into a training set and a test set; Construct a convolutional neural network, use the different anonymous regions and their multi-modal data sets in the training set as the input data of the convolutional neural network, and use the association strengths between various nodes in the training set as the output data of the convolutional neural network; Use the training set to train the convolutional neural network to obtain an initial convolutional neural network, use the test set to verify the model of the initial convolutional neural network, and output the initial convolutional neural network with a test error threshold less than or equal to the preset value as the association evaluation model of the nodes.

[0010] Further, the process of anonymizing the multi-modal data set of the vehicle to obtain the corresponding anonymized data set includes: Anonymize the multi-modal data corresponding to each node marked as a strongly associated state in the spatio-temporal association graph respectively. The anonymization process includes trajectory generalization, monitoring perturbation, and event hiding; The trajectory generalization refers to generalizing the coordinates of each trajectory point in the GPS trajectory information into geographical regions. The monitoring perturbation refers to blurring each target in the monitoring image of the roadside monitoring information. The event hiding refers to hiding the event content in the V2X event information; Mark the multi-modal data after anonymization as anonymized data, and anonymize the multi-modal data in the same multi-modal data set that is in a strongly associated state respectively to obtain the corresponding anonymized data set.

[0011] Further, the process of using the association evaluation model to judge whether the obtained anonymized data set meets the anonymization standard includes: Input a single anonymized data set and its anonymous region into the association evaluation model, use the association evaluation model to output the association strengths between various nodes corresponding to the single anonymized data set, and obtain the association states between various nodes; If each node is in a weakly associated state, it is judged that it meets the anonymization standard. If each node is in a strongly associated state, it is judged that it does not meet the anonymization standard, and it is anonymized again until it becomes a weakly associated state.

[0012] Compared with the prior art, the beneficial effects of the present invention are: The present invention obtains the corresponding road coefficients according to the daily average traffic volume of each traffic road and the number of roadside devices, and divides all traffic roads into several different anonymous areas according to the road coefficients, so as to classify the multi-modal data anonymization problem into each separate sub-area for processing respectively. By incorporating the multi-modal data of different vehicles into the same multi-modal data set for unified processing, it is beneficial to improve the overall efficiency of data anonymization; By constructing the spatio-temporal correlation graph of each multi-modal data set and obtaining the correlation strength and correlation state between each node therein, it is possible to accurately determine whether there is a strong correlation relationship between different nodes, and perform anonymization processing on it to eliminate its connection. By constructing a correlation evaluation model, it is possible to quickly determine whether the data after anonymization processing meets the anonymization standard, and perform anonymization processing again on the non-conforming data, which can better ensure the security of the data. Brief Description of the Drawings

[0013] Figure 1 It is a flowchart of the present invention. Detailed Embodiment

[0014] As Figure 1 shown, the spatio-temporal correlation anonymization processing method for multi-modal data in the vehicle-road cooperation scenario includes the following steps: Step S1: Obtain the road information of the traffic road, use GIS technology to construct the GIS road map of the traffic road according to the road information, obtain the road coefficient of the traffic road, and divide the traffic road into different anonymous areas according to the road coefficient in the GIS road map; Step S2: Obtain the multi-modal data sets of different vehicles in the same anonymous area, construct the spatio-temporal correlation graph in each anonymous area, obtain the edge coefficient and its corresponding correlation strength between different nodes in the spatio-temporal correlation graph, and obtain the correlation state between different nodes according to the correlation strength; Step S3: Construct a correlation evaluation model for nodes according to different multi-modal data sets and their corresponding correlation strengths, perform anonymization processing on the multi-modal data set of the vehicle to obtain the corresponding anonymized data set, and use the correlation evaluation model to judge whether the obtained anonymized data set meets the anonymization standard.

[0015] It should be further noted that, in the specific implementation process, the process of obtaining the road information of the traffic road and using GIS technology to construct the GIS road map of the traffic road according to the road information includes: Collect road information of all traffic roads within a single area. The road information refers to various data related to traffic roads and necessary for constructing a GIS road map, including the route distribution and specification parameters of all traffic roads within the single area, the geographical locations of traffic intersections in all traffic roads, the deployment locations of roadside devices on both sides of all traffic roads, etc.; Use GIS technology to construct a GIS road map of all traffic roads within the single area based on the collected road information. In the constructed GIS road map, the road information of all traffic roads can be viewed. On this basis, road information of different areas can be obtained, and then their corresponding GIS road maps can be constructed. At this time, the GIS road map does not include the various data obtained subsequently.

[0016] It should be further noted that in the specific implementation process, the process of obtaining the road coefficient of a traffic road and dividing the traffic road into different anonymous areas according to the road coefficient in the GIS road map includes: In the GIS road map, taking a single traffic intersection as the center, all traffic roads within its preset fixed range are used as its first anonymous area. The first anonymous area is for the traffic intersection. The same method is used to obtain the first anonymous area corresponding to each traffic intersection in the GIS road map; Obtain the road coefficient of the other traffic roads in the GIS road map except for the first anonymous area, denoted as P di , where i = 1, 2,..., n, i is the preset road number of each traffic road, and n is the total number of traffic roads in the GIS road map; ; Among them, are all preset weight values, C ai is the average daily traffic volume of the corresponding traffic road, C bi is the total number of roadside devices on both sides of the corresponding traffic road, excluding the roadside devices within the first anonymous area. The larger the road coefficient, the more data the corresponding traffic road generates; According to the road coefficient of each traffic road, divide its other part except for the first anonymous area to divide it into several second anonymous areas with the same length. Among them, the larger the road coefficient, the more the total number of second anonymous areas divided by the corresponding traffic road; Upload the divided anonymous areas to the GIS road map for synchronization. The anonymous areas include the first anonymous area and the second anonymous area, and bind the roadside devices on both sides of the traffic road to their corresponding anonymous areas.

[0017] It should be further noted that in the specific implementation process, the process of obtaining the multi-modal data sets of different vehicles in the same anonymous area and constructing the spatio-temporal association graph in each anonymous area includes: Taking a single anonymous area as an example, set a construction period. When a construction period is reached, incorporate the multi-modal data of all vehicles that appear in this single anonymous area during this single construction period into the same multi-modal data set; The multi-modal data includes the GPS trajectory information and V2X event information of the vehicle, the roadside monitoring information of the roadside equipment. The GPS trajectory information includes the timestamp, trajectory points, etc. The roadside monitoring information includes the timestamp, monitoring images, etc. The V2X event information includes the timestamp, event content, etc.; For the GPS trajectory information, aggregate every 5 trajectory points into a trajectory node, and take the mean of the timestamps of these 5 trajectory points as the timestamp of this trajectory node. For the roadside monitoring information, regard each target in the monitoring image as a monitoring node, including vehicles, pedestrians, traffic signs, etc. For the V2X event information, regard each V2X event information as an event node; Construct the spatio-temporal association graph of this single anonymous area during this single construction period according to the various nodes (including trajectory nodes, monitoring nodes, event nodes) obtained above. In the constructed spatio-temporal association graph, each node is its corresponding timestamp in terms of time. In terms of space, the trajectory node and the event node are the coordinates corresponding to their timestamps, and the monitoring node is the coordinate of the center of the target it identifies; The spatio-temporal association graph at this time only contains each node and does not contain the edges between each node. Use the same method to construct the spatio-temporal association graphs of this single anonymous area in different construction periods and the spatio-temporal association graphs of different anonymous areas in different construction periods respectively.

[0018] It should be further noted that in the specific implementation process, the process of obtaining the edge coefficients between different nodes and their corresponding association strengths in the spatio-temporal association graph and obtaining the association status between different nodes according to the association strength includes: Taking a single spatio-temporal association graph as an example, the edge coefficients include time similarity, space similarity, and semantic similarity. Taking any two nodes in this single spatio-temporal association graph as an example, obtain the time similarity S t 、space similarity S v 、semantic similarity S w ; ; Among them, t 1 、t 2 are the timestamps corresponding to the two nodes respectively, with the unit of millisecond, and t 0 is a preset time parameter, equal to 1500 milliseconds; ; Wherein, (x 1 , y 1 , z 1 ), (x 2 , y 2 , z 2 ) are the coordinates corresponding to two nodes respectively, with the unit of meter, and d 0 is a preset spatial parameter, equal to 20 meters; ; Wherein, A and B are the text contents corresponding to two nodes respectively; According to the time similarity S t between two nodes, spatial similarity S v , and semantic similarity S w , obtain the association strength between the two, denoted as G; ; If two nodes are in the first anonymous area, adopt the preset weight value , and the sum of the three is 1. If two nodes are in the second anonymous area, adopt the preset weight value , and the sum of the three is 2; Set the association threshold G 0 , and adopt the same method to obtain the association strength between each node in this single spatio-temporal association graph, and compare it with the association threshold respectively to obtain the association state between each node. The association state includes strong association state and weak association state; If G > G 0 , then mark the corresponding two nodes as in the strong association state. If G ≤ G 0 , then mark the corresponding two nodes as in the weak association state, and upload the association strength and its association state between different nodes as the edges between the corresponding nodes to the spatio-temporal association graph for synchronization.

[0019] It should be further noted that in the specific implementation process, the process of constructing the association evaluation model of nodes according to different multi-modal data sets and their corresponding association strengths includes: According to the multi-modal data sets in different anonymous areas during different construction periods, and the association strength between each node in its spatio-temporal association graph, generate the corresponding association evaluation set, and divide the association evaluation set into a training set and a test set; Construct a convolutional neural network, use different anonymous areas and their multi-modal data sets in the training set as the input data of the convolutional neural network, and use the association strength between each node in the training set as the output data of the convolutional neural network; Train a convolutional neural network using a training set to obtain an initial convolutional neural network, verify the model of the initial convolutional neural network using a test set, and output the initial convolutional neural network with a test error threshold less than or equal to a preset value as the associated evaluation model of the node.

[0020] It should be further noted that in the specific implementation process, the process of anonymizing the multi-modal data set of the vehicle to obtain the corresponding anonymized data set includes: Anonymize the multi-modal data corresponding to each node marked as a strong association state in the spatio-temporal association graph respectively. The anonymization process includes trajectory generalization, monitoring perturbation, and event hiding; The trajectory generalization means generalizing the coordinates corresponding to each trajectory point in the GPS trajectory information into geographical regions. For example, converting longitude and latitude coordinates into the names of cities, regions, and streets; The monitoring perturbation means blurring each target recognized in the monitoring image in the roadside monitoring information. For example, adding Gaussian noise to the monitoring image to make it blurred; The event hiding means hiding the event content in the V2X event information. For example, deleting the acceleration and speed or replacing the specific values with range values.

[0021] Mark the multi-modal data after the above anonymization process as anonymized data, and anonymize the multi-modal data in a strong association state in the same multi-modal data set respectively to obtain the corresponding anonymized data set.

[0022] It should be further noted that in the specific implementation process, the process of using the associated evaluation model to judge whether the obtained anonymized data set meets the anonymization standard includes: Taking a single anonymized data set as an example, input the single anonymized data set and its anonymized area into the associated evaluation model, use the associated evaluation model to output the association strength between the corresponding nodes of the single anonymized data set, and obtain the association state between the nodes; If each node is in a weak association state, it is judged that it meets the anonymization standard and no other operations are performed on it. If each node is in a strong association state, it is judged that it does not meet the anonymization standard and it is anonymized again until it becomes a weak association state.

[0023] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A method for anonymizing spatiotemporal correlation of multimodal data in a vehicle-road collaboration scenario, characterized in that: The following steps are involved: Step S1: Obtaining road information of traffic roads, using GIS technology to construct a GIS road map of traffic roads according to the road information, obtaining road coefficients of traffic roads, and dividing the traffic roads into different anonymous areas according to the road coefficients in the GIS road map; Step S2: obtaining a multimodal data set of different vehicles in the same anonymous area, constructing a spatiotemporal association graph in each anonymous area, obtaining edge coefficients between different nodes and their corresponding association strengths in the spatiotemporal association graph, and obtaining the association status between different nodes according to the association strengths; Step S3: construct a node association evaluation model based on different multimodal data sets and their corresponding association strengths, anonymize the vehicle's multimodal data set to obtain a corresponding anonymized data set, and use the association evaluation model to determine whether the obtained anonymized data set meets the anonymization standard.

2. The method for anonymizing spatiotemporal correlation of multimodal data in a vehicle-road collaboration scenario according to claim 1 is characterized in that: The process of obtaining road information and building a GIS road map includes: Collect road information of all traffic roads in a single area, including route distribution and specification parameters of all traffic roads in the single area, geographical locations of traffic intersections on all traffic roads, and deployment locations of roadside equipment on both sides of all traffic roads; GIS technology is used to construct a GIS road map of all traffic roads in the single area based on the collected road information.

3. The method for anonymizing spatiotemporal correlation of multimodal data in a vehicle-road collaboration scenario according to claim 2 is characterized in that: The process of obtaining the road coefficient and dividing the traffic road into different anonymous areas includes: In the GIS road map, each traffic intersection is taken as the center, and all traffic roads within its preset fixed range are taken as its first anonymous area; Obtain the road coefficient P of other traffic roads except the first anonymous area in the GIS road map di , i=1, 2, ..., n, i is the road number of each traffic road, n is the total number of traffic roads in the GIS road map; ； is the preset weight value, C ai is the average daily traffic volume of the corresponding traffic road, C bi is the total number of roadside equipment on both sides of the corresponding traffic road, excluding the roadside equipment in the first anonymous area; According to the road coefficient of each traffic road, the other part except the first anonymous area is divided into a number of second anonymous areas. The larger the road coefficient is, the more the total number of second anonymous areas divided by the corresponding traffic road is. The divided anonymous areas are uploaded to the GIS road map for synchronization, including the first anonymous area and the second anonymous area, and the roadside equipment on both sides of the traffic road are bound to its anonymous area.

4. The method for anonymizing the spatiotemporal association of multimodal data in a vehicle-road collaboration scenario according to claim 3 is characterized in that: The process of obtaining a multimodal data set and constructing a spatiotemporal correlation graph includes: A build cycle is set, and when a build cycle is reached, multimodal data of all vehicles that appear in a single anonymous area within a single build cycle are included in the same multimodal data set; The multimodal data includes GPS track information and V2X event information of the vehicle, and roadside monitoring information of the roadside equipment, wherein the GPS track information includes a timestamp and a track point, the roadside monitoring information includes a timestamp and a monitoring image, and the V2X event information includes a timestamp and event content; Corresponding trajectory nodes, event nodes, and monitoring nodes are generated for GPS trajectory information, V2X event information, and roadside monitoring information, and a spatiotemporal correlation graph of the single anonymous area within the single construction cycle is constructed based on each node.

5. The method for anonymizing spatiotemporal correlation of multimodal data in a vehicle-road collaboration scenario according to claim 4 is characterized in that: The process of obtaining the association strength and association status between different nodes includes: The edge coefficients include temporal similarity, spatial similarity, and semantic similarity. The temporal similarity S between two nodes in a single spatiotemporal association graph is obtained. t , spatial similarity S v , semantic similarity S w ; ； t1 and t2 are the timestamps corresponding to the two nodes, and t0 is the preset time parameter; ； (x1, y1, z1) and (x2, y2, z2) are the coordinates of the two nodes, and d0 is the preset spatial parameter; ； A and B are the text contents corresponding to the two nodes; According to the temporal similarity, spatial similarity, and semantic similarity between two nodes, the correlation strength G between the two nodes is obtained; ； If two nodes are in the first anonymous area, the preset weight value is adopted , the sum of the three is 1. If two nodes are in the second anonymous area, the preset weight value is adopted. , the sum of the three is 2; Set the association threshold, obtain the association strength between each node in the single spatiotemporal association graph, and compare it with the association threshold to obtain the association status between each node, including strong association status and weak association status. The association strength and association status between different nodes are uploaded to the spatiotemporal association graph as the edges between corresponding nodes for synchronization.

6. The method for anonymizing the spatiotemporal association of multimodal data in a vehicle-road collaboration scenario according to claim 5 is characterized in that: The process of building a node's associated evaluation model includes: Generate an association evaluation set based on the multimodal data sets of different anonymous regions in different construction cycles and the association strength between each node in their spatiotemporal association graph, and divide the association evaluation set into a training set and a test set; Construct a convolutional neural network, use the different anonymous regions in the training set and their multimodal data sets as the input data of the convolutional neural network, and use the correlation strength between each node in the training set as the output data of the convolutional neural network; The convolutional neural network is trained using the training set to obtain an initial convolutional neural network, the initial convolutional neural network is model verified using the test set, and the initial convolutional neural network that is less than or equal to a preset test error threshold is output as the node association evaluation model.

7. The method for anonymizing spatiotemporal association of multimodal data in a vehicle-road collaboration scenario according to claim 6 is characterized in that: The process of anonymizing a multimodal data set to obtain an anonymized data set includes: Anonymizing the multimodal data corresponding to each node marked as a strong correlation state in the spatiotemporal correlation graph, wherein the anonymization processing includes trajectory generalization, monitoring disturbance, and event hiding; The trajectory generalization refers to generalizing the coordinates of each trajectory point in the GPS trajectory information into a geographical area, the monitoring disturbance refers to blurring each target of the monitoring image in the roadside monitoring information, and the event hiding refers to hiding the event content in the V2X event information; The multimodal data after anonymization processing is marked as anonymized data, and the multimodal data in a strongly associated state in the same multimodal data set are anonymized separately to obtain the corresponding anonymized data set.

8. The method for anonymizing the spatiotemporal association of multimodal data in a vehicle-road collaboration scenario according to claim 7 is characterized in that: The process of determining whether an anonymized data set meets the anonymization criteria includes: Inputting a single anonymized data set and its anonymous region into the association evaluation model, using the association evaluation model to output the association strength between each node corresponding to the single anonymized data set, and obtaining the association status between each node; If each node is in a weakly associated state, it is judged to meet the anonymization standard. If each node is in a strongly associated state, it is judged to not meet the anonymization standard and is re-anonymized until it becomes a weakly associated state.

Images