Internet of vehicles identity authentication and privacy protection method and system based on federated learning
By adopting federal -based learning -based authentication and privacy protection methods in the Internet of Vehicle, the problem of model training and data privacy security in the Internet of Vehicle network has been solved, and the effective protection of data privacy and the improvement of the security of the Internet of the Internet.
Patent Information
- Application Number
- CN202411661846.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-20
- Publication Date
- 2025-01-21
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
In the Internet of Vehicles, how to balance model training and data privacy security is a technical problem that needs to be solved urgently.
Adopt a federal -based Internet -based authentication authentication and privacy protection method to authenticate the identity by obtaining the federal chain identification code of the vehicle, and only transmits the encrypted model parameters during the model training process, instead of transmitting the original data to ensure data privacy privacy And security.
Effectively ensure the local data privacy and security of the vehicle, and achieve the training of global models through federal learning, and improve the safety and privacy protection capabilities of the networking system.
Smart Images
Figure CN119337428A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of Internet of Vehicles data security management, and in particular to an Internet of Vehicles identity authentication and privacy protection method and system based on federated learning. Background Art
[0002] Through the new generation of information and communication technology, the Internet of Vehicles realizes all-round network connection between vehicles and cloud platforms, vehicles and vehicles, vehicles and roads, vehicles and people, and vehicles, mainly realizing the "three-network integration", that is, integrating the in-vehicle network, inter-vehicle network and in-vehicle mobile Internet. The Internet of Vehicles uses sensor technology to perceive the status information of vehicles, and uses wireless communication networks and modern intelligent information processing technologies to realize intelligent management of traffic, intelligent decision-making of traffic information services and intelligent control of vehicles.
[0003] In order to realize intelligent decision-making of the Internet of Vehicles and intelligent control of vehicles, data transmission and reception are essential. Especially when training models, how to balance model training and data privacy security is a technical problem that needs to be solved urgently. Summary of the invention
[0004] One of the purposes of the present invention is to provide a method and system for Internet of Vehicles identity authentication and privacy protection based on federated learning, which has solved the above-mentioned technical problems.
[0005] An embodiment of the present invention provides a method for identity authentication and privacy protection of Internet of Vehicles based on federated learning, comprising:
[0006] Get the federated chain identification code assigned to the historical federated learning task sent by the vehicle;
[0007] Authenticate the vehicle’s identity based on the federated chain identification code;
[0008] When authentication is passed, public parameters and model training tasks are sent to the vehicle;
[0009] Receive the return data returned by the vehicle after executing the model training task;
[0010] The global model parameters are updated and adjusted based on the returned data.
[0011] Preferably, the allocation rules of the federated chain identification code are as follows:
[0012] Construct a distributed topology structure corresponding to the federated learning task;
[0013] According to the task number of the federated learning task and the device number of each data transfer device in the distributed topology structure, they are numbered in order by level;
[0014] Determine the number corresponding to the position of the vehicle in the distributed topology structure as the federation chain identification code.
[0015] Preferably, the authentication rules are as follows:
[0016] When the number of federated chain identification codes is greater than or equal to a preset number threshold, extract the federated chain identification code corresponding to the nearest number threshold and construct it as authentication data;
[0017] Matching the authentication data with verification data in a pre-configured authentication repository;
[0018] When there is a matching item, the authentication is passed;
[0019] When the number is less than the preset number threshold, the obtained federated chain identification codes are arranged in order to form a checksum data string;
[0020] Match the data string to be verified with the standard data string in the corresponding pre-authentication library;
[0021] When there is a matching item, authentication is performed in a manner that assists authentication of the instruction data of the authentication interface.
[0022] Preferably, the Internet of Vehicles identity authentication and privacy protection method based on federated learning also includes:
[0023] Before sending the model training task to the vehicle, send the data classification statistical rules to the vehicle;
[0024] Receive the data classification and statistical results corresponding to the data classification and statistical rules returned by the vehicle;
[0025] Based on the data classification and statistical results, determine whether to send standard data in a pre-configured standard database to the vehicle;
[0026] When it is determined that the data needs to be sent, the standard data to be sent is extracted from the standard database based on the statistical results of the data classification and sent.
[0027] Preferably, the Internet of Vehicles identity authentication and privacy protection method based on federated learning also includes:
[0028] Update the data classification statistics results based on the standard data sent;
[0029] Using the updated data classification statistical results, retrieve the corresponding verification data from the pre-configured verification library;
[0030] The returned data is verified based on the verification data, and the returned data that fails the verification is discarded.
[0031] The present invention also provides a vehicle network identity authentication and privacy protection system based on federated learning, including: an acquisition module, an authentication module and a training module;
[0032] Among them, the acquisition module obtains the federal chain identification code assigned when the vehicle participates in the historical federated learning task; the authentication module authenticates the identity of the vehicle based on the federal chain identification code; when the authentication is passed, the training module sends public parameters and model training tasks to the vehicle, receives the return data extracted after the execution of the model training task returned by the vehicle, and updates and adjusts the global model parameters based on the return data.
[0033] Preferably, the allocation rules of the federated chain identification code are as follows:
[0034] Construct a distributed topology structure corresponding to the federated learning task;
[0035] According to the task number of the federated learning task and the device number of each data transfer device in the distributed topology structure, they are numbered in order by level;
[0036] Determine the number corresponding to the position of the vehicle in the distributed topology structure as the federation chain identification code.
[0037] Preferably, the authentication rules of the authentication module are as follows:
[0038] When the number of federated chain identification codes is greater than or equal to a preset number threshold, extract the federated chain identification code corresponding to the nearest number threshold and construct it as authentication data;
[0039] Matching the authentication data with verification data in a pre-configured authentication repository;
[0040] When there is a matching item, the authentication is passed;
[0041] When the number is less than the preset number threshold, the obtained federated chain identification codes are arranged in order to form a checksum data string;
[0042] Match the data string to be verified with the standard data string in the corresponding pre-authentication library;
[0043] When there is a matching item, authentication is performed in a manner that assists authentication of the instruction data of the authentication interface.
[0044] Preferably, the Internet of Vehicles identity authentication and privacy protection system based on federated learning further includes: a data analysis and completion module;
[0045] The data analysis and completion module performs the following operations before sending the model training task to the vehicle:
[0046] Send data classification and statistical rules to vehicles;
[0047] Receive the data classification and statistical results corresponding to the data classification and statistical rules returned by the vehicle;
[0048] Based on the data classification and statistical results, determine whether to send standard data in a pre-configured standard database to the vehicle;
[0049] When it is determined that the data needs to be sent, the standard data to be sent is extracted from the standard database based on the statistical results of the data classification and sent.
[0050] Preferably, the Internet of Vehicles identity authentication and privacy protection system based on federated learning further includes: a verification module;
[0051] The verification module performs the following operations:
[0052] Update the data classification statistics results based on the standard data sent;
[0053] Using the updated data classification statistical results, retrieve the corresponding verification data from the pre-configured verification library;
[0054] The returned data is verified based on the verification data, and the returned data that fails the verification is discarded.
[0055] The present invention has the following advantages: a federated learning method is used to train a global model, and each participating vehicle only needs to perform training iterations locally and then return features such as trained model parameters. The platform aggregates the returned data to obtain new parameters of the global model. During this process, the original data of the vehicle is not transmitted, which effectively ensures the privacy and security of the vehicle's local data. In addition, the assigned federal chain identification code is used during vehicle identity authentication, which can effectively hide the vehicle's information and further ensure the confidentiality of privacy.
[0056] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the accompanying drawings.
[0057] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS
[0058] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:
[0059] Figure 1 A schematic diagram of a method for identity authentication and privacy protection of Internet of Vehicles based on federated learning in an embodiment of the present invention;
[0060] Figure 2Schematic diagram of a vehicle network identity authentication and privacy protection system based on federated learning in an embodiment of the present invention. DETAILED DESCRIPTION
[0061] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.
[0062] The embodiment of the present invention provides a method for identity authentication and privacy protection of Internet of Vehicles based on federated learning, such as Figure 1 As shown, including:
[0063] Step 1: Obtain the federated chain identification code assigned to the historical federated learning task sent by the vehicle;
[0064] When it is necessary to authenticate the identity of a vehicle connected to the Internet of Vehicles, obtain the historically assigned federated chain identification code sent by the vehicle; the federated chain identification code is proof that the vehicle has participated in federated learning and can be directly used as the basis for identity authentication; each time the platform performs federated learning to update the parameters of the global model, it will assign a federated chain identification code to the participating vehicles. The allocation rules of the federated chain identification code are as follows: construct a distributed topology structure corresponding to the federated learning task; number the task number of the federated learning task and the device number of each data transfer device in the distributed topology structure in order of hierarchy; determine the number corresponding to the position of the vehicle in the distributed topology structure as the federated chain identification code;
[0065] Federated learning is a machine learning framework. Participating vehicles only need to submit encrypted model parameters, weights, gradients and other features to the platform. There is no need to transmit the original data stored locally. It can meet the training of the global model of the platform system and effectively ensure the privacy of the vehicle's own data.
[0066] The distributed topology result is a tree-like hierarchical structure, with the first layer being the platform node (server), the second layer being the data transmission node (roadside unit, repeater, etc.), and the third layer being the parameter learner (each vehicle); for example, the vehicle's federated chain identification code includes: a field representing the task number of this federated learning task, a field representing the device number of the platform node, a field representing the data transmission node to which it is connected, and a field representing the serial number of the device connected to the data transmission node, etc.;
[0067] Step 2: Authenticate the vehicle’s identity based on the federated chain identification code;
[0068] Each time a federated learning task is performed, the federated chain identification code is reallocated, so the number of federated chain identification codes of the vehicle will gradually increase. Therefore, the effectiveness and efficiency of identification can be taken into account by controlling the number of federated chain identification codes required for identity authentication. The authentication rules are as follows: When the number of federated chain identification codes is greater than or equal to a preset number threshold (any value from 3 to 10), the federal chain identification code corresponding to the most recent number threshold is extracted to construct the authentication data; the authentication data is matched with the verification data in the pre-configured authentication library; when there is a matching item, the authentication is passed; when the number is less than the preset number threshold, the obtained federal chain identification codes are arranged in sequence to form a data string with verification; the data string to be verified is matched with the corresponding standard data string in the pre-authentication library; when there is a matching item, the authentication is performed in a manner to assist the authentication of the instruction data of the authentication interface. That is, first determine the number of obtained federated chain identification codes. When the number exceeds the threshold, only the threshold number of recently assigned federated chain identification codes need to be extracted to match the verification data in the authentication library to achieve authentication. When the authentication is passed, the vehicle is allowed to participate in the federated learning task. The verification data in the authentication library will be updated after each federated learning, that is, the authentication data (or verification data) corresponding to each vehicle will be updated in combination with the assigned federated chain identification code. For vehicles whose number does not meet the requirements, the pre-authentication library can be used for authentication, but it must be combined with the authentication of the instruction data in the assisting authentication interface. The instruction data includes: pre-configured passwords, gestures and other identification information. By separating the authentication library and the pre-authentication library, the matching operation time can be reduced and the efficiency can be improved.
[0069] Step 3: After authentication is passed, send public parameters and model training tasks to the vehicle;
[0070] Passing the authentication means that the identity is legitimate and is allowed to participate in this federated learning task. After the platform server configures the public parameters of this federated school, it sends them to each vehicle to start this machine learning; the public parameters include global model parameters, global iteration rounds, system public key, etc.
[0071] Step 4: Receive the return data returned by the vehicle after executing the model training task;
[0072] The returned data includes: parameters, weights, gradients and other features of the trained model;
[0073] Step 5: Update and adjust the global model parameters based on the returned data.
[0074] Process the returned data through aggregation operations to obtain new global model parameters;
[0075] In one embodiment, the method for identity authentication and privacy protection of Internet of Vehicles based on federated learning further includes:
[0076] Before sending the model training task to the vehicle, send the data classification statistical rules to the vehicle;
[0077] Receive the data classification and statistical results corresponding to the data classification and statistical rules returned by the vehicle;
[0078] Based on the data classification and statistical results, determine whether to send standard data in a pre-configured standard database to the vehicle;
[0079] When it is determined that the data needs to be sent, the standard data to be sent is extracted from the standard database based on the statistical results of the data classification and sent.
[0080] The working principle and beneficial effects of the above technical solution are:
[0081] Since the amount of data stored locally in the vehicle is sometimes small and cannot meet the requirements of model training, it is necessary to classify and count the vehicle data before conducting the training task, and then supplement it with data from the standard database to ensure that the data used by the vehicle to train the model meets the training requirements; the data classification and statistical rules vary according to the global model to be trained, so first send the data classification and statistical rules to the vehicle, the vehicle performs data classification and statistics, and feeds back the data classification and statistical results, and the platform decides whether to send standard data according to the situation; when it needs to be sent, extract the standard data from the pre-configured standard database and send it; the data classification and statistical rules include: the total number of statistical data and / or the number of each type of data, etc.; the distinction between each type of data The data can be distinguished by the different keywords corresponding to the type of data contained in the data; wherein, based on the data classification statistical results, it is determined whether to send the standard data in the pre-configured standard database to the vehicle; and when it is determined that it needs to be sent, the standard data to be sent is extracted from the standard database based on the data classification statistical results and sent, specifically: determine whether the statistical values corresponding to the various statistical items in the data classification statistical results are greater than the corresponding preset thresholds, when they are all greater, there is no need to send; otherwise, it is necessary to send; when it is necessary to send, arrange the statistical values of the various statistical items in order to form a call parameter set; use the call parameter set as an index to retrieve the corresponding standard data from the standard database and send it; the call parameter set in the standard database is associated with the standard data in a one-to-one correspondence.
[0082] In one embodiment, the method for identity authentication and privacy protection of Internet of Vehicles based on federated learning further includes:
[0083] Update the data classification statistical results based on the sent standard data; that is, classify the standard data, and then add the value of the standard data after statistical classification to the corresponding statistical classification item; however, the data classification statistical results need to be updated only when the standard data needs to be supplemented, and if there is no supplement, the original data classification statistical results can be used directly;
[0084] Using the updated data classification statistical results, corresponding verification data is retrieved from the pre-configured verification library; firstly, the statistical values of each statistical classification item in the data classification statistical results are sorted in order to form a retrieval parameter set, and then the verification data corresponding to the retrieval parameter set is retrieved from the verification library; the verification data is essentially the allowable range corresponding to each data item of the returned data;
[0085] The returned data is verified based on the verification data, and the returned data that fails the verification is eliminated. That is, when all data items in the returned data are within the allowable range guided by the verification data, the verification is passed, otherwise it fails; through verification, the influence of the returned data of malicious vehicles on the training of the global model is avoided.
[0086] Since the vehicle is dynamic, that is, the corresponding transfer data transmission device of the vehicle may be converted, and the topology of the federated chain will change at this time, which is not conducive to recording the federated chain identification code for identity authentication. In order to solve the above technical problems, the vehicle network identity authentication and privacy protection method based on federated learning also includes: obtaining the first position of the vehicle that has passed the identity authentication and the second position of the transfer transmission device to which it is connected; based on the first position and the second position, determining the first time when the vehicle is disconnected from the transfer transmission device; obtaining the parameters of the corresponding model training of the vehicle (bandwidth, available computing power, etc.), and determining the second time required for model training based on the parameters; when the second time is greater than the first time, deleting the vehicle from the federated chain, and the vehicle does not participate in this training task;
[0087] The steps for determining the first time are as follows: obtaining a map of the coverage radius area of the transit transmission device with the second position as the center; taking the intersection point of the edge of the coverage area and the road on the map as the end point and the first position of the vehicle as the starting point, performing vehicle path planning; determining the shortest and longest values of the vehicle's travel time on each planned path as the upper and lower limits of the first time;
[0088] The steps for determining the second time are as follows: arrange the parameters in order to form a time analysis set; match the time analysis set with the standard set corresponding to each analysis time interval in the pre-configured analysis library corresponding to the global model, and use the analysis time interval corresponding to the standard set matched with the time analysis set as the second time; the analysis library is constructed in advance based on the parameters of each vehicle and the statistical completion time interval during the historical training process of the global model; among them, the available computing power is obtained by multiplying the maximum computing power of the vehicle by the operating condition coefficient; the operating condition coefficient is obtained by querying the pre-configured coefficient correspondence table by the current operating condition of the vehicle.
[0089] In order to further improve the confidentiality and security of the verification, in one embodiment, the vehicle network identity authentication and privacy protection method based on federated learning further includes:
[0090] After the federated learning is completed, the vehicles participating in the federated learning are grouped and associated in pairs, and the assigned federated chain identification codes are exchanged and marked; the two vehicles are associated with each other, and there is no identical mark, that is, one mark corresponds to a group of associated vehicles;
[0091] When the next federated learning is verified, if two vehicles marked with the same tag are detected at the same time, the two vehicles will send communication connection requests to each other. After the two vehicles are connected, they will exchange verification. After the verification is passed, the federated chain identification code will be exchanged again. If two vehicles with the same tag are not detected (one tag detects only one vehicle, three vehicles or more), the detected vehicles will not participate in this federated learning task; and the vehicles marked with three vehicles or more will be placed in the blacklist;
[0092] Get the federated chain identification code assigned to the historical federated learning task sent by the vehicle;
[0093] Authenticate the vehicle’s identity based on the federated chain identification code;
[0094] When authentication is passed, public parameters and model training tasks are sent to the vehicle;
[0095] Receive the return data returned by the vehicle after executing the model training task;
[0096] The global model parameters are updated and adjusted based on the returned data.
[0097] Among them, the exchange verification includes: random number verification, device number verification, and verification of the previous federated chain identification code of the exchange, one or more combinations thereof; wherein the random number verification is to synchronously generate and store a random number when exchanging and marking the federated chain identification code, and when verifying, the stored random number is sent to each other for verification; the device number verification is to synchronously store the device number of the other party when exchanging and marking the federated chain identification code, and when verifying, the stored device number is sent to each other for verification; the previous federated chain identification code verification of the exchange is to synchronously store the previous federated chain identification code of the other party when exchanging and marking the federated chain identification code, and when verifying, the previous federated chain identification code is sent to each other for verification;
[0098] Since association verification is required, when performing pairwise association, predictive analysis is required to analyze the probability of associated vehicles appearing at the same time, and group association is performed on this basis. The specific association analysis steps are as follows:
[0099] Get all the task numbers corresponding to the historical learning of the vehicles participating in the federated learning;
[0100] Construct the learning task axis of federated learning. The learning task axis is connected by line segments representing each learning task. That is, each line segment of the learning task axis corresponds to the task number of the historical learning in sequence.
[0101] Sort the vehicles according to the proportion of all the task numbers of each vehicle on the line segment of the learning task axis;
[0102] Extract the first vehicle in the ranking, and calculate its participation behavior correlation with other vehicles according to the following formula. The correlation calculation formula is as follows:
[0103]
[0104] In the formula, G ij represents the participation behavior correlation between the currently extracted vehicle and the vehicle at the jth position in the sorting, that is, the participation behavior correlation between the vehicle at the ith position in the sorting and the vehicle at the jth position in the sorting; is the number of line segments in the union of the sets of line segments on the learning task axis for the i-th vehicle and the j-th vehicle respectively; is the number of line segments in the intersection of the set of line segments of the i-th vehicle and the j-th vehicle on the learning task axis; β1 and β2 are the first weight coefficient and the second weight coefficient pre-configured respectively;
[0105] The vehicle computing performance data (upload rate, download rate, single-cycle floating-point computing capability, peak floating-point computing capability, etc.) and online time of both vehicles corresponding to the participating behavior correlation are obtained, and the first secondary correlation and the second secondary correlation are calculated respectively. The calculation formula is as follows:
[0106]
[0107] In the formula, S ij is the first secondary correlation between the vehicle at the i-th position in the ranking and the vehicle at the j-th position in the ranking; C ik is the k parameters in the calculated performance data of the i-th vehicle in the ranking; C jk is the kth parameter in the calculated performance data of the jth vehicle in the ranking; γ k is the preset third weight coefficient corresponding to the kth parameter; K is the total number of parameters in the calculation performance data;
[0108] In the formula, Z ij represents the second secondary association between the i-th vehicle in the ranking and the j-th vehicle in the ranking; Indicates the total time that either the i-th vehicle or the j-th vehicle is online; Indicates the total time that the i-th vehicle and the j-th vehicle are online at the same time; β1 and β2 are the third weight coefficient and the fourth weight coefficient pre-configured respectively;
[0109] The vehicle having the largest sum of the participation behavior correlation degree, the first secondary correlation degree and the second secondary correlation degree is associated with the extracted vehicles.
[0110] The reason for considering computing performance in the analysis of association is that when the next federated learning has computing performance requirements, one of them cannot be associated with the other after being screened out; that is, when the platform gives the task requirements and the vehicle itself judges that it cannot meet them, it will not enter the verification stage when it actively does not participate. At this time, it is screened out and cannot be associated with the other associated vehicle for verification. The analysis of online status and participation behavior mainly considers the objective participation and subjective participation. The objective participation is based on the need for the vehicle to be online on the platform, and the subjective performance is the historical participation. Comprehensive analysis of the vehicle's participation behavior, computing performance, and online status ensures the probability that the associated vehicles can participate in the next federated task at the same time, so as to ensure the smooth progress of the next federated learning.
[0111] The present invention also provides a vehicle network identity authentication and privacy protection system based on federated learning, such as Figure 2 As shown, it includes: an acquisition module 1, an authentication module 2 and a training module 3;
[0112] Among them, the acquisition module 1 obtains the federal chain identification code assigned when the vehicle participates in the historical federated learning task; the authentication module 2 authenticates the identity of the vehicle based on the federal chain identification code; when the authentication is passed, the training module 3 sends public parameters and model training tasks to the vehicle, receives the return data extracted after the execution of the model training task returned by the vehicle, and updates and adjusts the global model parameters based on the return data.
[0113] The allocation rules of the federated chain identification code are as follows:
[0114] Construct a distributed topology structure corresponding to the federated learning task;
[0115] According to the task number of the federated learning task and the device number of each data transfer device in the distributed topology structure, they are numbered in order by level;
[0116] Determine the number corresponding to the position of the vehicle in the distributed topology structure as the federation chain identification code.
[0117] Among them, the authentication rules of the authentication module are as follows:
[0118] When the number of federated chain identification codes is greater than or equal to a preset number threshold, extract the federated chain identification code corresponding to the nearest number threshold and construct it as authentication data;
[0119] Matching the authentication data with verification data in a pre-configured authentication repository;
[0120] When there is a matching item, the authentication is passed;
[0121] When the number is less than the preset number threshold, the obtained federated chain identification codes are arranged in order to form a checksum data string;
[0122] Match the data string to be verified with the standard data string in the corresponding pre-authentication library;
[0123] When there is a matching item, authentication is performed in a manner that assists authentication of the instruction data of the authentication interface.
[0124] In one embodiment, the Internet of Vehicles identity authentication and privacy protection system based on federated learning further includes: a data analysis and completion module;
[0125] The data analysis and completion module performs the following operations before sending the model training task to the vehicle:
[0126] Send data classification and statistical rules to vehicles;
[0127] Receive the data classification and statistical results corresponding to the data classification and statistical rules returned by the vehicle;
[0128] Based on the data classification and statistical results, determine whether to send standard data in a pre-configured standard database to the vehicle;
[0129] When it is determined that the data needs to be sent, the standard data to be sent is extracted from the standard database based on the statistical results of the data classification and sent.
[0130] In one embodiment, the Internet of Vehicles identity authentication and privacy protection system based on federated learning further includes: a verification module;
[0131] The verification module performs the following operations:
[0132] Update the data classification statistics results based on the standard data sent;
[0133] Using the updated data classification statistical results, retrieve the corresponding verification data from the pre-configured verification library;
[0134] The returned data is verified based on the verification data, and the returned data that fails the verification is discarded.
[0135] In one embodiment, after authentication is passed and before sending public parameters and model training tasks to the vehicle, the training module further performs the following operations:
[0136] Obtaining a first position of a vehicle that has passed identity authentication and a second position of a transfer transmission device to which the vehicle is connected;
[0137] Determining a first time when the vehicle is disconnected from the transit transmission device based on the first location and the second location;
[0138] Obtain the parameters of the corresponding model training of the vehicle, and determine the second time required for model training based on the parameters; when the second time is greater than the first time, delete the vehicle from the federation chain, and the vehicle does not participate in this training task;
[0139] The steps for determining the first time are as follows: obtaining a map of the coverage radius area of the transit transmission device with the second position as the center; taking the intersection point of the edge of the coverage area and the road on the map as the end point and the first position of the vehicle as the starting point, performing vehicle path planning; determining the shortest and longest values of the vehicle's travel time on each planned path, which are the lower limit and upper limit of the first time respectively;
[0140] The steps for determining the second time are as follows: arrange the parameters in order to form a time analysis set; match the time analysis set with the standard set corresponding to each analysis time interval in the pre-configured analysis library corresponding to the global model, and use the analysis time interval corresponding to the standard set matched with the time analysis set as the second time; the analysis library is constructed in advance based on the parameters of each vehicle and the statistical completion time interval during the historical training process of the global model; among them, the available computing power is obtained by multiplying the maximum computing power of the vehicle by the operating condition coefficient; the operating condition coefficient is obtained by querying the pre-configured coefficient correspondence table by the current operating condition of the vehicle.
[0141] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include these modifications and variations.
Claims
1. A method for identity authentication and privacy protection in Internet of Vehicles based on federated learning, characterized in that: include: Get the federated chain identification code assigned to the historical federated learning task sent by the vehicle; Authenticate the vehicle’s identity based on the federated chain identification code; When authentication is passed, public parameters and model training tasks are sent to the vehicle; Receive the return data returned by the vehicle after executing the model training task; Update and adjust global model parameters based on returned data; After authentication, before sending public parameters and model training tasks to the vehicle, it also includes: Obtaining a first position of a vehicle that has passed identity authentication and a second position of a transfer transmission device to which the vehicle is connected; Determining a first time when the vehicle is disconnected from the transit transmission device based on the first location and the second location; Obtain the parameters of the corresponding model training of the vehicle, and determine the second time required for model training based on the parameters; when the second time is greater than the first time, delete the vehicle from the federation chain, and the vehicle does not participate in this training task; The steps for determining the first time are as follows: obtaining a map of the coverage radius area of the transit transmission device with the second position as the center; taking the intersection point of the edge of the coverage area and the road on the map as the end point and the first position of the vehicle as the starting point, performing vehicle path planning; determining the shortest and longest values of the vehicle's travel time on each planned path, which are the lower limit and upper limit of the first time respectively; The steps for determining the second time are as follows: arrange the parameters in order to form a time analysis set; match the time analysis set with the standard set corresponding to each analysis time interval in the pre-configured analysis library corresponding to the global model, and use the analysis time interval corresponding to the standard set matched with the time analysis set as the second time; the analysis library is constructed in advance based on the parameters of each vehicle and the statistical completion time interval during the historical training process of the global model; among them, the available computing power is obtained by multiplying the maximum computing power of the vehicle by the operating condition coefficient; the operating condition coefficient is obtained by querying the pre-configured coefficient correspondence table by the current operating condition of the vehicle.
2. The method for identity authentication and privacy protection of Internet of Vehicles based on federated learning as claimed in claim 1, characterized in that: The allocation rules of the federated chain identification code are as follows: Construct a distributed topology structure corresponding to the federated learning task; According to the task number of the federated learning task and the device number of each data transfer device in the distributed topology structure, they are numbered in order by level; Determine the number corresponding to the position of the vehicle in the distributed topology structure as the federation chain identification code.
3. The method for identity authentication and privacy protection of Internet of Vehicles based on federated learning as claimed in claim 1, characterized in that: The authentication rules are as follows: When the number of federated chain identification codes is greater than or equal to a preset number threshold, extract the federated chain identification code corresponding to the nearest number threshold and construct it as authentication data; Matching the authentication data with verification data in a pre-configured authentication repository; When there is a matching item, the authentication is passed; When the number is less than the preset number threshold, the obtained federated chain identification codes are arranged in order to form a checksum data string; Match the data string to be verified with the standard data string in the corresponding pre-authentication library; When there is a matching item, authentication is performed in a manner that assists authentication of the instruction data of the authentication interface.
4. The method for identity authentication and privacy protection of Internet of Vehicles based on federated learning as claimed in claim 1, characterized in that: Also includes: Before sending the model training task to the vehicle, send the data classification statistical rules to the vehicle; Receive the data classification and statistical results corresponding to the data classification and statistical rules returned by the vehicle; Based on the data classification and statistical results, determine whether to send standard data in a pre-configured standard database to the vehicle; When it is determined that the data needs to be sent, the standard data to be sent is extracted from the standard database based on the statistical results of the data classification and sent.
5. The method for identity authentication and privacy protection of Internet of Vehicles based on federated learning as claimed in claim 4, characterized in that: Also includes: Update the data classification statistics results based on the standard data sent; Based on the updated data classification statistics, the corresponding verification data is retrieved from the pre-configured verification library; The returned data is verified based on the verification data, and the returned data that fails the verification is discarded.
6. A vehicle network identity authentication and privacy protection system based on federated learning, characterized in that: include: acquisition module, certification module and training module; Among them, the acquisition module obtains the federated chain identification code assigned when the vehicle participates in the historical federated learning task sent by the vehicle; the authentication module authenticates the identity of the vehicle based on the federated chain identification code; when the authentication is passed, the training module sends public parameters and model training tasks to the vehicle, receives the return data extracted after the execution of the model training task returned by the vehicle, and updates and adjusts the global model parameters based on the return data; After authentication, before sending public parameters and model training tasks to the vehicle, the training module also performs the following operations: Obtaining a first position of a vehicle that has passed identity authentication and a second position of a transfer transmission device to which the vehicle is connected; Determining a first time when the vehicle is disconnected from the transit transmission device based on the first location and the second location; Obtain the parameters of the corresponding model training of the vehicle, and determine the second time required for model training based on the parameters; when the second time is greater than the first time, delete the vehicle from the federation chain, and the vehicle does not participate in this training task; The steps for determining the first time are as follows: obtaining a map of the coverage radius area of the transit transmission device with the second position as the center; taking the intersection point of the edge of the coverage area and the road on the map as the end point and the first position of the vehicle as the starting point, performing vehicle path planning; determining the shortest and longest values of the vehicle's travel time on each planned path, which are the lower limit and upper limit of the first time respectively; The steps for determining the second time are as follows: arrange the parameters in order to form a time analysis set; match the time analysis set with the standard set corresponding to each analysis time interval in the pre-configured analysis library corresponding to the global model, and use the analysis time interval corresponding to the standard set matched with the time analysis set as the second time; the analysis library is constructed in advance based on the parameters of each vehicle and the statistical completion time interval during the historical training process of the global model; among them, the available computing power is obtained by multiplying the maximum computing power of the vehicle by the operating condition coefficient; the operating condition coefficient is obtained by querying the pre-configured coefficient correspondence table by the current operating condition of the vehicle.
7. The vehicle network identity authentication and privacy protection system based on federated learning as claimed in claim 6, characterized in that: The allocation rules of the federated chain identification code are as follows: Construct a distributed topology structure corresponding to the federated learning task; According to the task number of the federated learning task and the device number of each data transfer device in the distributed topology structure, they are numbered in order by level; Determine the number corresponding to the position of the vehicle in the distributed topology structure as the federation chain identification code.
8. The vehicle network identity authentication and privacy protection system based on federated learning as claimed in claim 6, characterized in that: The authentication rules of the authentication module are as follows: When the number of federated chain identification codes is greater than or equal to a preset number threshold, extract the federated chain identification code corresponding to the nearest number threshold and construct it as authentication data; Matching the authentication data with verification data in a pre-configured authentication repository; When there is a matching item, the authentication is passed; When the number is less than the preset number threshold, the obtained federated chain identification codes are arranged in order to form a checksum data string; Match the data string to be verified with the standard data string in the corresponding pre-authentication library; When there is a matching item, authentication is performed in a manner that assists authentication of the instruction data of the authentication interface.
9. The Internet of Vehicles identity authentication and privacy protection system based on federated learning as claimed in claim 6, characterized in that: Also includes: Data analysis and completion module; The data analysis and completion module performs the following operations before sending the model training task to the vehicle: Send data classification and statistical rules to vehicles; Receive the data classification and statistical results corresponding to the data classification and statistical rules returned by the vehicle; Based on the data classification and statistical results, determine whether to send standard data in a pre-configured standard database to the vehicle; When it is determined that the data needs to be sent, the standard data to be sent is extracted from the standard database based on the statistical results of the data classification and sent.
10. The vehicle network identity authentication and privacy protection system based on federated learning as claimed in claim 9, characterized in that: Also includes: Verification module; The verification module performs the following operations: Update the data classification statistics results based on the standard data sent; Based on the updated data classification statistics, the corresponding verification data is retrieved from the pre-configured verification library; The returned data is verified based on the verification data, and the returned data that fails the verification is discarded.