Method and system for secure sharing of Internet of Vehicles data based on blockchain and dynamic reputation

Abstract

The present invention discloses a method and system for secure sharing of Internet of Vehicles (IoV) data based on blockchain and dynamic reputation, relating to the technical fields of data security and privacy protection. Training tasks are published via blockchain, and model parameters and the reputation values ​​of participating nodes are stored in a decentralized manner. The legitimacy of vehicle nodes and RSU nodes is verified. Secret sharing and the Multi-Krum algorithm are combined to enable RSUs to filter and aggregate model parameters uploaded by vehicle nodes. During the reputation evaluation process, a subjective logic model is improved, weights are set for different data sets, and Bayesian theory is used to eliminate uncertain interaction events that occur in vehicle nodes during training. The reputation of RSU nodes and vehicle nodes is evaluated based on the historical interaction behaviors of participating nodes, achieving secure on-chain storage of global model parameters and the reputation values ​​of participating nodes. This method enables secure data sharing applications in IoV scenarios while protecting local privacy and the security of training models.

Description

Technical Field

[0001] The present invention relates to the technical field of data security and privacy protection, and in particular to a method and system for securely sharing Internet of Vehicles data based on blockchain and dynamic reputation. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] The Internet of Vehicles (IoV) is a crucial infrastructure for transitioning from single-vehicle intelligence to the connected, autonomous driving era. Federated learning's model sharing mechanism enables more secure and efficient information exchange within the IoV. Each participating vehicle node can store training data locally without sharing it, sharing only trained model parameters with the RSU node. However, several challenges remain in IoV scenarios:

[0004] (1) Since RSUs are usually semi-trusted or untrusted entities, they can infer part or even all of the vehicle's original data based on the vehicle's local model parameters, exposing the vehicle's privacy information to the risk of leakage;

[0005] (2) Vehicle nodes may launch poisoning attacks or use low-quality datasets for model training during local training, thereby damaging the global model;

[0006] (3) During the sharing process, there is no pre-established trust relationship between participating nodes, which may lead to security risks during collaborative training.

[0007] To address these challenges, some schemes combine secret sharing technology to ensure that the server can only decrypt parameters after at least k participating nodes upload them, thereby preventing inference attacks from malicious nodes. However, such schemes lack verification of the legitimacy of participating nodes, and there is a risk of external adversaries impersonating legitimate nodes.

[0008] In terms of resisting poisoning, some works have proposed using clean datasets collected in advance to verify the model parameters uploaded by users, thereby removing abnormal updates; however, this centralized server-based solution still has the risk of single point failure.

[0009] In order to enable participating nodes to trust each other and obtain high-quality model parameters, some researchers have introduced subjective logic models to conduct safe and efficient reputation evaluation of vehicle nodes. However, such schemes lack the evaluation of the reputation of RSU nodes. RSU plays an important role in the model training process, responsible for verifying the data sharing size of vehicle nodes, assisting in aggregation, and decrypting the global model. This is related to the accuracy of the training model and the security and efficiency of collaborative learning, and even the normal operation of the training process. At the same time, there are cases where vehicle nodes may experience uncertain interaction events during the training process, which will seriously affect the effectiveness and accuracy of the reputation evaluation of participating nodes. Summary of the Invention

[0010] To address the above problems, the present invention proposes a method and system for secure sharing of Internet of Vehicles (IoV) data based on blockchain and dynamic reputation, which avoids the risk of external adversaries impersonating legitimate nodes, solves the single point failure problem of centralized servers, improves the effectiveness and accuracy of reputation evaluation, and realizes data security sharing applications in IoV scenarios while protecting local privacy and training model security.

[0011] In order to achieve the above object, the present invention adopts the following technical solutions:

[0012] In a first aspect, the present invention provides a method for securely sharing Internet of Vehicles data based on blockchain and dynamic reputation, comprising:

[0013] Publish model training tasks and select RSU nodes and vehicle nodes that participate in the model training tasks based on their reputation values;

[0014] The selected RSU node downloads the global model from the blockchain and sends it to the vehicle nodes that have joined the model training task within their respective coverage areas;

[0015] The vehicle node iteratively trains the local model based on the global model and local data, encrypts the generated local update parameters using a secret sharing algorithm to obtain a secret share, and then signs the secret share and uploads it to the RSU node.

[0016] The RSU node calculates the Euclidean distance between the secret shares, signs them, and uploads them to the blockchain;

[0017] The blockchain reconstructs the Euclidean distance through Lagrangian interpolation, selects legal vehicle nodes, aggregates the secret shares uploaded by the legal vehicle nodes at the RSU nodes, and reconstructs the aggregation results through Lagrangian interpolation to obtain the global model of aggregation update.

[0018] The RSU node downloads the aggregated updated global model from the blockchain and sends it to the corresponding vehicle node so that the vehicle node can conduct the next round of model training until the model converges. During each round of model training, the reputation values ​​of the RSU node and the vehicle node will be stored on the chain after reputation evaluation.

[0019] As an optional implementation, the method for securely sharing data in the Internet of Vehicles also includes verifying the legitimacy of the vehicle node and the RSU node; specifically, including:

[0020] After receiving the secret share uploaded by the vehicle node, the RSU node verifies whether the identity and timestamp of the vehicle node are valid;

[0021] After the blockchain receives the Euclidean distance and secret share uploaded by the RSU node, it verifies whether the identity and timestamp of the RSU node are valid.

[0022] As an optional implementation method, the reputation evaluation of RSU includes: RSU node verifies the size of local update parameters uploaded by vehicle node, RSU node verifies the update of global model parameters, and RSU node verifies the secret share of other RSU nodes; if the above events are valid and reliable after verification, they will be regarded as positive interaction events.

[0023] As an optional implementation, a weight η is set for the number of positive interaction events, where 0<η≤1, and the weight of the number of negative interaction events is set to 1. Then the reputation value of the RSU node is:

[0024]

[0025] in, They represent the degree of trust, distrust and uncertainty of task publisher i towards RSU node j respectively.

[0026] As an optional implementation method, the trust level, distrust level, and uncertainty level of task publisher i to RSU node j are:

[0027]

[0028] Where r and s are the number of positive interaction events and the number of negative interaction events, respectively; c is a constant.

[0029] As an optional implementation, the reputation value of the vehicle node is:

[0030]

[0031] in, is the uncertainty degree of task publisher i about vehicle node j, s and f are the number of positive interaction events and the number of negative interaction events, respectively.

[0032] As an optional implementation method, when evaluating the reputation of a vehicle node, Bayesian theory is used to predict the probability of a positive interaction event occurring in the vehicle node, so as to eliminate uncertain interaction events occurring in the vehicle node.

[0033] In a second aspect, the present invention provides a vehicle network data security sharing system based on blockchain and dynamic reputation, including: a task publisher, a blockchain, an RSU node and a vehicle node;

[0034] The task publisher is used to publish model training tasks and select RSU nodes and vehicle nodes that participate in the model training tasks based on their reputation values;

[0035] The selected RSU node downloads the global model from the blockchain and sends it to the vehicle nodes that have joined the model training task within their respective coverage areas;

[0036] The vehicle node iteratively trains the local model based on the global model and local data, encrypts the generated local update parameters using a secret sharing algorithm to obtain a secret share, signs the secret share, and uploads it to the RSU node.

[0037] The RSU node calculates the Euclidean distance between secret shares, signs them, and uploads them to the blockchain;

[0038] The blockchain reconstructs the Euclidean distance through Lagrangian interpolation, selects legal vehicle nodes, aggregates the secret shares uploaded by the legal vehicle nodes at the RSU nodes, and reconstructs the aggregation results through Lagrangian interpolation to obtain a global model of aggregation updates.

[0039] The RSU node downloads the aggregated updated global model from the blockchain and sends it to the corresponding vehicle node so that the vehicle node can conduct the next round of model training until the model converges. During each round of model training, the reputation values ​​of the RSU node and the vehicle node will be stored on the chain after reputation evaluation.

[0040] In a third aspect, the present invention provides an electronic device comprising a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein the computer instructions, when executed by the processor, perform the method described in the first aspect.

[0041] In a fourth aspect, the present invention provides a computer-readable storage medium for storing computer instructions, wherein when the computer instructions are executed by a processor, the method described in the first aspect is performed.

[0042] Compared with the prior art, the present invention has the following beneficial effects:

[0043] Compared with existing technologies, the present invention protects the model parameters of vehicle nodes during the entire federated learning training process by using secret sharing and the Multi-Krum algorithm, effectively avoiding inference attacks and collusion attacks by untrusted RSU nodes; the model parameters are screened before aggregation, and the Euclidean distance between secret shares is calculated to effectively remove poisonous updates, thereby reducing the communication and computing pressure on vehicle terminals, ensuring effective defense against poisoning attacks launched by malicious vehicle nodes without leaking privacy, and ultimately achieving safe and efficient Internet of Vehicles data sharing.

[0044] Compared with the existing technology that does not use blockchain, the present invention uses blockchain instead of the central server to screen and aggregate model parameters, avoiding the problem of untrusted central servers returning erroneous results and single point failure. At the same time, when the vehicle nodes and RSU nodes upload model parameters, the legitimacy of the identities of the vehicle nodes and RSU nodes are verified to prevent external adversaries from impersonating legitimate users to upload false data, realizing decentralized and secure federated learning, and ultimately ensuring the secure on-chain storage of global models and reputation values.

[0045] Compared with the existing technology using blockchain, during each training process, the task publisher of the present invention conducts effective and accurate reputation evaluation of RSU and vehicle nodes based on the model quality, and stores the reputation value securely on the chain; during the reputation evaluation process, based on the subjective logic model, the weight is set in combination with the size of the data set to increase the reputation evaluation of the RSU node, and the Bayesian theory is used to eliminate uncertain interaction events that may occur in the vehicle nodes during the training process, thereby achieving effective and accurate reputation evaluation of the RSU nodes and vehicle nodes, and ultimately realizing traceable, verifiable and privacy-protected vehicle federated learning.

[0046] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0048] Figure 1 A model diagram of the secure sharing method for Internet of Vehicles data based on blockchain and dynamic reputation provided in Example 1 of the present invention;

[0049] Figure 2 This is a flowchart of privacy protection training and poisoning attack detection provided by Example 1 of the present invention. DETAILED DESCRIPTION

[0050] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0051] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0052] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0053] In the absence of conflict, the embodiments of the present invention and the features thereof may be combined with each other.

[0054] Example 1

[0055] This embodiment provides a method for securely sharing Internet of Vehicles data based on blockchain and dynamic reputation, including:

[0056] Publish model training tasks and select RSU nodes and vehicle nodes that participate in the model training tasks based on their reputation values;

[0057] The selected RSU node downloads the global model parameters from the blockchain and sends them to the vehicle nodes that join the model training task within their respective coverage areas;

[0058] The vehicle node iteratively trains the local model based on the global model parameters and local data, encrypts the generated local update parameters using a secret sharing algorithm to obtain a secret share, and then signs the secret share and uploads it to the RSU node.

[0059] The RSU node calculates the Euclidean distance between the secret shares, signs them, and uploads them to the blockchain;

[0060] The blockchain reconstructs the Euclidean distance through Lagrangian interpolation, selects legal vehicle nodes, aggregates the secret shares uploaded by the legal vehicle nodes at the RSU nodes, and reconstructs the aggregation results through Lagrangian interpolation to obtain the global model of aggregation update.

[0061] The RSU node downloads the aggregated updated global model parameters from the blockchain and sends them to the corresponding vehicle node so that the vehicle node can conduct the next round of model training until the model converges. During each round of model training, the reputation values ​​of the RSU node and the vehicle node will be stored on the chain after reputation evaluation.

[0062] In this embodiment, the method includes the participation of four parties: a task publisher, a blockchain module, a roadside unit (RSU) module, and a vehicle node module. Specifically:

[0063] (1) Task publisher: Build a machine learning model based on its own needs and publish federated learning training tasks through the blockchain. Interested vehicle nodes can apply to participate in the federated learning training tasks. As more and more vehicle nodes join in and contribute to model training, the task publisher will eventually obtain an ideal machine learning model. At the same time, the task publisher needs to evaluate the local model quality of the participating nodes and generate a reputation value based on the subjective logic model as a measure of the credibility of the participating nodes in the federated learning process, so that other task publishers can select participating nodes with better reputation in the federated learning process.

[0064] (2) Blockchain module: This module stores the reputation evaluation results of the task publisher on the RSU and vehicle nodes, is responsible for registering the vehicle nodes and RSU nodes within the system, and aggregates the global model. Due to the traceability and tamper-proof nature of the blockchain, when a participating node sends a low-quality local model, the relevant information in the data block can serve as persistent and transparent evidence. In addition, the task publisher ensures the secure sharing of the reputation evaluation results through the access control policy pre-defined in the blockchain. The access records of other task publishers to the reputation evaluation results will be stored in the blockchain.

[0065] (3) Roadside Unit (RSU) module: The RSU node is a wireless device fixed on both sides of the road or placed at a specific location on the roadside. It is located near the vehicle node and has certain computing and storage capabilities. It can collect, process and forward data uploaded by vehicle nodes within its coverage area. It is mainly responsible for distributing training tasks, verifying the data sharing size of vehicle nodes, assisting the blockchain in removing toxic gradient updates and aggregating global models, thereby expanding the data communication range of the Internet of Vehicles and ensuring security in the data sharing process.

[0066] During the model download phase, the RSU node downloads the global model from the blockchain and distributes the global model to the vehicle nodes participating in the training within the coverage area; during the model upload phase, the RSU node is responsible for processing the secret share uploaded by the vehicle node and uploading it to the blockchain.

[0067] (4) Vehicle node module: that is, the mobile user end, which collects, stores and pre-processes data, participates in the training process of federated learning, and then uploads the secret share of local model parameters to the nearby RSU node, and then repeats the training process of model parameters and performs a new round of iteration until the accuracy of the global model reaches the ideal expected value; the uploaded content needs to declare the size of the local data and attach the corresponding training time, thereby indicating the size of its own data contribution.

[0068] The following combination Figure 1-Figure 2 The process of this method is introduced in detail.

[0069] 1. Task publishing stage: The task publisher builds a machine learning model based on their own needs and uploads the initialized model to the blockchain.

[0070] 2. System node registration phase: The blockchain completes the registration of all RSU nodes and vehicle nodes in the system, and the information of successfully registered nodes is stored on the blockchain.

[0071] 3. Participating node selection phase: RSU nodes and vehicle nodes that want to join the federated learning training task send an application to the task publisher. The application content includes signing the summary of their respective identity and dataset information with the latest timestamp, and then becoming a candidate node for model training and aggregation; the task publisher downloads the reputation value of the candidate nodes in the recent period from the blockchain and selects the node with the higher reputation value to participate in the task.

[0072] 4. Model distribution phase: The RSU node that successfully joins the training task downloads the global model parameters from the blockchain, and then the RSU distributes the global model parameters of the current round to the vehicle nodes that successfully join the training task within their respective coverage areas.

[0073] 5. Privacy protection training stage:

[0074] 5.1. Local training: The vehicle node updates the local model based on the downloaded global model, then enters the iterative training phase of the model, using local data to train the local model and generate local updates.

[0075] 5.2. Generate secret share: The vehicle node calls the Shmair (k, N) secret sharing algorithm to encrypt the local update parameters to obtain the secret share, signs the secret share with the private key, and uploads it to the RSU node.

[0076] Secret sharing is a cryptographic technique that shares information among multiple parties to ensure it is protected from destruction, tampering, and loss. Secret sharing uses a specific algorithm to split a secret into shares and distribute them to multiple participants. Secret recovery requires the joint efforts of multiple participants according to a protocol; individual secret shares are useless. Key secret sharing algorithms include the Chinese Remainder Theorem, Shmair, and Blakley.

[0077] This embodiment uses the Shmair algorithm, which divides a secret S into N secret shares and distributes them to N participants. To recover the original secret S, at least k participants must collaborate on decryption. If fewer than k participants collaborate, the original secret S cannot be reconstructed. This algorithm is based on Lagrange interpolation and consists of two phases: secret distribution and secret reconstruction.

[0078] 1) The algorithm for the secret distribution phase is as follows:

[0079] For a secret s∈Z p , the distributor starts from Z p Randomly select t-1 coefficients a1, a2, ..., a t-1 Construct a polynomial:

[0080] f(x)=a0+a1x+a2x 2 +…+a t-1 x t-1 (mod p)

[0081] Where s = f(0), calculate y i =f(x i ), x i ∈[1,n],(x i ,y i ) are sent to participants P i , i∈[1,n].

[0082] 2) The algorithm for the secret reconstruction phase is as follows:

[0083] At least t participants reconstruct the original secret s from the secret shares using the Lagrange interpolation formula:

[0084]

[0085] in, If fewer than t parties provide shares, no information about the secret is leaked.

[0086] 6. Poisoning attack detection phase: When the RSU node receives the secret share uploaded by the vehicle node, it first verifies whether the identity and timestamp of the vehicle node are valid. After verification, considering that malicious vehicle nodes may upload poisonous updates, it is necessary to remove poisonous updates as much as possible during global aggregation before aggregation.

[0087] This embodiment uses the Multi-Krum algorithm to remove toxic updates generated by malicious users. Specifically, the RSU node calculates the Euclidean distance between the secret shares uploaded by the received vehicle nodes, and then the RSU node signs the Euclidean distance using its own private key and uploads it to the blockchain.

[0088] The Multi-Krum algorithm is a Byzantine fault-tolerant machine learning algorithm based on Euclidean distance. It is an algorithm that ensures that distributed machine learning can still converge even when there are Byzantine faults. It can remove toxic updates generated by malicious users. The algorithm is as follows:

[0089] The number of users in an area is m. Assuming z is the number of Byzantine nodes, the Euclidean distance between the gradient uploaded by each vehicle node and its nearest mz-2 gradients is added as the quality score of the gradient:

[0090]

[0091] Among them, i→j represents the gradient Ideal gradient The nearest mz-2 gradients.

[0092] Finally, the mz gradients with the lowest quality scores are selected as legal updates and aggregated.

[0093] 7. Model aggregation stage:

[0094] After receiving the Euclidean distance and secret share uploaded by the RSU node, the blockchain first verifies whether the identity and timestamp of the RSU node are valid;

[0095] After verification, the Euclidean distance between the gradient of the i-th user and the gradient of the j-th user is reconstructed through Lagrange interpolation based on the model parameters uploaded by the RSU node, and then the legal vehicle node is selected according to the Multi-Krum algorithm;

[0096] The RSU node downloads the legitimate vehicle node from the blockchain, sums the secret shares uploaded by the legitimate vehicle node locally, and uploads them to the blockchain;

[0097] The blockchain reconstructs the aggregation results through Lagrange interpolation;

[0098] The RSU node downloads the aggregated updated global model parameters from the blockchain and sends them to the vehicle nodes within the coverage area. The vehicle nodes use the aggregated updated global model parameters for the next round of training and repeat the above steps until the model converges or reaches the ideal accuracy.

[0099] 8. Credit Assessment Stage:

[0100] The reputation evaluation model based on subjective logic uses the concepts of evidence space and concept space to describe and measure the trust relationship between one party and another. The evidence space consists of historical interaction events, which are divided into positive events and negative events. The concept space consists of a series of probabilistic reputation evaluation opinions. The subjective reputation opinion of party i on party j is composed of triples expresses and satisfies

[0101]

[0102] in, They represent the degree of trust, distrust and uncertainty of i towards j respectively; r and s are the number of positive interaction events and the number of negative interaction events respectively; c is a constant, and its value is related to the impact of the number of uncertain interaction events on reputation opinion.

[0103] An interaction event is the process by which a vehicle node downloads global model parameters and then uploads them once based on iterative training of local data. A positive interaction event is detected by the RSU node poisoning attack scheme, while a negative interaction event is considered otherwise. When a vehicle node does not upload any parameters, it indicates an uncertain interaction event. Historical interaction events are the total number of interaction events among participating nodes within a time interval Δt: t = {s, f}, where s and f are the number of positive and negative interaction events, respectively.

[0104] 8.1. The credit evaluation of RSU includes three types of events:

[0105] 1) The RSU node verifies the size of the uploaded data set of the participating vehicle nodes; 2) The RSU node verifies the update of the global model parameters; 3) The RSU verifies the secret share of other RSUs.

[0106] If the RSU node considers these events to be valid and reliable after verification, it will be regarded as a positive interaction event. The concept space mainly consists of the reputation opinions of the task publisher on the participating RSU nodes.

[0107] Considering that different RSU nodes may contribute differently to model training, the number of vehicles within the coverage area of ​​different RSU nodes and the size of the collected data set are different, which makes the time for assisting the blockchain to remove toxic models and aggregate global models different, which means that they make different contributions to model training. Therefore, this reputation evaluation scheme sets a weight η for the number of positive interaction events r to reflect the size of the RSU node's contribution; at the same time, in order to reduce the occurrence of negative interaction events, the weight of s is set to 1, 0<η≤1; the original algorithm is rewritten as:

[0108]

[0109] Then the reputation value of task publisher i to RSU node j is expressed as:

[0110]

[0111] 8.2. Reputation evaluation of vehicle nodes;

[0112] Based on the subjective logic model and the concept of historical interaction events, we can obtain:

[0113]

[0114] in, It is the probability of successful parameter transmission and represents the communication quality.

[0115] Considering that vehicle node j may have uncertain interaction events, which will affect the accuracy of the reputation evaluation of the vehicle node, this embodiment uses the Bayesian formula to predict the probability of a positive interaction event occurring at the vehicle node, namely:

[0116]

[0117] Here, the historical interaction event t={s, f} of vehicle node j is taken as the precondition E, and the behavior of affirmative interaction event is taken as event H.

[0118] Assuming that the probability of event H occurring under the condition that event E occurs follows Beta distribution, the correlation coefficient of the impact of the uncertain interaction behavior of vehicle node j on reputation is The mathematical expectation of the Beta distribution is expressed as:

[0119]

[0120] Among them, the correlation coefficient It represents the probability that vehicle node j will show a positive interaction event when an uncertain interaction occurs.

[0121] Combining the above three formulas, in a federated learning task, the direct credit value of task publisher i to vehicle node j is for:

[0122]

[0123] This embodiment proposes a secure data sharing method for the Internet of Vehicles (IoV) based on blockchain and dynamic reputation. Training tasks are published via blockchain, replacing the parameter server in traditional federated learning. Model parameters and the reputation values ​​of participating nodes are stored in a decentralized manner. The legitimacy of vehicle nodes and RSU nodes is verified. Secret sharing of model parameters is supported, and a Multi-Krum algorithm is used to remove toxic model parameters. An RSU layer is added to filter and aggregate model parameters uploaded by vehicle nodes. During the reputation assessment process, a subjective logic model is improved to assign weights to different datasets. Bayesian theory is used to eliminate uncertain interaction events that occur during training. Based on the historical interaction behavior of participating nodes, accurate and effective reputation assessments are performed on both RSU and vehicle nodes. Ultimately, global model parameters and the reputation values ​​of participating nodes are securely stored on-chain. This approach mitigates the risk of external adversaries impersonating legitimate nodes, addresses the single point of failure of centralized servers, and improves the effectiveness and accuracy of reputation assessment. This approach enables secure data sharing applications in IoV scenarios while protecting local privacy and the security of trained models.

[0124] Example 2

[0125] This embodiment provides a secure sharing system for Internet of Vehicles data based on blockchain and dynamic reputation, including: a task publisher, a blockchain, an RSU node, and a vehicle node;

[0126] The task publisher is used to publish model training tasks and select RSU nodes and vehicle nodes that participate in the model training tasks based on their reputation values;

[0127] The selected RSU node downloads the global model from the blockchain and sends it to the vehicle nodes that have joined the model training task within their respective coverage areas;

[0128] The vehicle node iteratively trains the local model based on the global model and local data, encrypts the generated local update parameters using a secret sharing algorithm to obtain a secret share, signs the secret share, and uploads it to the RSU node.

[0129] The RSU node calculates the Euclidean distance between secret shares, signs them, and uploads them to the blockchain;

[0130] The blockchain reconstructs the Euclidean distance through Lagrangian interpolation, selects legal vehicle nodes, aggregates the secret shares uploaded by the legal vehicle nodes at the RSU nodes, and reconstructs the aggregation results through Lagrangian interpolation to obtain a global model of aggregation updates.

[0131] The RSU node downloads the aggregated updated global model from the blockchain and sends it to the corresponding vehicle node so that the vehicle node can conduct the next round of model training until the model converges. During each round of model training, the reputation values ​​of the RSU node and the vehicle node will be stored on the chain after reputation evaluation.

[0132] It should be noted that the above modules correspond to the steps described in Example 1, and the examples and application scenarios implemented by the above modules and the corresponding steps are the same, but are not limited to the contents disclosed in the above Example 1. It should be noted that the above modules, as part of the system, can be executed in a computer system such as a set of computer-executable instructions.

[0133] In further embodiments, there is also provided:

[0134] An electronic device includes a memory and a processor, and computer instructions stored in the memory and executed by the processor, wherein when the computer instructions are executed by the processor, the method described in Example 1 is performed. For the sake of brevity, no further details are given here.

[0135] It should be understood that in this embodiment, the processor may be a central processing unit (CPU), or may be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), off-the-shelf field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0136] The memory may include a read-only memory and a random access memory, and provides instructions and data to the processor. A portion of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0137] A computer-readable storage medium is used to store computer instructions, and when the computer instructions are executed by a processor, the method described in Example 1 is performed.

[0138] The method in Example 1 can be directly implemented as a hardware processor, or can be implemented using a combination of hardware and software modules in the processor. The software module can be located in a storage medium mature in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, it will not be described in detail here.

[0139] Those skilled in the art will appreciate that the units, i.e., algorithm steps, of the various examples described in conjunction with this embodiment can be implemented using electronic hardware or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0140] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without any creative work are still within the scope of protection of the present invention.

Claims

1. A secure sharing method for Internet of Vehicles data based on blockchain and dynamic reputation, characterized by: include: Publish model training tasks and select RSU nodes and vehicle nodes that participate in the model training tasks based on their reputation values; The selected RSU node downloads the global model from the blockchain and sends it to the vehicle nodes that have joined the model training task within their respective coverage areas; The vehicle node iteratively trains the local model based on the global model and local data, encrypts the generated local update parameters using a secret sharing algorithm to obtain a secret share, and then signs the secret share and uploads it to the RSU node. The RSU node calculates the Euclidean distance between the secret shares, signs them, and uploads them to the blockchain; The blockchain reconstructs the Euclidean distance through Lagrangian interpolation, selects legal vehicle nodes, aggregates the secret shares uploaded by the legal vehicle nodes at the RSU nodes, and reconstructs the aggregation results through Lagrangian interpolation to obtain the global model of aggregation update. The RSU node downloads the aggregated updated global model from the blockchain and sends it to the corresponding vehicle node so that the vehicle node can conduct the next round of model training until the model converges. During each round of model training, the reputation values ​​of the RSU node and the vehicle node will be stored on the chain after reputation evaluation.

2. The method for secure sharing of Internet of Vehicles data based on blockchain and dynamic reputation as claimed in claim 1, characterized in that: The secure sharing method for IoV data also includes verifying the legitimacy of vehicle nodes and RSU nodes; specifically, it includes: After receiving the secret share uploaded by the vehicle node, the RSU node verifies whether the identity and timestamp of the vehicle node are valid; After the blockchain receives the Euclidean distance and secret share uploaded by the RSU node, it verifies whether the identity and timestamp of the RSU node are valid.

3. The method for secure sharing of Internet of Vehicles data based on blockchain and dynamic reputation as claimed in claim 1, characterized in that: The reputation evaluation of RSU includes: RSU node verifies the size of the local update parameters uploaded by the vehicle node, RSU node verifies the update of global model parameters, and RSU node verifies the secret share of other RSU nodes; if the above events are all valid and reliable after verification, they will be regarded as positive interaction events.

4. The method for secure sharing of Internet of Vehicles data based on blockchain and dynamic reputation as claimed in claim 3, characterized in that: The weight η is set for the number of positive interaction events, and 0<η≤1, and the weight of the number of negative interaction events is set to 1. The reputation value of the RSU node is: in, They represent the degree of trust, distrust and uncertainty of task publisher i towards RSU node j respectively.

5. The method for secure sharing of Internet of Vehicles data based on blockchain and dynamic reputation as claimed in claim 4, characterized in that: The trust, distrust and uncertainty of task publisher i to RSU node j are: Where r and s are the number of positive interaction events and the number of negative interaction events, respectively; c is a constant.

6. The method for secure sharing of Internet of Vehicles data based on blockchain and dynamic reputation as claimed in claim 1, characterized in that: The reputation value of a vehicle node is: in, is the uncertainty degree of task publisher i about vehicle node j, s and f are the number of positive interaction events and the number of negative interaction events, respectively.

7. The method for secure sharing of Internet of Vehicles data based on blockchain and dynamic reputation as claimed in claim 6, characterized in that: When evaluating the reputation of vehicle nodes, Bayesian theory is used to predict the probability of positive interaction events occurring in vehicle nodes, so as to eliminate uncertain interaction events occurring in vehicle nodes.

8. The Internet of Vehicles data security sharing system based on blockchain and dynamic reputation is characterized by: include: Task publisher, blockchain, RSU node and vehicle node; The task publisher is used to publish model training tasks and select RSU nodes and vehicle nodes that participate in the model training tasks based on their reputation values; The selected RSU node downloads the global model from the blockchain and sends it to the vehicle nodes that have joined the model training task within their respective coverage areas; The vehicle node iteratively trains the local model based on the global model and local data, encrypts the generated local update parameters using a secret sharing algorithm to obtain a secret share, signs the secret share, and uploads it to the RSU node. The RSU node calculates the Euclidean distance between secret shares, signs them, and uploads them to the blockchain; The blockchain reconstructs the Euclidean distance through Lagrangian interpolation, selects legal vehicle nodes, aggregates the secret shares uploaded by the legal vehicle nodes at the RSU nodes, and reconstructs the aggregation results through Lagrangian interpolation to obtain a global model of aggregation updates. The RSU node downloads the aggregated updated global model from the blockchain and sends it to the corresponding vehicle node so that the vehicle node can conduct the next round of model training until the model converges. During each round of model training, the reputation values ​​of the RSU node and the vehicle node will be stored on the chain after reputation evaluation.

9. An electronic device, characterized in that: The method comprises a memory and a processor, and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method according to any one of claims 1 to 7 is completed.

10. A computer-readable storage medium, characterized in that Used to store computer instructions, which, when executed by a processor, complete the method according to any one of claims 1 to 7.

Images