Federal distributed processing method and system based on traffic data elements

Abstract

The invention discloses a federated distributed processing method and system based on traffic data elements, and aims to solve the problems of privacy disclosure, value evaluation deficiency, difficulty in cross-domain collaboration and the like in traditional traffic data processing. A double-layer architecture used by localized value evaluation and federated data is adopted, a value vector is generated based on four dimensions through a value calculation module deployed at a data source end, and the value vector is registered to a trusted data space. When cross-domain data collaboration is needed, a federal coordination engine selects an optimal data source according to a value registry, and federal operation of data immobility and calculation dynamic is realized through technologies such as federal learning, secure multi-party calculation or homomorphic encryption. The intelligent contract manages the full life cycle of the task, and automatically settles the income according to the contribution degree. The system realizes organic unification of data privacy protection, scientific value quantification, efficient federation calculation and transparent automatic settlement, and provides a trusted infrastructure for a data element market of an intelligent traffic system.

Description

Technical Field

[0001] This application relates to the fields of traffic data processing and artificial intelligence management technology, and in particular to a federally distributed processing method and system based on traffic data elements. Background Technology

[0002] With the rapid development of intelligent transportation systems, traffic data exhibits highly distributed and heterogeneous characteristics. Due to the lack of quantitative assessment of data value under the premise of privacy protection, the selection of data sources in federated computing tasks is often blind and random, failing to prioritize high-quality, scarce data. This affects the convergence speed and final accuracy of federated learning models, and wastes computing resources and communication bandwidth. Furthermore, the lack of automated methods for contribution assessment and settlement, relying on manual intervention, significantly increases the operational complexity and latency of federated computing systems, and introduces the technical risk of collaboration failure due to settlement disputes. Existing technologies face the following core challenges:

[0003] First, the conflict between data privacy and centralized processing.

[0004] Traditional traffic data processing requires aggregating data from multiple sources, such as vehicles, parking lots, and charging stations, onto a central server for unified processing. However, this centralized architecture faces serious privacy risks. Traffic data often contains sensitive information such as users' travel trajectories and behavioral habits. Once centrally stored, it becomes an easy target for hackers and struggles to meet increasingly stringent data protection regulations.

[0005] Second, data valuation lacks scientific standards.

[0006] Current technologies lack scientific methods for evaluating the value of distributed traffic data elements. Data from different sources vary significantly in terms of quality, scarcity, utility, and timeliness, but existing systems cannot accurately quantify the true value of each data source without aggregating the original data, leading to difficulties in pricing data transactions and unfair distribution of revenue.

[0007] Third, cross-domain data collaborative computing is inefficient.

[0008] Intelligent transportation AI systems require the integration of multi-regional and multi-dimensional data for complex calculations, such as route planning, traffic prediction, and accident risk analysis. Traditional methods either transmit all data to a central node for processing (high privacy risks and high bandwidth consumption) or abandon cross-domain collaboration (resulting in inaccurate calculations), failing to strike a balance between privacy protection and computational efficiency.

[0009] Fourth, the data usage and settlement mechanism is opaque.

[0010] When multiple data providers participate in collaborative computing tasks, existing technologies struggle to accurately track the actual contributions of each party, hindering fair and automated value settlement. Manual negotiation is costly, time-consuming, and prone to disputes, severely restricting the healthy development of the data element market.

[0011] Fifth, computational trust and verification difficulties.

[0012] In a distributed environment, verifying that all participants have indeed performed computations according to the protocol and that the submitted results are genuine and reliable has long been a difficult technical challenge. The lack of a trusted computing environment and verification mechanisms makes federated computing vulnerable to malicious node attacks or data contamination. Summary of the Invention

[0013] Based on this, embodiments of this application provide a federated distributed processing method and system based on traffic data elements, solving the technical challenge of value assessment and joint calculation of multi-source heterogeneous traffic data while protecting the privacy of data sources. The system achieves the goal of keeping data stationary while computation moves, ensuring value flow and preventing information leakage through a two-layer architecture of localized data element calculation and federated data operations.

[0014] Firstly, a federated distributed processing method based on traffic data elements is provided, which includes:

[0015] S1. Each traffic data source deploys a value calculation engine locally to calculate the value vector of local data based on a four-dimensional model of quality, scarcity, utility, and timeliness.

[0016] S2. After digitally signing the value vector, write it into the data value registry on the blockchain;

[0017] S3. Receive the computing requirements submitted by the data user, query the data value registry, automatically select the optimal data source combination based on the value score, historical contribution and quotation, and issue a federated task description; wherein, the federated task description includes at least the computing type, encryption parameters and task identifier.

[0018] S4. Execute the privacy computing protocol corresponding to the computing type locally, output the encrypted intermediate result and return it to the federated coordination engine;

[0019] S5 manages the entire lifecycle of federated tasks, including task release, data source bidding, execution monitoring, contribution evaluation, and automatic settlement.

[0020] Optionally, the value vector is calculated based on a four-dimensional model of quality, scarcity, utility, and timeliness, wherein:

[0021] The value vector is calculated based on a four-dimensional model of quality, scarcity, utility, and timeliness, where:

[0022] The calculation of the quality indicators depends on the consistency verification results of local data with preset benchmark data or cross-source data;

[0023] The generation of the scarcity index depends on querying the data value registry to count the number and coverage of data sources that can provide similar data.

[0024] The prediction of the utility index is based on the actual contribution of similar data in historical federated mission records to the mission results.

[0025] The timeliness index reflects the degree of matching between the data update cycle and the real-time requirements of the federated tasks;

[0026] The value vector is used as a quantitative basis for data source selection and contribution assessment in federal tasks.

[0027] Optionally, the value vector is digitally signed and then written into the data value registry on the blockchain, specifically including:

[0028] The data value registry stores the value vector, feature summary, and digital signature of each data source in the form of a blockchain ledger. It also includes an identity authentication module to verify the legitimacy of the data source and a value query interface for authorized parties to retrieve data sources that meet specific conditions and their value scores.

[0029] Optionally, the process of selecting the optimal combination of data sources in S3 is completed by the federated coordination engine, which includes:

[0030] The task parser breaks down data usage requests into federated subtasks;

[0031] The data source optimizer selects the optimal combination of data sources based on value vectors, historical contribution records, and current availability.

[0032] The federated protocol scheduler selects and schedules federated learning, secure multi-party computation, or homomorphic encryption protocols based on the task type.

[0033] The results aggregator securely aggregates encrypted intermediate results returned from various data sources.

[0034] The contribution evaluator calculates contribution scores based on the actual performance of each data source in the task.

[0035] Optionally, the privacy computing protocol described in S4 supports at least one of the following:

[0036] Federated learning client: Train machine learning models on local data and upload model parameter updates;

[0037] Secure multi-party computation nodes participate in cross-domain statistical queries using secret sharing technology;

[0038] Homomorphic encryption computing units perform computational operations on encrypted data to support outsourced computing scenarios.

[0039] Optionally, S5 manages the entire lifecycle of federated tasks, including task deployment, data source bidding, execution monitoring, contribution evaluation, and automatic settlement. The tasks include the following specific state transition processes:

[0040] The task has been created and the data user has posted the task and deposited a security deposit.

[0041] In the bidding process, the data source submits its bid and the smart contract automatically selects the winning bidder.

[0042] The bidder has committed to participating and the successful bidder has pledged a performance bond as collateral.

[0043] In progress, each data source is executing local computing tasks;

[0044] Aggregation in progress; the federation coordinator aggregates and encrypts intermediate results.

[0045] Completed status; verify the correctness of the aggregation results;

[0046] The settlement status indicates that the earnings will be automatically distributed based on the contribution level and the deposit will be refunded.

[0047] In the disputed state, the objections of the participating parties to the settlement results are handled.

[0048] Optionally, the method further includes dynamically adjusting the value vector according to the following formula after each federated task is completed:

[0049] Q_new=Q_old×(1+α×contribution_quality)

[0050] S_new=S_old×(1+β×task_scarcity_factor)

[0051] U_new=U_old×(1+γ×utility_proven)

[0052] Wherein, contribution_quality is the task contribution quality score, task_scarcity_factor is the task scarcity factor, utility_proven is the utility proof factor, and α, β, and γ are configurable adjustment coefficients.

[0053] Optionally, when the privacy computing protocol is federated learning, the secure aggregation method includes at least:

[0054] Differential privacy federated averaging adds carefully calibrated noise to local gradient updates;

[0055] Homomorphic encryption weighted average, performing aggregation operations directly on the ciphertext;

[0056] Secure summation based on secure multi-party computation ensures that the coordinator cannot know the specific updates of any single data source.

[0057] Optionally, the method further includes a multi-layered security protection mechanism, specifically including:

[0058] Data source layer protection ensures that the original data never leaves the local machine and uses TLS 1.3 encryption for transmitted content;

[0059] Federal computing layer protection employs threshold cryptography to prevent collusion by malicious participants;

[0060] Contract and settlement layer protection uses zero-knowledge proof technology to verify the correctness of calculations;

[0061] The regulatory compliance layer provides optional regulatory audit interfaces and supports major data protection regulations, including at least the GDPR and the Personal Information Protection Act.

[0062] Secondly, a federated distributed processing system based on traffic data elements is provided, which includes:

[0063] The local value calculation module for data sources is used to deploy value calculation engines locally for each traffic data source, and calculate the value vector of local data based on a four-dimensional model of quality, scarcity, utility and timeliness.

[0064] A trusted data space, constructed based on blockchain technology, is used to digitally sign the value vector and write it into the data value registry on the blockchain.

[0065] The federated coordination engine is used to receive computing requests submitted by data users, query the data value registry, automatically select the optimal combination of data sources based on value scores, historical contributions, and quotes, and issue federated task descriptions; wherein, the federated task descriptions include at least computing type, encryption parameters, and task identifier.

[0066] The federated computing agent module is deployed at each data source end to execute the privacy computing protocol corresponding to the computing type locally, output encrypted intermediate results and return them to the federated coordination engine.

[0067] The smart contract management module, deployed on the blockchain, is used to manage the entire lifecycle of federated tasks, including task release, data source bidding, execution monitoring, contribution evaluation, and automatic settlement.

[0068] The beneficial effects of the technical solutions provided in this application include at least the following:

[0069] (1) It solves the privacy leakage risk caused by centralized processing of traffic data. Each data source completes value assessment and privacy calculation only locally.

[0070] (2) It has realized the reliable quantification and dynamic discovery of data value. Through the four-dimensional value model of localized computing and the registration mechanism based on blockchain, distributed data has obtained measurable, verifiable and evolving value identification under the premise of privacy protection, laying a reliable technical foundation for the market circulation of data elements.

[0071] (3) The resource scheduling and coordination efficiency of the federated computing system have been optimized. The federated coordination engine selects the best data sources based on the value registry, ensuring that the data sources participating in the task are optimal in terms of quality and scarcity, thereby improving the training efficiency of the federated learning model, the accuracy of secure multi-party computation, and the reliability of the results of homomorphic encryption analysis (as shown in Examples 1, 2, and 3).

[0072] (4) The entire process is automated and trustless. By encoding the task state machine and settlement rules through smart contracts, the entire process from task release to revenue distribution is fully automated, which greatly reduces the complexity of system operation and the risk of human intervention, and ensures the fairness and transparency of the collaboration process. Attached Figure Description

[0073] To more clearly illustrate the embodiments of this application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0074] Figure 1 This is a system architecture diagram for the valuation and use of federalized transportation data elements in this application;

[0075] Figure 2 This is a diagram of the local value calculation engine architecture for the data source in this application;

[0076] Figure 3 This is a flowchart of the value vector registration and verification process for this application;

[0077] Figure 4 Optimization flowchart for the federalization path planning of this application;

[0078] Figure 5 This is a flowchart of the federal learning process for this application;

[0079] Figure 6This is an example diagram of a secure multi-party computation protocol (secret sharing) for this application;

[0080] Figure 7 This is a flowchart of the homomorphic encryption computation process in this application;

[0081] Figure 8 This is a diagram illustrating the dynamic update mechanism for the data value in this application.

[0082] Figure 9 This is a flowchart illustrating the fully automated execution process of federated tasks based on smart contracts in this application.

[0083] Figure 10 This is a diagram of the multi-layered security protection system of this application;

[0084] Figure 11 This is a three-phase deployment roadmap for this application;

[0085] Figure 12 This is a matrix diagram of the core technological innovations of this application. Detailed Implementation

[0086] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0087] In the description of this application, the terms "comprising," "having," and any variations thereof are intended to cover non-exclusive inclusion, such as a process, method, system, product, or apparatus that includes a series of steps or units, not necessarily limited to those steps or units that are expressly listed, but may also include other steps or units that are not expressly listed but are inherent to these processes, methods, products, or apparatuses, or steps or units added based on further optimizations conceived in this application.

[0088] This system is a key sub-module of the "Vehicle Full-Domain AI Management and Autonomous Operation System Based on Trusted Data Space," specifically designed to address the technical challenge of value assessment and joint computation of multi-source heterogeneous traffic data while protecting data source privacy. The system achieves its goal of "data remaining stationary while computation moves, value flowing without information leakage" through a two-tiered architecture of localized data element computation and federated data operations.

[0089] The core technological innovation of this application lies in: localizing the data value assessment process, enabling each data source to calculate and register the value vector of its own data without exposing the original data; at the same time, enabling collaborative use of cross-domain data through privacy computing technologies such as federated learning, secure multi-party computation, and homomorphic encryption; and finally, automatically settling value and distributing benefits based on the actual contribution of each party in the federated task through a smart contract mechanism.

[0090] To achieve the above objectives, this application adopts the following technical solution:

[0091] This application provides a federated distributed processing method based on traffic data elements, including the following steps:

[0092] S1. Each traffic data source deploys a value calculation engine locally to calculate the value vector of local data based on a four-dimensional model of quality, scarcity, utility, and timeliness.

[0093] S2. After digitally signing the value vector, write it into the data value registry on the blockchain;

[0094] S3. Receive the computing requirements submitted by the data user, query the data value registry, automatically select the optimal data source combination based on the value score, historical contribution and quotation, and issue a federated task description; wherein, the federated task description includes at least the computing type, encryption parameters and task identifier.

[0095] S4. Execute the privacy computing protocol corresponding to the computing type locally, output the encrypted intermediate result and return it to the federated coordination engine;

[0096] S5 manages the entire lifecycle of federated tasks, including task release, data source bidding, execution monitoring, contribution evaluation, and automatic settlement.

[0097] This application provides a federated distributed processing system based on traffic data elements, including:

[0098] The local value calculation module for data sources is used to deploy value calculation engines locally for each traffic data source, and calculate the value vector of local data based on a four-dimensional model of quality, scarcity, utility and timeliness.

[0099] A trusted data space, constructed based on blockchain technology, is used to digitally sign the value vector and write it into the data value registry on the blockchain.

[0100] The federated coordination engine is used to receive computing requests submitted by data users, query the data value registry, automatically select the optimal combination of data sources based on value scores, historical contributions, and quotes, and issue federated task descriptions; wherein, the federated task descriptions include at least computing type, encryption parameters, and task identifier.

[0101] The federated computing agent module is deployed at each data source end to execute the privacy computing protocol corresponding to the computing type locally, output encrypted intermediate results and return them to the federated coordination engine.

[0102] The smart contract management module, deployed on the blockchain, is used to manage the entire lifecycle of federated tasks, including task release, data source bidding, execution monitoring, contribution evaluation, and automatic settlement.

[0103] like Figure 1 As shown, the system architecture for the evaluation and use of federalized traffic data elements in this application is presented. It adopts a layered architecture design, consisting of a data source layer, a federated computing layer, a coordination and scheduling layer, and an application interface layer from bottom to top.

[0104] Application layer: AI Manager data request interface

[0105] This layer receives various data requests from the vehicle-wide AI management system, including application scenarios such as route planning optimization, traffic flow prediction, and accident risk analysis. Each request specifies the required data type (such as real-time traffic data, historical trajectory data, and multi-source accident data), quality requirements, timeliness requirements, and budget range.

[0106] Coordination Layer: Federal Scheduling and Coordination Engine

[0107] The federated coordination engine is the central hub of the system and is responsible for: (1) receiving and analyzing data requests from the AI ​​manager and breaking them down into sub-tasks that can be executed in a federated manner; (2) selecting the optimal combination of data sources based on the data value registry, taking into account both the value of data elements and computational costs; (3) coordinating the execution of operations such as federated learning and secure multi-party computation; and (4) aggregating the encrypted intermediate results returned by each data source and finally generating computational results that satisfy the request.

[0108] Computation layer: Local data source processing module

[0109] Two core modules are deployed for each data source (vehicle cluster, roadside facilities, traffic center, etc.):

[0110] Local value assessment module: Based on a four-dimensional model, calculate the quality index (Q), scarcity index (S), utility index (U), and timeliness index (T) of local data, generate a value vector, and encrypt and upload it to the trusted data space.

[0111] The Federated Computing Agent Module implements functions such as the Federated Learning Client (FL Client), Secure Multi-Party Computation Node (MPCNode), and Homomorphic Encrypted Computation (HE Compute), and participates in local computation for federated tasks.

[0112] Foundation Layer: Core of Trusted Data Space

[0113] The Trusted Data Space is built on blockchain technology, providing core services such as smart contract management, a data element value registry, federated task scheduling records, and value settlement and payment. All value registrations, task execution records, and contribution evaluation results are stored on the blockchain in an immutable manner, ensuring transparency and traceability.

[0114] The system supports three main types of participants: data source providers (such as vehicle clusters, roadside facilities, and traffic management centers), data requesters (such as AI managers and third-party application service providers), and system operators (responsible for maintaining the infrastructure of the federal coordination engine and trusted data space).

[0115] The value vector proposed in this application is not a simple commercial label, but a type of technical metadata used to optimize the performance of a federated distributed computing system. It is generated locally from the data source and registered in a trusted manner through a blockchain. This provides the federated coordination engine with the technical basis for selecting data sources and offers smart contracts a computational benchmark for contribution evaluation and revenue settlement. Thus, it achieves optimal matching of computing and data resources at the system level, improving the overall execution efficiency and result quality of federated tasks.

[0116] Localized data element calculation process

[0117] Step 1: Value pre-calculation at the data source end

[0118] Deploy a data element value calculation engine locally on each data source. This engine performs value assessments without exposing the raw data.

[0119] like Figure 2 The diagram illustrates the architecture of the local value calculation engine for the data source, which specifically includes:

[0120] Quality assessment dimension (Q_local)

[0121] The accuracy, completeness, and timeliness of local data are calculated. Accuracy is measured by sampling comparison or consistency verification with a standard reference source; completeness is quantified by statistical missing rate and redundancy; and timeliness is assessed by measuring data update delay and staleness. The final quality metrics are derived through weighted aggregation.

[0122] Scarcity assessment dimension (S_local)

[0123] The system analyzes the geographical coverage, data source type uniqueness, and supply of local data. It queries the registry of the trusted data space to count the number and coverage of data sources that can provide similar data. The more unique the coverage and the fewer the sources that can provide this type of data, the higher the scarcity score.

[0124] Utility assessment dimension (U_local)

[0125] Predict the value of local data for different application scenarios. The system maintains a historical task database, recording the actual utility of various data in scenarios such as path optimization, event detection, and traffic prediction. Based on machine learning models, it estimates the potential contribution of local data to future tasks and generates scenario-based utility scores.

[0126] Timeliness assessment dimension (T_local)

[0127] Examine the real-time and seasonal characteristics of the data. Real-time performance is measured by the time interval between data generation and availability (e.g., update cycle); seasonality is identified by analyzing the time correlation and periodic patterns of the data. For some applications, the value of historical data decays over time, and the system adjusts for this using a time decay function.

[0128] Step 2: Trustworthy on-chaining of value vectors

[0129] After the data source generates a value vector locally, it needs to be registered on the blockchain in the trusted data space to form an immutable value certificate.

[0130] like Figure 3 The diagram illustrates the value vector registration and verification process, which is detailed below:

[0131] The data source generates a complete value vector, containing specific values ​​in four dimensions.

[0132] Generate a data feature summary, including key information such as coverage area, update frequency, data type, accuracy statement, and timestamp.

[0133] The combination of value vector and feature digest is digitally signed using the private key of the data source to ensure that the data source is verifiable and the content has not been tampered with.

[0134] The signed value package will be uploaded to the blockchain via a smart contract interface.

[0135] The smart contract verifies the validity of the digital signature and, upon confirmation, permanently stores the value record on the blockchain.

[0136] This mechanism ensures the authenticity and non-repudiation of value registration. Any attempt to tamper with registered value will result in signature verification failure and automatic rejection by the system. Simultaneously, the value registration record is linked to the data source identity, providing a reliable foundation for subsequent credit assessment and contribution tracing.

[0137] The federalized data handling process in this application specifically includes:

[0138] When the AI ​​Manager needs to perform collaborative computations across multiple data sources, the system adopts a federated operation mode to ensure that the original data is always kept locally on the respective data source, and only encrypted intermediate computation results are transmitted.

[0139] Scenario: The AI ​​manager requires multi-source traffic data for route planning and optimization.

[0140] Traditional approach: Collecting all raw data → High risk of privacy breaches

[0141] This system employs a federated computing approach, where data remains unchanged.

[0142] like Figure 4 The diagram illustrates the federated path planning optimization process, with detailed process descriptions:

[0143] The first step is for the AI ​​Manager to submit a path optimization request to the Federal Coordination Engine, specifying the starting point A, the destination B, the time window (now), and the optimization goal (shortest time).

[0144] The second step involves the Federal Coordination Engine querying the value registry to identify all data sources covering the path from A to B, and selecting the optimal combination of high-quality, scarce, and utility-matching data sources (sources X, Y, and Z) based on their value vector scores and historical contribution records.

[0145] The third step involves the coordination engine initiating a federated computing task to the selected data source, assigning a task identifier (FL_001), computing type (secure aggregation), participant list (X, Y, Z), and task parameters.

[0146] The fourth step involves each data source performing sub-task calculations locally. For example, data source X calculates the real-time travel time for road segment A to C, data source Y calculates the congestion index for road segment C to D, and data source Z calculates the accident risk probability for road segment D to B. All calculations are performed locally, and only encrypted intermediate results are output.

[0147] Fifth, the coordination engine uses a secure aggregation protocol to collect encrypted intermediate results from various data sources. The system can use homomorphic encryption to perform calculations directly on the ciphertext, or use a secure multi-party computation (MPC) protocol to aggregate results while ensuring the privacy of all parties' inputs.

[0148] The sixth step involves the coordination engine generating a final route suggestion based on the aggregation results, including the recommended route (A→C→D→B), estimated travel time (25 minutes), and confidence level assessment (92%).

[0149] In the seventh step, the coordination engine returns the path planning results to the AI ​​manager.

[0150] Step 8: After the AI ​​manager confirms acceptance of the result, it triggers the smart contract to execute the automatic payment process. The smart contract distributes the revenue proportionally based on the actual contribution of each data source in the task.

[0151] In the ninth step, the smart contract automatically transfers the revenue to the respective data source accounts, while simultaneously updating their credit scores and utility metrics in the trusted data space (such as increasing the U value of source X).

[0152] The core characteristics of the entire process:

[0153] The AI ​​manager obtained high-quality route planning results, but never accessed the raw traffic data from any data source.

[0154] Each data source received fair benefits from data use, and its sensitive data was effectively protected.

[0155] The system records transactions and contributions, providing a basis for subsequent value adjustments.

[0156] The specific federal operational technologies supported in this application include:

[0157] 1. Federated Learning

[0158] Application scenario: Training traffic flow prediction models

[0159] like Figure 5 The diagram illustrates the federated learning process, which includes:

[0160] Federated learning is primarily used in scenarios requiring the training of machine learning models, such as traffic flow prediction models and accident risk assessment models. Traditional centralized training requires aggregating all training data, while federated learning allows models to be trained on distributed data without the need for centralized data.

[0161] The specific operation process is as follows: The coordinator initializes a global model and distributes its parameters to all participating data sources. Each data source independently trains its model on its local data and calculates the gradient updates for the model parameters. The data source uploads its locally calculated gradient updates to the coordinator, rather than uploading the original training data. The coordinator uses a safe aggregation algorithm to perform a weighted average of all local updates to generate new global model parameters. This process is repeated iteratively until the model converges to satisfactory accuracy.

[0162] The system supports multiple secure aggregation methods, including differential privacy federated averaging, homomorphic encrypted weighted averaging, and secure summation based on secure multi-party computation. These methods ensure that even the coordinator cannot deduce the original data from any single data source from the uploaded gradients.

[0163] Federated learning is particularly well-suited for handling non-independent and identically distributed (Non-IID) traffic data. Traffic data from different geographical locations and time periods often exhibit significant differences, and traditional centralized training may produce models biased towards a particular type of data. Federated learning, by training on distributed heterogeneous data, can obtain a more generalizable global model.

[0164] 2. Secure Multi-Party Computation

[0165] Application scenario: Cross-regional traffic statistics query

[0166] Example: Calculate the average commute time between city A and city B without exposing their respective detailed data.

[0167] like Figure 6 The diagram illustrates a secure multi-party computation protocol (secret sharing example), which includes:

[0168] Secure multi-party computation is primarily used in scenarios where multiple data sources are required to jointly execute statistical queries or numerical calculations, but none of the parties wish to expose their specific data.

[0169] The system employs a secret sharing protocol for secure computation. First, City A divides its total commute time into multiple random shares, none of which can be used to reconstruct the original data. City B performs the same operation on its data. Each party distributes its different shares to other participants and computation nodes. Then, each participant performs computation locally on its received share, obtaining partial results. Finally, through a secure result reconstruction protocol, the parties exchange partial results and calculate the final average commute time.

[0170] Throughout the process, no single participant can access the raw data from the others, yet all parties collectively obtain accurate statistical results. The system supports basic operations such as addition and multiplication, and can be combined to implement complex statistical analysis functions.

[0171] 3. Homomorphic Encryption

[0172] Application scenario: Privacy-preserving traffic data analysis services

[0173] Example: Service providers perform calculations directly on encrypted data to obtain analysis results.

[0174] like Figure 7 The diagram illustrates the homomorphic encryption computation process, which includes:

[0175] Homomorphic encryption allows computational operations to be performed directly on ciphertext, and the decrypted result is equivalent to performing the same operation on the plaintext. This characteristic enables data sources to outsource encrypted data processing to untrusted computing service providers without worrying about data leakage.

[0176] The application scenario is a privacy-preserving traffic data analysis service. The data source encrypts local traffic data using a public key before uploading it to the computing cloud. The computing cloud directly performs various analytical operations on the encrypted data, such as encrypted addition, encrypted multiplication, and complex calculations combining these, including statistical analysis and machine learning inference. After the calculation is complete, the cloud returns the encrypted result to the authorized party, who uses their private key to decrypt and obtain the plaintext analysis result.

[0177] The advantage of homomorphic encryption is that the computer has no view of the original data, while it supports complex data analysis operations. The system supports both partial and full homomorphic encryption schemes. Partial homomorphic encryption is computationally efficient but only supports a limited number of operations, while full homomorphic encryption supports arbitrary operations but incurs higher computational overhead. The system automatically selects the optimal encryption scheme based on the specific application scenario's requirements and performance constraints.

[0178] The dynamic adjustment of data element value in federal operations in this application includes:

[0179] Federal Contribution Assessment and Value Adjustment: The value of data elements is not static and fixed, but should be dynamically adjusted based on their performance in actual federal tasks. The system establishes a value adjustment mechanism based on feedback from federal tasks.

[0180] like Figure 8 The diagram illustrates a dynamic data value update mechanism, which includes:

[0181] After each federated task is completed, the system evaluates the actual contribution quality of each data source, the manifestation of data scarcity in the task, and the degree of matching between expected and actual utility. Based on these evaluation metrics, the system updates the value vector of the data sources.

[0182] The adjustment of quality metrics is based on the task contribution quality score, which comprehensively considers the accuracy of local calculation results, the completeness of data coverage, and the timeliness of response. If the data source performs well in the task, its quality metric will be improved accordingly, and vice versa.

[0183] The scarcity metric is adjusted based on the task scarcity factor, which is the reciprocal of the number of data sources that can provide similar data. If a data source provides unique data that cannot be replaced by other sources, its scarcity score will increase significantly.

[0184] The adjustment of the utility index is based on the utility proof factor, which is the ratio of the actual utility generated by the data in this task to the expected utility. If the actual contribution of the data in the task exceeds expectations, it indicates that its utility has been underestimated, and the system will adjust the utility index upward.

[0185] The system uses adjustment coefficients (α, β, γ) to control the magnitude of value updates, avoiding drastic value fluctuations caused by accidental factors in a single task. All value update records are stored on the blockchain, forming a traceable history of value evolution.

[0186] The automatic execution process of the smart contract in this application is as follows:

[0187] like Figure 9 The diagram illustrates the fully automated execution of federated tasks based on smart contracts, specifically including:

[0188] The system automates the entire process of federated tasks and settles value through smart contracts deployed on the blockchain. The smart contracts define the complete state machine of a task from deployment to settlement, ensuring that all operations are transparent, traceable, and tamper-proof.

[0189] The state transition of a smart contract includes the following stages:

[0190] Created Status: The AI ​​Manager publishes the requirements for the federated task, specifying the task type, data requirements, quality standards, budget range, and deadline, while also depositing a security deposit into the smart contract.

[0191] Bidding Status: Eligible data sources can view task details and submit bids, outlining the data value they can offer, their price quote, and service commitments. The smart contract selects the winning data source based on a comprehensive evaluation of its value score, historical credit history, and bid price.

[0192] Committed Status: The winning data source needs to stake a certain amount of tokens as a performance guarantee. If there is any default, cheating, or malicious behavior during the task execution process, the staked tokens will be forfeited. After all participants confirm, the task enters the execution phase.

[0193] Executing Status: Each data source executes its computational tasks locally, and the federated coordinator monitors the execution progress. Smart contracts have timeout mechanisms; if a data source fails to submit results within the specified time, it will be marked as in default.

[0194] Aggregating: After all participants submit their encrypted local computation results, the federated coordinator performs a secure aggregation operation. The aggregation process can be executed efficiently off-chain, but the critical verification hash value is stored on-chain to ensure the result is verifiable.

[0195] Completed Status: After aggregation is complete, the smart contract verifies the correctness of the result. Verification can employ zero-knowledge proof technology, allowing verification of computational correctness without re-executing the entire computation. Once verification passes, the task is marked as complete.

[0196] Settled Status: The smart contract automatically deducts fees from the pre-deposited margin and distributes them to each data source based on their contribution assessment. The distribution ratio is proportional to the contribution, achieving payment based on contribution. Simultaneously, the data source's performance collateral tokens are refunded, and their credit score is updated.

[0197] Disputed Status: If any participant disagrees with the contribution assessment or settlement results, they can file a dispute within the specified time. The smart contract freezes the relevant funds and initiates arbitration proceedings. Arbitration can be conducted by a pre-designated trusted third party or a decentralized arbitration network, and the arbitration result is automatically enforced.

[0198] The automated execution mechanism of smart contracts eliminates the complexity of manual negotiation and settlement, significantly reducing transaction costs and improving the efficiency of data flow. All state transitions and fund flows are recorded on the blockchain, which can be audited and traced by any participant, ensuring fairness and transparency.

[0199] The security and privacy protection mechanisms in this application are as follows:

[0200] The multi-layered protection system refers to the system's establishment of a four-layer security protection system covering data sources, computing, contracts, and supervision.

[0201] like Figure 10 The diagram illustrates a multi-layered security protection system, which specifically includes:

[0202] Data source layer protection ensures that raw data never leaves the device or local server. Local value computation uses differential privacy technology to add carefully calibrated noise, ensuring that even if the value vector and feature summary are made public, sensitive information in the raw data cannot be deduced. All network communications are end-to-end encrypted using Transport Layer Security Protocol version 1.3 (TLS 1.3) to prevent man-in-the-middle attacks and eavesdropping.

[0203] The federated computing layer employs a cryptographically secure aggregation protocol. Secure aggregation ensures that even the coordinator cannot know the specific contributions of any individual data source, only the aggregated result. The system supports dynamic joining and leaving of participants, automatically adjusting key distribution and the aggregation protocol. To prevent collusion attacks by malicious participants, the system uses a threshold cryptography scheme; only collusion by more than a threshold number of participants can break the attack, and the practical configuration makes such collusion economically infeasible.

[0204] Contract and settlement layer protection ensures the correctness of contract logic and the absence of backdoors through open-source smart contract code and third-party security audits. The system uses zero-knowledge proof technology to verify computational correctness; the data source can prove that it correctly executed the computational task without exposing input data and intermediate processes. Considering the potential threat of quantum computing to existing cryptography, the system reserves interfaces for quantum-resistant cryptographic algorithms, allowing for smooth upgrades when necessary.

[0205] The regulatory compliance layer supports the principle of data minimization, collecting and processing only the minimum data necessary to complete the task. The system provides an optional regulatory audit interface, allowing authorized regulatory agencies to verify the system's compliance without compromising the privacy of participating parties. The system design complies with the requirements of major data protection regulations such as the EU General Data Protection Regulation (GDPR) and China's Personal Information Protection Law, and supports users' legal rights such as the right to data erasure and data portability.

[0206] Deployment recommendations include: a gradual deployment strategy.

[0207] Given the complexity and innovativeness of the system, a phased and incremental deployment strategy is recommended to gradually accumulate experience and expand the application scope.

[0208] like Figure 11 The diagram shows a three-phase deployment roadmap:

[0209] The first phase focuses on building the infrastructure and validating the core mechanisms. This includes deploying a local value computation engine to pilot data sources to perform basic quality, scarcity, utility, and timeliness assessments; implementing a simple federated averaging algorithm to support basic federated learning tasks; establishing a data value registry to allow data sources to register value vectors and query public information from other sources; and testing the basic functionality of smart contracts on a small scale.

[0210] The second phase focuses on developing decentralized coordination and advanced privacy computing capabilities. It involves deploying a complete federated coordination engine to support the automatic decomposition of complex tasks and the selection of optimal data sources. It also integrates secure multi-party computation protocols to support applications such as cross-domain statistical queries. Furthermore, it enhances smart contract functionality to achieve fully automated task management and value settlement. Finally, it expands the number and types of participating data sources, covering more geographical regions and data scenarios.

[0211] The third phase aims to create a fully autonomous federated data market. This includes providing an open interface for data source self-registration, allowing any data provider that meets the technical standards to join the system. It will support various complex federated learning tasks, including deep learning model training and reinforcement learning collaborative optimization. A data value discovery mechanism will be established, automatically determining fair data pricing through supply and demand matching and auction mechanisms. Finally, it will foster an ecosystem of third-party service providers, offering value-added services such as data preprocessing, model optimization, and results visualization.

[0212] Each stage has clearly defined technical indicators and business objectives. The next stage will only begin after the stage objectives have been achieved and the system stability has been fully verified.

[0213] The core technological innovations of this application are mainly reflected in the following aspects:

[0214] like Figure 12 The following is a matrix of core technology innovations:

[0215] Innovation Point 1: A Two-Tier Architecture of Localized Value Assessment and Federated Data Use

[0216] Existing technologies either centrally evaluate data (leading to high privacy risks) or fail to scientifically quantify the value of distributed data. This application pioneers a two-tier architecture: value assessment is performed locally at the data source, outputting only value vectors and feature summaries without exposing the raw data; data usage is achieved through federated computing, ensuring "data remains stationary while value flows." This innovation fundamentally resolves the contradiction between data privacy protection and value circulation.

[0217] Innovation Point Two: Four-Dimensional Value Model and Dynamic Adjustment Mechanism

[0218] To address the lack of scientific standards for assessing the value of traffic data elements, this application proposes the QSUT four-dimensional value assessment model, which comprehensively considers quality, scarcity, utility, and timeliness. More importantly, the system innovatively introduces a dynamic value adjustment mechanism based on federated task feedback, enabling data value to be updated in real time according to actual usage effects, thus achieving a closed loop in value discovery.

[0219] Innovation Point 3: Fully Automated Settlement Driven by Smart Contracts

[0220] Existing data transactions rely on manual negotiation, which is costly and prone to disputes. This application designs a complete smart contract state machine to automate the entire process from task issuance, data source bidding, federated computation execution to value settlement. A contribution evaluation algorithm is embedded in the smart contract to ensure the fairness and transparency of payment based on contribution, significantly reducing transaction costs.

[0221] Innovation Point 4: Deep Integration of Multi-Privacy Computing Technologies

[0222] Unlike existing solutions that rely on a single privacy-preserving computation technology, this application systematically integrates multiple privacy-preserving computation technologies, including federated learning, secure multi-party computation, homomorphic encryption, and zero-knowledge proofs. The system automatically selects the optimal combination of technologies based on different application scenarios (model training, statistical querying, outsourced computation, and result verification), achieving the best balance between privacy protection strength, computational efficiency, and functional completeness.

[0223] Innovation Point Five: Trusted Data Space Blockchain Infrastructure

[0224] To address trust issues in distributed environments, this application constructs a blockchain-based trusted data space, enabling fully auditable data value registration, federated task recording, and contribution traceability. The immutability of blockchain ensures the credibility of value assessment and the non-repudiation of settlement. Simultaneously, a hybrid architecture combining on-chain storage of verification hashes and off-chain execution of intensive computation balances the blockchain's trustworthiness with performance requirements.

[0225] Innovation Point Six: Dedicated Optimization for Traffic Scenarios

[0226] In response to the spatiotemporal characteristics, real-time requirements, and heterogeneity of traffic data, this application specifically designs a lightweight value assessment algorithm (supporting deployment on edge devices), an efficient federated aggregation protocol (adapting to mobile network environments), and a highly adaptable task decomposition strategy (processing non-independent and identically distributed data) to ensure the availability and performance of the system in practical traffic applications.

[0227] The following are several specific embodiments of this application based on the above method and system:

[0228] Example 1: Training a traffic flow prediction model based on federated learning

[0229] The application background includes: A smart traffic management platform needs to train a city-level traffic flow prediction model to support real-time route planning and traffic signal optimization. The platform needs to integrate historical traffic flow data from three different regional traffic management centers (Area A, Area B, and Area C). However, due to data privacy and data sovereignty concerns, each regional center is unwilling to centrally upload the raw dataset.

[0230] The system deployment includes: deploying a data source local value calculation module and a federated computing agent module in each of the three regional traffic management centers (A, B, and C); deploying a federated coordination engine on the intelligent traffic management platform; and deploying a trusted data space and smart contract management module on the blockchain network.

[0231] Step 1: Local Value Assessment and Registration

[0232] The local value calculation engine of the Traffic Management Center in Area A is activated to conduct a four-dimensional evaluation of its historical traffic flow data:

[0233] Quality Assessment (Q): The system sampled 10% of the data and compared it with the GPS floating car data, calculating an accuracy rate of 93%. Data integrity was checked, revealing a missing data rate of 2%. The overall quality indicator Q_A was assessed to be 0.91.

[0234] Scarcity Assessment (S): The system queries the value registry of the trusted data space and finds that there are only two data sources that can provide road network data for area A (area A itself and a private floating car service provider, but the latter has a coverage of only 40%). Based on the uniqueness of coverage and the scarcity of supply, S_A is calculated to be 1.25.

[0235] Utility Assessment (U): System analysis of historical task records revealed that data in area A was used in five traffic prediction tasks, with an average contribution score of 0.85. Based on task matching and historical performance, the estimated U_A is 2.1.

[0236] Timeliness Assessment (T): Data in Area A is updated every minute and includes complete historical data for the past three years. Considering the requirements of traffic forecasting for real-time performance and historical depth, the assessment value of T_A is 1.1.

[0237] The system in Zone A generates a value vector V_A equal to square brackets 0.91 comma 1.25 comma 2.1 comma 1 square brackets, and generates a feature digest (coverage area, update frequency, data type, accuracy declaration, etc.). The value vector and feature digest are digitally signed using the private key of Zone A, and then uploaded to the blockchain via a smart contract interface. After the smart contract verifies the signature validity, the value record is permanently stored.

[0238] Areas B and C perform similar local value assessment processes, generating and registering their respective value vectors V_B and V_C.

[0239] Step Two: Federated Learning Task Release

[0240] The intelligent traffic management platform submits a model training request to the federal coordination engine:

[0241] Task type: Federated learning (traffic flow prediction model)

[0242] Data requirements: Historical traffic flow data covering the entire city, spanning at least two years.

[0243] Quality requirements: Accuracy rate no less than 90%, missing rate no more than 5%.

[0244] Budget range: 5,000 data coins

[0245] Deadline: Complete model training within seven days

[0246] The platform deposited 5,000 data coins into the smart contract as a security deposit.

[0247] Step 3: Data Source Selection and Bidding

[0248] The federal coordination engine queries the value registry to identify all data sources covering the target area. It finds that the value vectors of data sources in zones A, B, and C all meet the quality requirements, and the combination of the three can cover the entire city's road network.

[0249] The coordination engine calculates a comprehensive score: Score equals alpha multiplied by Q, plus beta multiplied by S, plus gamma multiplied by U, plus delta multiplied by T, where alpha, beta, gamma, and delta are weighting coefficients. Based on the score ranking, regions A, B, and C are all selected to participate in the task.

[0250] The smart contract sends bidding invitations to three data sources. Each data source submits a bid (based on its own value vector and historical contribution record). The smart contract evaluates the bids based on the value score and the bids, determines the winning bidder, and notifies the three data sources.

[0251] Step Four: Execution of Federated Learning

[0252] Each of the three data sources pledged 1,000 data coins as a performance guarantee, and the task entered the execution phase.

[0253] The federated coordination engine initializes a global traffic flow prediction model (e.g., a time series prediction model based on a Long Short-Term Memory (LSTM) network) and distributes the initial model parameters G_0 to three data sources, A, B, and C.

[0254] In region A, the model is trained on local data. After ten training epochs, the gradient Delta_A for updating the local model parameters is calculated. To protect gradient privacy, differential privacy noise is added to Delta_A in region A to generate Delta_A_DP.

[0255] Similar local training is performed in regions B and C, generating Delta_B_DP and Delta_C_DP.

[0256] Three data sources upload encrypted gradient updates to the federated coordination engine. The coordination engine calculates a weighted average using a secure aggregation protocol (e.g., secure summation based on homomorphic encryption): Delta_avg equals SecureAggregate (left bracket: Delta_A_DP comma; Delta_B_DP comma; Delta_C_DP right bracket).

[0257] The coordination engine updates the global model: G_1 equals G_0 plus eta multiplied by Delta_avg, where eta is the learning rate.

[0258] The updated global model G_1 was distributed to three data sources, and the above training and aggregation process was repeated. After one hundred iterations, the model achieved a prediction accuracy of 94% on the validation set, meeting the task requirements.

[0259] Step 5: Contribution Assessment and Settlement

[0260] The federal coordination engine assesses the contribution of each data source:

[0261] contribution_quality_A equals local result accuracy multiplied by data coverage integrity multiplied by timeliness, which equals 0.93 multiplied by 0.98 multiplied by 1.0, which equals 0.91.

[0262] Similarly, we calculate that contribution_quality_B equals 0.88 and contribution_quality_C equals 0.85.

[0263] Smart contracts distribute rewards based on contribution:

[0264] The revenue from Area A equals 5,000 multiplied by 0.91 divided by the left parenthesis (0.91 plus 0.88 plus 0.85), which is approximately equal to 1,728 data coins.

[0265] The earnings from Zone B are approximately equivalent to 1667 data coins.

[0266] The revenue from Zone C is approximately equivalent to 1,605 data coins.

[0267] The smart contract automatically deducts fees from the pre-deposited margin and transfers the funds to the three data source accounts. Simultaneously, it refunds each account's performance deposit of one thousand data coins.

[0268] Step Six: Dynamic Update of Value Vector

[0269] Based on the feedback from this task, the system updates the value vectors of each data source:

[0270] Q_A_new equals Q_A_old multiplied by the left parenthesis plus alpha multiplied by contribution_quality_A, and the right parenthesis equals 0.91 multiplied by the left parenthesis plus 0.1 multiplied by 0.91, and the right parenthesis approximately equals 0.99.

[0271] U_A_new equals U_A_old multiplied by the left parenthesis, plus gamma multiplied by the right parenthesis, where utility_proven_A equals the actual utility divided by the expected utility, approximately 1.05. Therefore, U_A_new is approximately 2.43.

[0272] The updated value vector is re-registered to the blockchain via a smart contract, forming a traceable history of value evolution.

[0273] The technical benefits include: Through this embodiment, the intelligent traffic management platform successfully trained a high-precision traffic flow prediction model without accessing any raw traffic data. The data privacy of the three regional traffic management centers was fully protected, while they also received fair benefits from data usage. The entire process, from task deployment to settlement, was fully automated, taking only seven days, far shorter than traditional manual negotiation methods (which typically take several months).

[0274] Example 2: Cross-city commuting time statistics based on secure multi-party computation

[0275] The application context includes: A regional transportation planning agency needs to analyze the average commute time between city A and city B to assess the necessity of constructing a new highway. However, neither city's transportation management department is willing to directly share its detailed vehicle trajectory data.

[0276] The system deployment includes: deploying data source modules and federated computing agents in the transportation management departments of City A and City B; and deploying a federated coordination engine in the regional transportation planning agency.

[0277] Step 1: Value Registration

[0278] City A and City B each complete their local value assessment, generate value vectors, and register them in the trusted data space.

[0279] Step 2: Statistical Query Request

[0280] The regional transportation planning agency submitted a query request: to calculate the average commute time between cities A and B, with the data covering the most recent three months.

[0281] The federal coordination engine selects City A and City B as data sources and initiates a federal task based on secure multi-party computation.

[0282] Step 3: Execution of the Secret Sharing Protocol

[0283] City A calculates the total commuting time Sum_A and the number of commutes Count_A between City A and City B in its local data.

[0284] City A splits Sum_A into three random shares: Sum_A1, Sum_A2, and Sum_A3, such that Sum_A1 plus Sum_A2 plus Sum_A3 equals Sum_A. Similarly, it splits Count_A.

[0285] City A sends shares Sum_A1 and Count_A1 to City B, sends Sum_A2 and Count_A2 to compute node C (a trusted computing environment provided by the federated coordination engine), and retains Sum_A3 and Count_A3.

[0286] City B performs similar secret sharing and distribution operations.

[0287] Step 4: Local Computing and Security Aggregation

[0288] City A calculates locally: S_A equals Sum_A3 plus Sum_B2 (received from City B) plus an extra share (received from node C).

[0289] For City B, the calculation is as follows: S_B equals Sum_B3 plus Sum_A2 (received from City A) plus an additional share.

[0290] Compute node C in its secure environment calculates: S_C equals the sum of all shares it receives.

[0291] The three parties exchange partial results, but no single party can restore the original data Sum_A or Sum_B of the other party.

[0292] Ultimately, the average commute time was calculated using the secure aggregation protocol: the sum of the left bracket S_A plus S_B plus S_C and the right bracket divided by the sum of the left bracket C_A plus C_B plus C_C and the right bracket equals forty-five minutes.

[0293] Step 5: Settlement and Value Update

[0294] The smart contract allocates query fees based on the data coverage and engagement of the two cities. The value vectors of the two cities are updated based on the accuracy and timeliness of the query results.

[0295] The technological benefits include: through secure multi-party computation, regional transportation planning agencies obtained accurate statistics on average commute times, but neither city's traffic management department disclosed detailed vehicle trajectory data. The entire query process took less than an hour, and the results were verifiable via smart contracts, ensuring transparency and credibility.

[0296] Example 3: Traffic Data Outsourcing Analysis Based on Homomorphic Encryption

[0297] The application context includes: A third-party traffic data analytics service provider needs to perform statistical analysis on historical parking data from multiple parking lots to generate parking heatmaps and predictive models. However, the parking lot operators are concerned about data leaks and are unwilling to provide plaintext data.

[0298] The system deployment includes: deploying data source modules and homomorphic encrypted computing units in each parking lot; deploying a computing cloud platform from a third-party service provider; and providing key management services through a trusted data space.

[0299] Step 1: Data Encryption and Upload

[0300] Each parking lot uses a public key to homomorphically encrypt its historical parking data (such as the number of cars parked per hour, average parking duration, etc.).

[0301] Encrypted data such as E (left bracket) data_A (right bracket), E (left bracket) data_B (right bracket), etc., are uploaded to the computing cloud of a third-party service provider.

[0302] Step 2: Ciphertext Calculation

[0303] Third-party service providers perform statistical analysis directly on encrypted data without knowing the original data content:

[0304] Ciphertext addition: Calculate the total number of parked cars. (Left parenthesis sum right parenthesis equals E; left parenthesis data_A right parenthesis circle plus E; left parenthesis data_B right parenthesis circle plus dots.)

[0305] Ciphertext multiplication: operations such as calculating weighted averages

[0306] By leveraging the homomorphic nature of addition and multiplication in homomorphic encryption, service providers can perform complex statistical analysis and machine learning inference.

[0307] Step 3: Result Return and Decryption

[0308] The service provider returns the encrypted analysis result (E, left bracket, result, right bracket) to the authorizing party (such as the regional transportation authority).

[0309] The authorizing party uses the private key to decrypt: result equals Decrypt (left parenthesis E left parenthesis result right parenthesis comma private_key right parenthesis), obtaining the plaintext parking heatmap and prediction model parameters.

[0310] Step 4: Settlement and Value Update

[0311] The smart contract allocates revenue based on the amount, quality, and complexity of data provided by each parking lot. The utility metric U for each parking lot is updated based on the usefulness of the analysis results.

[0312] The technical benefits include: Through homomorphic encryption, third-party service providers successfully completed data analysis tasks across multiple parking lots without accessing any of the parking lot's plaintext data. The parking lot's trade secrets were protected, while revenue from data usage was realized. The accuracy of the analysis results was comparable to that on plaintext data, but the computation time increased slightly (approximately three to five times that of plaintext computation, still within acceptable limits).

[0313] Industrial Application Prospects: The federated distributed processing system for traffic data elements proposed in this application has broad industrial application prospects and can play an important role in the following areas:

[0314] Intelligent Transportation Systems: Provides cross-regional and cross-departmental data collaboration capabilities for intelligent transportation management platforms, supports applications such as real-time route planning, traffic flow prediction, and accident risk analysis, while ensuring the privacy and data sovereignty of each data source.

[0315] Autonomous driving technology: Provides high-quality training data and real-time road condition information for autonomous driving systems, and trains perception and decision models through federated learning technology to avoid centralizing sensitive vehicle driving data to a single entity.

[0316] Urban planning and policy making: Providing government departments with traffic statistical analysis and policy evaluation tools based on multi-source data, supporting data-driven urban planning and traffic policy making while protecting the privacy of businesses and individuals.

[0317] Data element market development: Provide a reliable infrastructure for the transportation data element market, realize the scientific assessment of data value, transparent trading and fair settlement, and promote the orderly circulation and efficient allocation of data elements.

[0318] Cross-border data cooperation: Under the premise of complying with the data protection regulations of various countries, support cross-border transportation data cooperation, such as international logistics optimization and cross-border tourism flow analysis, to promote the intelligent collaboration of global transportation systems.

[0319] This application addresses a core pain point in the field of transportation data through technological innovation, providing a complete solution for data element management in intelligent transportation systems. The system's modular design and standardized interfaces facilitate easy integration into existing transportation information systems, exhibiting excellent scalability and compatibility. With the maturation of privacy-preserving computing technologies and the development of the data element market, this application is expected to be widely applied in fields such as intelligent transportation and smart cities, generating significant economic and social benefits.

[0320] In one embodiment of this application, a computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the above-described federated distributed processing method for traffic data elements.

[0321] In one embodiment of this application, a computer program product is provided, including a computer program / instructions, which, when executed by a processor, implements the steps of the above-described federated distributed processing method for traffic data elements.

[0322] The computer-readable storage medium and computer program product provided in this embodiment are similar in implementation principle and technical effect to the above method embodiments, and will not be repeated here.

[0323] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods.

[0324] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0325] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

Claims

1. A federated distributed processing method based on traffic data elements, characterized in that, The method includes: S1. Each traffic data source deploys a value calculation engine locally to calculate the value vector of local data based on a four-dimensional model of quality, scarcity, utility, and timeliness. S2. After digitally signing the value vector, write it into the data value registry on the blockchain; S3. Receive the computing requirements submitted by the data user, query the data value registry, automatically select the optimal data source combination based on the value score, historical contribution and quotation, and issue a federated task description; wherein, the federated task description includes at least the computing type, encryption parameters and task identifier. S4. Execute the privacy computing protocol corresponding to the computing type locally, output the encrypted intermediate result and return it to the federated coordination engine; S5 manages the entire lifecycle of federated tasks, including task release, data source bidding, execution monitoring, contribution evaluation, and automatic settlement.

2. The federated distributed processing method according to claim 1, characterized in that, The value vector is calculated based on a four-dimensional model of quality, scarcity, utility, and timeliness, where: The calculation of the quality indicators depends on the consistency verification results of local data with preset benchmark data or cross-source data; The generation of the scarcity index depends on querying the data value registry to count the number and coverage of data sources that can provide similar data. The prediction of the utility index is based on the actual contribution of similar data in historical federated mission records to the mission results. The timeliness index reflects the degree of matching between the data update cycle and the real-time requirements of the federated tasks; The value vector is used as a quantitative basis for data source selection and contribution assessment in federal tasks.

3. The federated distributed processing method according to claim 1, characterized in that, The value vector is digitally signed and then written into the data value registry on the blockchain, specifically including: The data value registry stores the value vector, feature summary, and digital signature of each data source in the form of a blockchain ledger. It also includes an identity authentication module to verify the legitimacy of the data source and a value query interface for authorized parties to retrieve data sources that meet specific conditions and their value scores.

4. The federated distributed processing method according to claim 1, characterized in that, The process of selecting the optimal combination of data sources in S3 is completed by the federated coordination engine, which includes: The task parser breaks down data usage requests into federated subtasks; The data source optimizer selects the optimal combination of data sources based on value vectors, historical contribution records, and current availability. The federated protocol scheduler selects and schedules federated learning, secure multi-party computation, or homomorphic encryption protocols based on the task type. The results aggregator securely aggregates encrypted intermediate results returned from various data sources. The contribution evaluator calculates contribution scores based on the actual performance of each data source in the task.

5. The federated distributed processing method according to claim 1, characterized in that, The privacy-preserving computation protocol described in S4 supports at least one of the following: Federated learning client: Train machine learning models on local data and upload model parameter updates; Secure multi-party computation nodes participate in cross-domain statistical queries using secret sharing technology; Homomorphic encryption computing units perform computational operations on encrypted data to support outsourced computing scenarios.

6. The federated distributed processing method according to claim 1, characterized in that, S5 manages the entire lifecycle of federated tasks, including task deployment, data source bidding, execution monitoring, contribution evaluation, and automatic settlement. The specific state transition processes for each task are as follows: The task has been created and the data user has posted the task and deposited a security deposit. In the bidding process, the data source submits its bid and the smart contract automatically selects the winning bidder. The bidder has committed to participating and the successful bidder has pledged a performance bond as collateral. In progress, each data source is executing local computing tasks; Aggregation in progress; the federation coordinator aggregates and encrypts intermediate results. Completed status; verify the correctness of the aggregation results; The settlement status indicates that the earnings will be automatically distributed based on the contribution level and the deposit will be refunded. In the disputed state, the objections of the participating parties to the settlement results are handled.

7. The federated distributed processing method according to claim 1, characterized in that, The method also includes dynamically adjusting the value vector according to the following formula after each federated task is completed: Q_new=Q_old×(1+α×contribution_quality); S_new=S_old×(1+β×task_scarcity_factor); U_new=U_old×(1+γ×utility_proven); Wherein, contribution_quality is the task contribution quality score, task_scarcity_factor is the task scarcity factor, utility_proven is the utility proof factor, and α, β, and γ are configurable adjustment coefficients.

8. The federated distributed processing method according to claim 1, characterized in that, When the privacy-preserving computation protocol is federated learning, the secure aggregation method includes at least the following: Differential privacy federated averaging adds carefully calibrated noise to local gradient updates; Homomorphic encryption weighted average, performing aggregation operations directly on the ciphertext; Secure summation based on secure multi-party computation ensures that the coordinator cannot know the specific updates of any single data source.

9. The federated distributed processing method according to claim 1, characterized in that, The method also includes a multi-layered security protection mechanism, specifically including: Data source layer protection ensures that the original data never leaves the local machine and uses TLS 1.3 encryption for transmitted content; Federal computing layer protection employs threshold cryptography to prevent collusion by malicious participants; Contract and settlement layer protection uses zero-knowledge proof technology to verify the correctness of calculations; The regulatory compliance layer provides optional regulatory audit interfaces and supports major data protection regulations, including at least the GDPR and the Personal Information Protection Act.

10. A federated distributed processing system based on traffic data elements, characterized in that, The system includes: The local value calculation module for data sources is used to deploy value calculation engines locally for each traffic data source, and calculate the value vector of local data based on a four-dimensional model of quality, scarcity, utility and timeliness. A trusted data space, constructed based on blockchain technology, is used to digitally sign the value vector and write it into the data value registry on the blockchain. The federated coordination engine is used to receive computing requests submitted by data users, query the data value registry, automatically select the optimal combination of data sources based on value scores, historical contributions, and quotes, and issue federated task descriptions; wherein, the federated task descriptions include at least computing type, encryption parameters, and task identifier. The federated computing agent module is deployed at each data source end to execute the privacy computing protocol corresponding to the computing type locally, output encrypted intermediate results and return them to the federated coordination engine. The smart contract management module, deployed on the blockchain, is used to manage the entire lifecycle of federated tasks, including task release, data source bidding, execution monitoring, contribution evaluation, and automatic settlement.

Images