Method for anonymizing vehicle data
By determining static and dynamic variables in vehicles, using machine learning clustering mechanisms and deep neural network technology to generate and transmit anonymous variables of similar vehicles, the problem of protecting user information and service quality when anonymous vehicle data is solved, and the problem of protecting user information and service quality is realized. Efficient data anonymization and security.
Patent Information
- Application Number
- CN202380033725.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2022-05-17
- Filing Date
- 2023-04-24
- Publication Date
- 2024-11-22
AI Technical Summary
The prior art is difficult to effectively protect the personal information of vehicle users when anonymizing vehicle data, while ensuring data accuracy for external services for vehicles. Especially in large fleets and non-homogeneous vehicle user groups, similar vehicle usage profiles exist. Under the circumstances, data protection and service quality are difficult to take into account.
By installing sensors in the vehicle to determine static and dynamic variables, using machine learning clustering mechanisms to generate similarity metrics, transmit variables of similar vehicles or manually generated variables to external services of the vehicle, creating a privacy layer to replace real information, combined with principal components Analysis and deep neural networks perform dimension reduction and encryption to achieve data anonymization.
Effectively hide vehicle information worthy of protection, ensure data protection, and provide high-quality external vehicle services, reducing data transmission delays and improving data security.
Smart Images

Figure CN119013711A_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a method for anonymizing vehicle data for the use of services external to the vehicle and also to a vehicle suitable for using the method. Background Art
[0002] In today's vehicles, many data from vehicles in a fleet are cyclically exchanged with servers outside the vehicle, such as servers of the vehicle manufacturer, fleet operator or provider of services outside the vehicle. With the technological advances and standard installation of vehicle sensor systems and sensor systems inside the vehicle, as well as the exchange of information recorded by these sensor systems, powerful monitoring possibilities for people, their data and their status are created at locations outside the vehicle. In this case, this data can contain personal data of the vehicle user or corresponding vehicle data that allow conclusions to be drawn about the person of the vehicle user. In particular, such data can be used to create a user profile, to create a kind of "fingerprint" of the corresponding vehicle, etc. Therefore, it needs to be protected to a greater or lesser extent from a data protection point of view. On the one hand, people who use the vehicle are concerned about the use of their data, but at the same time they want to be able to continue to use existing services.
[0003] DE 10 2020 003 188 A1 describes a method for protecting the personal data of a vehicle occupant, wherein the emotional state of the vehicle occupant in the vehicle is recorded as personal data and the masked emotional state is evaluated in a device outside the vehicle. It is then associated with an emotion group containing a plurality of emotion patterns, thereby depersonalizing the emotional state of the vehicle occupant to protect him.
[0004] DE 10 2021 001 378 B3 (not pre-published at the time of application) describes a method for anonymizing the movement data of road users equipped with position detection devices. The main goal here is to monitor traffic flow, however, in this case, the data that needs to be protected should be prevented from being transmitted, but there should still be sufficient data accuracy to ensure a reliable assessment of the traffic flow.
[0005] DE 10 2015 213 393 A1 describes a method for anonymizing vehicle data for the use of services outside the vehicle. In this case, in addition to the original vehicle data indicating the route section traveled by the vehicle, further artificial vehicle data indicating the artificially traveled route are generated. The original vehicle data and the artificial vehicle data are stored or transmitted together.
[0006] DE 10 2015 226 650 A1 discloses a method for anonymously transmitting a first value of a driving parameter of a vehicle to an external data receiving unit. Further values of the driving parameter are received from the vehicle, which values are transmitted to the vehicle from other vehicles. A second value of the driving parameter is calculated from the first value and the further values, so that the first value cannot be reconstructed by the external data receiving unit. The second value is transmitted to the external data receiving unit. Summary of the invention
[0007] In particular, it is an object of the present invention to provide an improved method for anonymizing vehicle data when using vehicle-external services.
[0008] According to the invention, this object is achieved by a method having the features of claim 1, in particular in the characterizing part of claim 1. Advantageous embodiments and improvements of the method according to the invention are given in the dependent claims. Claim 13 also describes a vehicle suitable for carrying out the method.
[0009] The method according to the invention is used to anonymize vehicle data for the use of vehicle external services. In this case, it is assumed that in a large fleet of identical vehicles and an associated large non-homogeneous / heterogeneous group of vehicle users, similar vehicle usage profiles exist with high probability. Furthermore, vehicle external services are used and when these services are used, information is sent from the vehicle to the provider of the vehicle external services.
[0010] Sensors installed in and on the vehicle continuously determine vehicle sensor values describing the vehicle state (e.g. position, speed, travel route, route information, vehicle age, etc.) and the state of the driver and passengers (e.g. attention level, driving style, etc.). According to the invention, within a fleet of vehicles of the same type with a non-homogeneous group of vehicle users, a first variable at least indirectly dependent on an integer number n of recorded vehicle sensor values and a second variable at least indirectly dependent on an integer number m of current vehicle sensor values are detected for each of the vehicles.
[0011] Therefore, two variables are determined for each vehicle in the fleet. The first variable is a feature set of up to n vehicle sensor values and is determined by the recorded vehicle sensor values. It is approximately constant in time (e.g., load set data), so it can be said here that it is a static variable. In the extreme case, the variable can also correspond directly to a single sensor value. In addition, the second variable is a feature set of up to m vehicle sensor values. It is determined by the current vehicle sensor value and changes dynamically over time (e.g., position, speed, acceleration, etc.), so it can be said here that it is a dynamic variable. In the extreme case, this variable can also correspond directly to a single sensor value.
[0012] The first variable, especially when it is based on multiple vehicle sensor values, makes it possible to identify each vehicle, like a fingerprint, because this variable is based on the individual vehicle and usage behavior and changes only slightly over time. Therefore, these first variables are particularly worthy of protection from a data protection point of view. On the other hand, the second variable reflects current data on the vehicle and its use. The identification of the vehicle can only be carried out indirectly (for example, through a combination of geolocation and local image information).
[0013] The similarity of the first variable is now determined for several or all vehicles in the fleet, according to which the vehicles are divided into classes that are similar or into classes that are dissimilar with respect to the similarity of at least one variable. In particular, vehicles having similar vehicle sensor values in the first and / or second variable can be determined. To this end, a preferred design of the present invention uses a clustering mechanism of machine learning (e.g., K-means, mean shift or expectation maximization (EM) clustering) to generate a similarity measure for the first and / or second variable. In this case, the similarity of at least the second variable is determined continuously, since the latter changes over time in the fleet.
[0014] Now, in the case of a request for a vehicle-external service, instead of transmitting variables of the requesting vehicle containing information related to the service, corresponding variables of a vehicle from a category of similar vehicles, calculated variables of corresponding variables of multiple vehicles from a category of similar vehicles, or artificially generated similar variables can be transmitted.
[0015] Thus, a privacy layer is created which, in the simplest case, uses sufficiently similar information from another vehicle in the fleet instead of the vehicle's real information for the respective vehicle-external service. In another design, the relevant information about a set of relevant information about similar vehicles can be determined, for example averaged or artificially generated.
[0016] Thus, when a first vehicle calls a vehicle-external service, variables related to the service are transmitted, which are determined (eg by averaging) from variables of similar vehicles or artificially generated data information.
[0017] Furthermore, usually not all information sent to the provider of the vehicle-external service is necessary for providing the service, but a large amount of information not relevant for the service is transmitted to the provider of the vehicle-external service, so that this additional information can be monetized as added value by the provider of the vehicle-external service. Usually, this information is based on a respective further variable. Therefore, according to a very advantageous development of the method according to the invention, it can be provided that a respective variable of a vehicle from a class of dissimilar vehicles, a calculated variable of respective variables of a plurality of vehicles from a class of dissimilar vehicles or an artificially generated similar variable is transmitted instead of further variables of the requesting vehicle.
[0018] Therefore, we implemented a “privacy layer” that replaces the vehicle’s information with information about similar vehicles related to services outside the vehicle and irrelevant information about dissimilar vehicles, thereby hiding the actual information about the vehicle that is worth protecting.
[0019] According to an advantageous embodiment of the invention, artificially generated variables are possible in both cases, but play a decisive role in particular in the case of less relevant information, and can be generated by generative machine learning methods.
[0020] To this end, models can be trained and optimized to generate variables that are as "real" as possible, which can then be used for random variables, especially second variables. This "realism" is achieved by using appropriate functions, which are either generic or defined according to the services that the vehicle wishes to use. Thus, it is also possible to set the content focus of n or especially m vehicle sensor values.
[0021] According to a very advantageous embodiment of the method according to the invention, the variables, i.e. the first variable and the second variable, which are at least indirectly dependent on an integer number of vehicle sensor values, can be formed as a set of n or m vehicle sensor values. The variables thus consist of the corresponding sets or, if a specific arrangement of the sensor values in a predetermined order is required, also of n-tuples or m-tuples.
[0022] An alternative design of the respective variables may provide that the respective variables are formed as vectors based on the respective vehicle sensor values in an n-dimensional or m-dimensional space.
[0023] Therefore, two vectors are determined for each vehicle in the fleet, the first variable being a characteristic n-dimensional vector determined from the recorded vehicle sensor values. The second vector corresponding to the second variable will be an m-dimensional vector determined from the current vehicle sensor values.
[0024] In particular, the first vector allows very simple identification of each vehicle, like a fingerprint, and is therefore particularly worth protecting. The second vector essentially reflects the current value of the vehicle during use and is less worth protecting in terms of data protection.
[0025] According to another very advantageous embodiment of the method of the invention, as an improvement of its embodiment as a vector, the two variables can also be implemented in such a way that the corresponding variables are formed as a mapping of the corresponding vector to a value with a smaller or at most the same dimension. This mapping can preferably represent a reduction in the spatial dimension. With the help of a suitable mapping / transformation, for example, a principal component analysis called PCA according to the English term "Principal Component Analysis", or also with the help of a mapping of a deep neural network, corresponding values with n' or m' dimensions can be generated from vectors with n or m dimensions, respectively. In this case, the number of dimensions is preferably reduced or at most equal to the number of previous dimensions, that is, never increased. Since the vector with n or m dimensions is now mapped into a space with n' or m' dimensions, respectively, it is compressed and encrypted in a certain way, so that it is no longer possible to directly infer the previous vector or the vehicle sensor value on which the vector is based from the new value. Therefore, if the provider of vehicle external services can process such a vector or preferably the mapped value generated from the vector to provide their services, a certain degree of anonymization can be achieved only in this way. Furthermore, if this is combined with the above-mentioned function of the “privacy layer”, very good data protection is achieved by transmitting relevant similar information and, if necessary, less relevant dissimilar information.
[0026] According to a very advantageous design of the method of the present invention, it can now also be provided that, in order to determine the categories of similar and dissimilar variables, a clustering mechanism of machine learning is used to determine a similarity measure, wherein the variables are classified based on default values and the determined similarity measure is compared with the default values. Therefore, this advantageous design of the method of the present invention uses a clustering mechanism of machine learning. These can include, for example, K-means, mean shift or expectation maximization (EM) clustering. Then a similarity measure of the first variable and / or the second variable can be generated by this mechanism, for example as a vector or mapping of corresponding vectors. Based on the default value, the determined similarity measure can then be classified by the default value. If the similarity measure is, for example, between 0 and 100%, then the dissimilar information between 0 and 80% or the sufficiently similar information between 80 and 100% can be distinguished by, for example, a default value of 80%.
[0027] In this case, according to a very advantageous embodiment of the method according to the invention, the default value can be varied depending on the vehicle-external service. For example, the data of the relevant vehicle sensors or their absolute values offer other possibilities for changing or parameterizing the value, since, for example, a different similarity can be used at low speeds than at correspondingly high speeds. An example here could be a weather application, for which the similarity with regard to the exact position is relatively low, so that it is sufficient to determine the position with an accuracy of a few kilometers or a few hundred meters. However, if the vehicle-external service is a navigation system, such an indication of the position is of course not sufficient; for this, the similarity between the values is much higher.
[0028] Another very advantageous variant of the method according to the invention provides that the variables are at least partially exchanged between the vehicles and a central data center, wherein the aggregation and evaluation of the information takes place in the central data center. Thus, information from and about the vehicles of the fleet can be exchanged via a data center, in particular a cloud, and the information can be aggregated and evaluated in the data center / cloud. Alternatively or additionally, according to the invention, it can also be provided that the variables are at least partially exchanged, aggregated and evaluated between the vehicles of the fleet. In a second method, the information can be aggregated and evaluated decentralized in the fleet by exchanging information between the vehicles in the fleet. In this case, it is advantageous that the vehicle can communicate directly decentralized with other nearby vehicles, in particular with vehicles with similar dynamic values (i.e. vehicles with similar current values (e.g. position, speed, acceleration, etc.)) because they are in the same location, for example traveling in the same direction.
[0029] While the former solution reduces data transmission time and latency, decentralized information processing can improve data security (because there is no central location storing all the information about the fleet) and is robust to the failure of individual nodes, especially central nodes. The two solutions can also be combined to allow for partially centralized and partially decentralized operations.
[0030] According to a very advantageous embodiment of the idea, in order to determine the difference between information that is relevant for a vehicle-external service and information that is less relevant for at least some vehicles in the fleet, information with similar first variables and information with respectively dissimilar second variables can be sent to the vehicle-external service, after which the service is evaluated. Thus, the fleet is used to test the vehicle-external service in order to determine which information is relevant for the service result (analysis). To this end, a plurality of orchestration requests with the first variable and the second variable are sent to the provider of the vehicle-external service, which cover a range of possible values for the requests. The responses to the requests transmitted by the provider of the vehicle-external service are analyzed for similarity in the data center / cloud and / or in the fleet.
[0031] To this end, a group of vehicles in the fleet first sends a request to a provider of the vehicle external service that has similar data on one of the two variables but different data in the corresponding other variable. It can thus be determined whether one of the variables itself is sufficient to use the vehicle external service.
[0032] A very advantageous design of the method provides that individual vehicle sensor values are subsequently recorded and other vehicle sensor values are randomized in order to determine vehicle sensor values that are relevant and less relevant for the use of the corresponding vehicle-external service. Thus, based on the above-mentioned query, individual data or groups of data from the variables are recorded and the remaining data are randomized during the request in order to obtain a final group that is relevant for the vehicle-external service. Thus, value ranges that are far from the corresponding variable can also be utilized to the maximum extent by means of a "privacy layer".
[0033] The method according to the invention is therefore now particularly suitable for use in vehicles of a fleet of essentially identical vehicles, i.e., for example, a fleet of passenger cars for private use, passenger cars used as company cars, vehicles of a brand or brand combination, commercial vehicles, etc. In this case, the vehicle according to the invention is equipped with a plurality of sensors and at least one communication interface, which is configured to carry out the method together with other vehicles and / or an external data center. BRIEF DESCRIPTION OF THE DRAWINGS
[0034] Further advantageous embodiments of the method according to the invention and of a vehicle for carrying out the method according to the invention can also be derived from the exemplary embodiments described in more detail below with reference to the drawings.
[0035] in:
[0036] Figure 1 A basic scenario for applying the method according to the invention is shown;
[0037] Figure 2 An exemplary two-dimensional representation of vehicle information is shown;
[0038] Figure 3 Shows Figure 2 An exemplary one-dimensional compression of a representation of ;
[0039] Figure 4 A first possible implementation for implementing the method according to the invention is shown; and
[0040] Figure 5 A second possible embodiment for implementing the method according to the invention is shown. DETAILED DESCRIPTION
[0041] exist Figure 1In the illustration of , a first vehicle 11 is shown, which is supposed to have a plurality of sensors (not shown) and a communication interface 1, via which a communication connection can be established to a cloud 2, via which a provider of a vehicle external service FES offers its services. Between the vehicle 11 and the cloud 2 of the provider of the FES there is a "privacy layer" denoted by 3, in which the vehicle data are anonymized for the use of the FES by the vehicle 11. For this purpose, vehicle data of other vehicles denoted by 12, 13, 14 in a fleet 10 of vehicles 11, 12, 13, 14 of the same type with a non-homogeneous / heterogeneous group of vehicle users are used.
[0042] In such a fleet 10 consisting of vehicles 11, 12, 13, 14 of the same type, and the accompanying large non-homogeneous group of vehicle users, similar vehicle data in the form of usage profiles / user profiles exist with a very high probability. This is now utilized to implement the privacy layer 3 indicated here.
[0043] The vehicle sensors installed in and on the respective vehicles 11, 12, 13, 14 of the fleet 10 continuously determine vehicle sensor values, which contain the status of the vehicle 11, 12, 13, 14, such as its position, speed, driving route, route information, as well as information about vehicle age, wear, upcoming service intervals, etc. In addition, the sensors can also determine the status of the driver of the vehicle 11, 12, 13, 14 or a person accompanying him, for example by detecting the driving style, but also by detecting the attention level of the vehicle driver via an attention monitoring module. For this purpose, for example, the eyelid twitching frequency and / or other physical data are evaluated.
[0044] Two characteristic vectors are now determined for each of the vehicles 11, 12, 13, 14. These represent the variables of the invention described here as an example which are at least indirectly dependent on the sensor values and can in principle also be replaced by a set of sensor values. However, in the following embodiments, these vectors are essentially regarded as variables.
[0045] Therefore, two vectors are determined for each vehicle in the fleet, namely a characteristic static n-dimensional value WK determined from the recorded vehicle sensor values and approximately constant in time. For example, this can be load collection data. The second vector is determined as a dynamic m-dimensional value WZ from the current vehicle sensor values, such as position, velocity, acceleration, etc. It will change continuously over time.
[0046] Firstly, the first vector WK now makes it possible to identify the respective vehicle 11, 12, 13, 14 of the fleet 10, since this value is based on the individual vehicle data and the usage behavior of the vehicle user. It does not change over time and is therefore particularly worth protecting from a data protection point of view. On the other hand, the second vector WZ reflects the current value of the vehicle 11, 12, 13, 14 and its usage. It is possible to identify the vehicle 11, 12, 13, 14 only indirectly via this value, for example by a combination of position data with local image information, etc. Therefore, from a data protection point of view, its protection value is slightly lower than that of the first vector.
[0047] The two vectors WK, WZ can now preferably be processed by space dimensionality reduction, for example by a suitable mapping, such as PCA, deep neural network, etc. Thus, WK_n with n dimensions becomes a space WK' with n' dimensions. Here, n'<=n. Thus, the space WK with n dimensions is mapped to the space WK' with n' dimensions, and is thus compressed and encrypted in some way. Then it can be used similarly for WZ.
[0048] In particular, vehicles 11, 12, 13, 14 with similar sensor values in WK' and / or WZ' space can be determined. The exemplary specific design scheme uses a machine learning clustering mechanism (e.g., K-means, mean shift, or expectation maximization (EM) clustering) to generate similarity metrics in WK and / or WZ space. In this case, similarities in WZ or WZ' space are continuously determined as these values change over time in the fleet 10.
[0049] exist Figure 2 In the diagram of , an exemplary two-dimensional WK representation of similar and dissimilar vehicle information is shown accordingly. Similar vehicle information is represented by a dashed border with an "×". These similarities are determined by the described clustering mechanism. Here, it is purely based on the two-dimensional first vector WK as an example and can of course be represented similarly to the corresponding second vector WZ. In this case, the number of dimensions can be much higher than the two dimensions shown here, but these dimensions are particularly suitable for a graphical representation.
[0050] Therefore, in Figure 3 As can be seen in the diagram Figure 2 A one-dimensional compression of the illustration. Therefore, the spatial dimensionality reduction or mapping WK' is represented here accordingly. Here, similar vehicles 11, 12, 13, 14 or vehicle information are again shown between the dashed lines and marked with "×", while vehicle data located outside these dashed lines, i.e. to the right and left next to them, are dissimilar vehicle data and have no additional markings.
[0051] In accordance with Figure 4In a first variant of the illustration, information from vehicles 11, 12, 13, 14 of a fleet 10 can be exchanged via a data center / cloud 4 and can be aggregated and evaluated in the data center / cloud 4. In this case, this central data center 4 represents a trusted entity. For example, this can be a backend server of a vehicle manufacturer.
[0052] When the first vehicle 11 calls the vehicle external service FES, vehicle information 6 related to the service is transmitted from the WK or WZ of a similar vehicle 12, or information 6 determined (e.g., by averaging) from its WK or WZ, instead of its vehicle information 5. In the case where the vehicle information 5 of the vehicle 11 is not related to the service, those vehicle information 7 from the WK and WZ of a dissimilar vehicle 13 or a random value are transmitted instead of its own vehicle information 5. Thus, we realize a "privacy layer" that replaces the information 5 of the vehicle 11 with the information 6 of the similar vehicle 12 related to the FES and the irrelevant information 7 of the dissimilar vehicle 13, thereby hiding the actual information 5 of the vehicle.
[0053] Generative machine learning methods (e.g. adversarial networks, decoder-encoder-networks) can be used to improve the artificial generation of random WZ. To this end, a model is trained and optimized to generate target vectors that are as "realistic" as possible, which can then be used for random WZ values. This "realism" is achieved by using appropriate features, which are either generic or defined based on the services that the vehicle driver wishes to use. Therefore, the content focus of the m-dimension can also be set.
[0054] In another embodiment, the FES is only allowed to access the space WK' or WZ' and the mapping specification. As a result, the user-specific value of WK cannot be reproduced without further knowledge, because the information in WK' is compressed (cf. Figure 3 ) and use an irreversible mapping. It is important here to combine information from multiple vehicles 11, 12, 13, 14 in order to reduce the dimensions of the spaces WK and WZ to WK' and WZ', respectively, so as to achieve a sufficiently good compression and identify similar vehicles 11, 12 in the space, thereby ensuring anonymity.
[0055] Alternatively or additionally, the information 5 can be aggregated and evaluated in a decentralized manner in the convoy 10 by exchanging information between the vehicles 11 , 12 , 13 of the convoy. Figure 5 Similar to Figure 4 While the former solution reduces data transmission time and latency, decentralized / decentralized information processing can improve data security (because no central location stores all information about the fleet 10) and is robust to failures of individual nodes, especially central nodes.
[0056] In the following, a fleet 10 is used to test the FES to determine which information is relevant to the results of the service. To this end, multiple orchestration requests with values WK and WZ are sent to the provider or cloud 2, which covers the possible value range of the request. The similarity of the response to the request transmitted by the FES is analyzed in the data center / cloud 4 and / or the fleet 10. To this end, a group of vehicles 11, 12, 13, 14 in the fleet 10 first send requests to the FES, which have similar values in one of the two value ranges (WK or WZ), but have dissimilar values in the corresponding other value range. Therefore, it can be determined whether a single category of values is sufficient to use the FES. On this basis, a single value or value group from WK and WZ is recorded, and the remaining values are randomized during the request to obtain the final group of vehicle information related to the use of the FES. Therefore, a value range as far away from WK as possible can also be utilized.
[0057] In the example of an application of a weather application provided by FES, information related to the FES application is determined by a scheduled fleet query. In this case, the location data and destination address stored in the navigation system and the arrival time predicted by the navigation system are identified as relevant information. If the first vehicle 11 now uses the weather application, it is not the location data of the first vehicle 11 that is transmitted, but the location data of the second vehicle 12 from a category with similar location data, or the location data generated by the average value of the vehicle 12 from a category with similar location data. Similarly, instead of the destination address of the first vehicle 11, the destination address of the third vehicle 12 from a category with a similar destination address or the destination address averaged by multiple similar third vehicles 13 is transmitted to the weather application. Similarly, the time to arrive at the destination is used by a fourth vehicle 12 with a similar arrival time or the average arrival time of a class of similar vehicles 12. For data unrelated to the use of FES, data of other vehicles 13 in categories that are not similar to the first value of the first vehicle 11 (or average data from these vehicles) can be used.
Claims
1. A method for anonymizing vehicle data for use of a vehicle external service (FES), It is characterized in that In a fleet (10) of vehicles (11, 12, 13, 14) of the same type with a non-homogeneous group of vehicle users, a first variable which depends at least indirectly on an integer number n of recorded vehicle sensor values and a second variable which depends at least indirectly on an integer number m of current vehicle sensor values are detected for each of the vehicles (11, 12, 13, 14), wherein a similarity of at least one variable is determined for a plurality of or all vehicles (11, 12, 13, 14) of the fleet (10), according to which the vehicles (11, 12, 13, 14) are divided into classes which are similar or dissimilar with respect to the similarity of the at least one variable, wherein in the case of a request for a vehicle external service (FES), instead of transmitting variables of the requesting vehicle (11) containing information relevant to the service, corresponding variables of vehicles (12) from the class of similar vehicles (12), calculated variables of corresponding variables of a plurality of vehicles (12) from the class of similar vehicles (12) or artificially generated similar variables are transmitted.
2. The method according to claim 1, It is characterized in that In the event that the vehicle external service (FES) requires further information that is less relevant to the execution of the service, corresponding variables of a vehicle (13) from a category of dissimilar vehicles (13), calculated variables of corresponding variables of a plurality of vehicles (13) from a category of dissimilar vehicles (13), or artificially generated similar variables are transmitted based on corresponding other variables instead of transmitting the variables of the requesting vehicle (11).
3. The method according to claim 1 or 2, It is characterized in that The random variables are generated by a generative machine learning method.
4. The method according to any one of claims 1, 2 or 3, It is characterized in that The corresponding variable is formed as a set of n or m vehicle sensor values.
5. The method according to any one of claims 1, 2 or 3, It is characterized in that The corresponding variables are formed as vectors (WK, WZ) in an n-dimensional or m-dimensional space based on the corresponding vehicle sensor values, respectively.
6. The method according to claim 5, It is characterized in that The corresponding variables are formed as a mapping of the corresponding vectors (WK, WZ) to values (WK', WZ') having smaller or at most the same dimension.
7. The method according to any one of claims 1 to 6, It is characterized in that To determine categories of similar and dissimilar variables, a machine learning clustering mechanism is used to determine a similarity measure, wherein the variables are classified based on default values and the determined similarity measure is compared to the default values.
8. The method according to claim 7, It is characterized in that The default values are parameterized or defined as a function of relevant vehicle sensor values, their absolute values and / or vehicle external services (FES).
9. The method according to any one of claims 1 to 8, It is characterized in that The variables are at least partially exchanged between the vehicles (11, 12, 13, 14) and a central data center (4), wherein aggregation and evaluation of the information takes place in the central data center (4).
10. The method according to any one of claims 1 to 9, It is characterized in that The variables are at least partially exchanged, aggregated and evaluated between the vehicles (11, 12, 13, 14) of the fleet (10).
11. The method according to any one of claims 1 to 10, It is characterized in that In order to determine the difference between information relevant to the vehicle external service (FES) and information less relevant for at least some vehicles (11, 12, 13, 14) in the fleet (10), information with similar first variables and information with respectively dissimilar second variables are sent to the vehicle external service (FES), after which the service is evaluated.
12. The method according to claim 11, It is characterized in that Individual vehicle sensor values are then recorded, and other vehicle sensor values are randomized to determine vehicle sensor values that are relevant and less relevant to usage of a corresponding vehicle external service (FES).
13. A vehicle (11, 12, 13, 14) having a plurality of sensors and at least one communication interface (1), which is configured to carry out the method together with other vehicles (11, 12, 13, 14) and / or an external data center (4).
