Data processing systems and methods, computing devices, computer-readable storage media

Abstract

The embodiment of the specification provides a data processing system and method, a computing device and a computer readable storage medium. The data processing system comprises an interactive execution subsystem, and the interactive execution subsystem comprises an agent interaction gateway. The agent interaction gateway is configured to receive an agent interaction request sent by a requesting agent, verify the requesting agent according to a decentralized identifier, an intention proof, a digital signature, a delegation credential and task execution data, forward the agent interaction request to a target agent in the case of passing the verification, enable the target agent to execute a target task according to a task execution parameter, and return a task execution result to the agent interaction gateway, and receive and return the task execution result to the requesting agent. Through the agent interaction gateway, the agent interaction request is verified in multiple dimensions, identity trust, authority delegation, cross-domain mutual trust, dynamic risk assessment and traceability are solved, and the safety and reliability of Agent interaction are comprehensively improved.

Description

Technical Field

[0001] The embodiments in this specification relate to the field of artificial intelligence technology, and in particular to data processing systems and methods, computing devices, and computer-readable storage media. Background Technology

[0002] With the development of large language models and AI Agent (intelligent agent) technologies, intelligent agents are gradually gaining the ability to autonomously execute complex tasks, and direct interaction between agents will become the mainstream interaction mode. However, this mode brings many security challenges, such as identity trust, authorization delegation, cross-domain mutual trust, dynamic risk assessment, and accountability. Traditional identity authentication and authorization systems are mostly designed for human users and are difficult to adapt to the needs of autonomous agent interaction. Therefore, there is an urgent need for a new identity authentication and authorization solution that is suitable for multi-agent scenarios. Summary of the Invention

[0003] In view of the above, embodiments of this specification provide a data processing system. One or more embodiments of this specification also relate to a data processing method, a computing device, a computer-readable storage medium, and a computer program product, to address the technical deficiencies existing in the prior art.

[0004] According to a first aspect of the embodiments of this specification, a data processing system is provided, including an interaction execution subsystem, wherein the interaction execution subsystem includes an intelligent agent interaction gateway, wherein... The intelligent agent interaction gateway is used to receive intelligent agent interaction requests sent by a requesting intelligent agent. The intelligent agent interaction request includes the decentralized identifier, intent proof, digital signature, delegation credential, and task execution data of the requesting intelligent agent. The task execution data is used to indicate the target task to be executed by the target intelligent agent and the task execution parameters corresponding to the target task. The requesting agent is verified based on the decentralized identifier, the intent proof, the digital signature, the delegation credential, and the task execution data. If the verification is successful, the agent interaction request is forwarded to the target agent so that the target agent can execute the target task according to the task execution parameters and return the task execution result to the agent interaction gateway. Receive and return the task execution result to the requesting agent.

[0005] According to a second aspect of the embodiments of this specification, a data processing method is provided, applied to an intelligent agent interaction gateway, comprising: Receive an agent interaction request sent by a requesting agent, wherein the agent interaction request includes the requesting agent's decentralized identifier, intent proof, digital signature, delegation credential and task execution data, wherein the task execution data is used to indicate the target task to be executed by the target agent and the task execution parameters corresponding to the target task; The requesting agent is verified based on the decentralized identifier, the intent proof, the digital signature, the delegation credential, and the task execution data. If the verification is successful, the agent interaction request is forwarded to the target agent so that the target agent can execute the target task according to the task execution parameters and return the task execution result to the agent interaction gateway. Receive and return the task execution result to the requesting agent.

[0006] According to a third aspect of the embodiments of this specification, a computing device is provided, comprising: Memory and processor; The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions, which, when executed by the processor, implement the steps of the above-described data processing method.

[0007] According to a fourth aspect of the embodiments of this specification, a computer-readable storage medium is provided that stores a computer program / instructions that, when executed by a processor, implement the steps of the data processing method described above.

[0008] According to a fifth aspect of the embodiments of this specification, a computer program product is provided, including a computer program / instructions that, when executed by a processor, implement the steps of the above-described data processing method.

[0009] One embodiment of this specification implements a data processing system, including an interactive execution subsystem. The interactive execution subsystem includes an intelligent agent interaction gateway, which is configured to receive intelligent agent interaction requests sent by a requesting intelligent agent. The intelligent agent interaction request includes the requesting intelligent agent's decentralized identifier, intent proof, digital signature, delegation credential, and task execution data. The task execution data indicates the target task to be executed by the target intelligent agent and the corresponding task execution parameters. The system verifies the requesting intelligent agent based on the decentralized identifier, intent proof, digital signature, delegation credential, and task execution data. If the verification is successful, the system forwards the intelligent agent interaction request to the target intelligent agent, enabling the target intelligent agent to execute the target task according to the task execution parameters and return the task execution result to the intelligent agent interaction gateway. The system also receives and returns the task execution result to the requesting intelligent agent.

[0010] Specifically, this data processing system, through the intelligent agent interaction gateway of the interactive execution subsystem, performs unified multi-dimensional verification on intelligent agent interaction requests carrying decentralized identifiers, intent proofs, digital signatures, delegation credentials, and task execution data. It can achieve decentralized and trusted authentication of intelligent agent identities based on decentralized identifiers, and effectively resist malicious agents forging identities by combining digital signature verification, thus solving the identity trust issue. Through strict verification of delegation credentials and task execution data, it achieves secure delegation of user permissions to intelligent agents and boundary control, preventing unauthorized execution and solving the permission delegation problem. The decentralized identifier possesses universal trust characteristics across platforms, organizations, and trust domains, breaking down trust barriers between different trust domains and enabling cross-domain agent authentication. Highly efficient mutual trust between agents solves cross-domain trust issues. During the verification process, the intelligent agent interaction gateway comprehensively analyzes intent proof, permission information, and interaction data in real time for dynamic discrimination. It can accurately identify abnormal requests and unauthorized behaviors, realize dynamic risk assessment of agent autonomous behavior, and address the uncertainty risk of their behavior. At the same time, intent proof fixes the content of interaction behavior, and digital signature ensures that the content is tamper-proof and the subject is non-repudiable, forming a credible behavior record. This provides a reliable basis for liability determination and behavior auditing, and enables traceability of interaction behavior. Thus, it comprehensively solves the problems of identity credibility, permission delegation, cross-domain trust, dynamic risk assessment, and traceability in multi-agent autonomous interaction scenarios, significantly improving the security, standardization, and reliability of interactions between agents. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of the structure of a data processing system provided in one embodiment of this specification; Figure 2 This is a flowchart of a multidimensional trust assessment in a data processing system provided in one embodiment of this specification; Figure 3 This is a flowchart illustrating the complete verification of inter-agent interaction executed by the interactive execution subsystem in a data processing system, as provided in one embodiment of this specification. Figure 4 This is a flowchart illustrating the agent identity registration process performed by the identity management subsystem in a data processing system, as provided in one embodiment of this specification. Figure 5 This is a flowchart illustrating user delegation and authorization performed by a delegation and authorization subsystem in a data processing system, as provided in one embodiment of this specification. Figure 6 This is a flowchart illustrating the dispute tracing process performed by a dispute tracing subsystem in a data processing system, as provided in one embodiment of this specification. Figure 7 This is a flowchart illustrating a data processing method provided in one embodiment of this specification; Figure 8 This is a structural block diagram of a computing device provided in one embodiment of this specification. Detailed Implementation

[0012] Many specific details are set forth in the following description to provide a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar extensions without departing from the spirit of this specification. Therefore, this specification is not limited to the specific implementations disclosed below.

[0013] The terminology used in one or more embodiments of this specification is for the purpose of describing particular embodiments only and is not intended to be limiting of the one or more embodiments of this specification. The singular forms “a,” “described,” and “the” as used in one or more embodiments of this specification and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0014] It should be understood that although the terms first, second, etc., may be used to describe various information in one or more embodiments of this specification, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, first may also be referred to as second without departing from the scope of one or more embodiments of this specification, and similarly, second may also be referred to as first. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to a determination."

[0015] Furthermore, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties. Moreover, the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0016] First, the terms and concepts used in one or more embodiments of this specification will be explained.

[0017] Agent: An AI agent with autonomous decision-making capabilities, capable of executing complex task sequences, and able to make service calls, access resources, and make behavioral decisions on behalf of users or itself.

[0018] A2A: Agent-to-Agent, a communication, collaboration, and interoperability model between intelligent agents that does not require direct human intervention.

[0019] Multidimensional Trust Chain: A three-dimensional trust assessment system consisting of identity trust, behavioral trust, reputation trust, and environmental trust.

[0020] Autonomous Identity: A digital identity that is independently owned and controlled by the Agent, and is not dependent on a single centralized authority for issuance and management.

[0021] Dynamic authorization strategy: A permission granting mechanism that dynamically adjusts based on real-time context, risk assessment, and trust score.

[0022] Proof of intent: A verifiable statement submitted by the agent when initiating an interaction, containing the intent of the action, the expected result, and the constraints.

[0023] Authorization credentials: Credentials that grant a user to an Agent with specific scope of permissions, used to access designated resources on behalf of the user.

[0024] Trust anchor: An entity that endorses the identity of an Agent, which can be a user, organization, certification authority, or other trusted Agent.

[0025] Large Language Model (LLM) is a type of deep learning-based artificial intelligence model. It learns the statistical patterns of language and world knowledge through pre-training on massive amounts of text data, thereby possessing powerful natural language understanding, generation, and reasoning capabilities, enabling it to perform various complex tasks such as translation, writing, and code generation.

[0026] With the rapid development of large language models and AI agent technologies, agents are evolving from simple conversational assistants into intelligent agents capable of autonomously executing complex task sequences. In the future, direct interaction between agents will gradually replace the traditional model of humans interacting with services through apps / web pages. For example, a user's personal agent can directly connect with an e-commerce agent to complete shopping decisions and transactions; a corporate financial agent can autonomously reconcile accounts and file taxes with bank agents and tax agents; and a smart home agent can collaborate with energy management agents and security agents.

[0027] However, this agent-to-agent interaction model brings entirely new challenges to identity authentication and authorization, including issues of identity trustworthiness, delegation of authority, cross-domain trust, dynamic risk assessment, and behavioral traceability. Existing identity authentication systems are primarily designed for human users and cannot meet the needs of autonomous agent interaction. Therefore, a completely new agent identity authentication and authorization system is needed to address these challenges.

[0028] This specification provides a data processing system, and also relates to a data processing method, a computing device, a computer-readable storage medium, and a computer program product, which will be described in detail in the following embodiments.

[0029] To address the challenges of identity trust, authorization delegation, cross-domain trust, dynamic risk assessment, and traceability brought about by the Agent-to-Agent interaction model, the data processing system provided in the embodiments of this specification implements an overall solution through the intelligent agent interaction gateway of the interaction execution subsystem. The specific implementation process is as follows: First, the intelligent agent interaction gateway receives an intelligent agent interaction request sent by the requesting intelligent agent. This request carries the requesting intelligent agent's decentralized identifier, intent proof, digital signature, delegation credential, and task execution data. The task execution data clearly defines the target task to be executed by the target intelligent agent and the corresponding task execution parameters. Then, the intelligent agent interaction gateway performs unified multi-dimensional verification based on the information in the intelligent agent interaction request. It uses the decentralized identifier to achieve trusted authentication of the intelligent agent's identity across trust domains, solving the problems of cross-domain mutual trust and identity trust. It verifies the authenticity and non-repudiation of the interaction request through digital signature verification. Combined with the verification of delegation credential and task execution data, it achieves secure delegation of user permissions to the intelligent agent and control of behavior boundaries, solving the problem of permission delegation. Simultaneously, the intelligent agent interaction gateway verifies... The authentication process integrates intent proof, authorization information, and interaction data for real-time dynamic judgment, accurately identifying abnormal requests and unauthorized behaviors. This enables dynamic risk assessment of agent autonomous behavior, addressing the risks arising from the uncertainty of their actions. Furthermore, the combination of intent proof and digital signature forms a complete and reliable record of interaction behavior, providing a reliable basis for accountability and auditing in case of subsequent agent behavior problems, thus solving traceability issues. After successful verification, the agent interaction gateway forwards the agent's interaction request to the target agent. Once the target agent executes the target task and returns the task execution result, the agent interaction gateway receives the task execution result and feeds it back to the requesting agent. This ensures secure, standardized, and controllable interaction between agents throughout the entire process, effectively compensating for the shortcomings of existing identity authentication systems that cannot adapt to agent autonomous interaction.

[0030] See Figure 1 , Figure 1 This is a schematic diagram of the structure of a data processing system provided in one embodiment of this specification. For example... Figure 1 As shown, the system includes: Interactive execution subsystem 102, the interactive execution subsystem 102 includes an intelligent agent interaction gateway, wherein, The intelligent agent interaction gateway is used to receive intelligent agent interaction requests sent by a requesting intelligent agent. The intelligent agent interaction request includes the decentralized identifier, intent proof, digital signature, delegation credential, and task execution data of the requesting intelligent agent. The task execution data is used to indicate the target task to be executed by the target intelligent agent and the task execution parameters corresponding to the target task. The requesting agent is verified based on the decentralized identifier, the intent proof, the digital signature, the delegation credential, and the task execution data. If the verification is successful, the agent interaction request is forwarded to the target agent so that the target agent can execute the target task according to the task execution parameters and return the task execution result to the agent interaction gateway. Receive and return the task execution result to the requesting agent.

[0031] The interactive execution subsystem can be understood as a set of functional modules in the data processing system that are responsible for scheduling and managing the interaction process between intelligent agents. It is used to perform execution tasks such as request reception, security verification, request forwarding, and result feedback.

[0032] An intelligent agent interaction gateway can be understood as a unified entry point and verification node for interactions between agents. It is responsible for receiving intelligent agent interaction requests, completing identity and permission verification, forwarding legitimate intelligent agent interaction requests, and returning execution results.

[0033] A requesting agent can be understood as an AI agent that actively initiates an interaction request. For example, a personal shopping agent or a corporate financial agent on the user side can be the initiator of this interaction request.

[0034] An agent interaction request can be understood as a request for an agent to initiate a project collaboration instruction to a target agent, which is used to inform the target agent of the specific tasks to be performed and related identity, permissions, intent information, etc.

[0035] Decentralized identifiers, or DIDs, are globally unique identifiers for requesting agents. They do not rely on a single platform for management and can be used to identify the agent's identity across organizations and domains.

[0036] Intent proof can be understood as a declaration made by the requesting agent regarding the purpose, type of operation, and constraints of the agent's interaction request. In other words, it is information used to explain the true purpose and scope of the action of the requesting agent initiating this agent interaction request. For example, intent proof could be a declaration by the purchasing agent that this interaction is to create a purchase order for 4,500 yuan, with the supplier number being XXX.

[0037] Digital signatures can be understood as verification information obtained by requesting an intelligent agent to sign decentralized identifiers, intent proofs, delegation credentials, and task execution data using its private key. In other words, it is verification information generated by requesting an intelligent agent to encrypt the content of the interaction request using its own private key, which is used to prove that the request source is authentic and the content has not been tampered with.

[0038] A proxy certificate can be understood as a certificate generated by user authorization that proves the requesting agent has the authority to perform the operation corresponding to the agent's interaction request. It limits the scope of operations that the requesting agent can perform. For example, a proxy certificate may limit the requesting agent to "allow the user to inquire about bills but not to transfer funds".

[0039] Task execution data can be understood as a set of data containing the target task to be executed and its related task execution parameters. It is used to clarify the specific work that the target agent needs to complete. For example, the task execution data is "execute the transfer task, and the task execution parameters are account number, amount, and time". That is, the target task is the specific matter that the requesting agent expects the target agent to complete, and the task execution parameters are the specific input information required to execute the target task. They are the necessary data to complete the target task.

[0040] The target intelligent agent can be understood as an AI agent that receives and executes the target task, such as an e-commerce service agent, a bank agent, or a tax agent. It is the task executor in this interaction.

[0041] The task execution result can be understood as the response information generated and returned by the target intelligent agent after it has completed the target task according to the task execution parameters. For example, the task execution result may be "transaction successful, reconciliation completed, application accepted, etc."

[0042] Specifically, the intelligent agent interaction gateway is also used for: The requesting agent is authenticated based on the decentralized identifier and the digital signature to obtain the authentication result; If the authentication of the requesting agent is successful based on the authentication result, the trust score of the requesting agent on multiple evaluation dimensions is used to verify the trust level of the requesting agent and obtain the trust verification result. If the trust verification of the requesting agent is successful based on the trust verification result, the intent verification of the requesting agent is performed based on the intent proof to obtain the intent verification result. If the intent verification of the requesting agent is successful based on the intent verification result, the authorization verification of the requesting agent is performed based on the delegation credential and the task execution data to obtain the authorization verification result; If the authorization verification of the requesting agent is successful based on the authorization verification result, the agent interaction request is forwarded to the target agent.

[0043] Identity verification can be understood as verifying the authenticity and validity of the identity of the requesting agent based on decentralized identifiers and digital signatures, that is, confirming the legitimacy of the interaction subject.

[0044] The authentication result can be understood as the verification conclusion output by the authentication process, including two authentication results: successful authentication and failed authentication.

[0045] Evaluation dimensions can be understood as different considerations used to evaluate the trustworthiness of an intelligent agent, such as identity trustworthiness, behavior trustworthiness, reputation trustworthiness, and environmental trustworthiness.

[0046] Trust score can be understood as a quantitative value calculated based on multiple evaluation dimensions to characterize the trustworthiness of an agent.

[0047] Trust verification can be understood as determining whether the agent's current trust level meets the interaction requirements based on the trust score, and verifying the agent's risk level.

[0048] Trust verification results can be understood as the verification conclusions output by the trust verification process, including two types of trust verification results: trust verification successful and trust verification failed.

[0049] Intent verification can be understood as verifying whether the interaction purpose of the intelligent agent is reasonable and whether there is any abnormal or malicious behavior tendency based on the intent.

[0050] The intent verification result can be understood as the verification conclusion output by the intent verification process, including two types of intent verification results: intent verification successful and intent verification failed.

[0051] Authorization verification can be understood as verifying whether the intelligent agent has the operation authority to execute the target task and whether there is any unauthorized behavior, based on the authorization certificate and task execution data.

[0052] The authorization verification result can be understood as the verification conclusion output by the authorization verification process, including two types of authorization verification results: authorization verification successful and authorization verification failed.

[0053] In practice, when processing agent interaction requests, the agent interaction gateway executes a multi-level verification process sequentially: first, it authenticates the requesting agent using a decentralized identifier and digital signature; after successful authentication, it verifies trustworthiness by combining the agent's trust scores across multiple evaluation dimensions; after successful trustworthiness verification, it verifies the reasonableness of the agent's interaction intent based on the intent proof; after successful intent verification, it verifies authorization for the agent based on the delegation credentials and task execution data; only when all verifications pass will the agent gateway forward the legitimate agent interaction request to the target agent.

[0054] The data processing system provided in one or more embodiments of this specification employs a multi-level progressive verification mechanism, which can comprehensively improve the security and reliability of intelligent agent interaction. First, identity verification ensures the authenticity and validity of the intelligent agent's identity, preventing forgery and impersonation attacks. Second, trust verification based on multi-dimensional trust scoring dynamically assesses the risk of intelligent agent behavior, adapting to the uncertainty of its autonomous interaction. Third, intent verification identifies abnormal interaction purposes, proactively avoiding malicious behavior. Finally, authorization verification strictly limits the scope of intelligent agent operations, ensuring that it executes tasks within delegated permissions. This multi-level verification achieves both trustworthy intelligent agent identity and controllable permissions, completes dynamic risk assessment, and provides credible evidence for behavior tracing, effectively solving problems such as cross-domain mutual trust and permission delegation, and compensating for the shortcomings of traditional identity authentication systems in adapting to autonomous agent interaction.

[0055] Specifically, the intelligent agent interaction gateway is also used for: Based on the decentralized identifier, obtain the corresponding decentralized identifier document from the blockchain; Extract the public key of the requesting agent from the decentralized identifier document, and use the public key to verify the digital signature to obtain the signature verification result; The decentralized identifier is verified based on the state information in the decentralized identifier document to obtain the identifier verification result. The state information is a state field recorded in the decentralized identifier document that is used to identify the lifecycle and validity of the requesting agent's identity. The authentication result is obtained based on the signature verification result and the identifier verification result; The step of determining that the authentication of the requesting agent was successful based on the authentication result includes: If both the signature verification result and the identifier verification result are found to be successful, the identity verification of the requesting agent is determined to be successful.

[0056] The decentralized identifier document can be understood as an identity document that is bound to a decentralized identifier and stored on the blockchain. This decentralized identifier document contains identity-related information such as the agent's public key and identity status. In other words, it requests the agent's on-chain identity profile, which can be queried across domains and is tamper-proof.

[0057] A public key can be understood as a public key that is paired with the private key of the requesting agent and used to verify whether the digital signature was signed by the private key of the requesting agent, ensuring that the data has not been tampered with.

[0058] The signature verification result can be understood as the verification conclusion obtained after verifying the digital signature using the public key, including verification passed (i.e., the digital signature matches the request content and is confirmed to be issued by the corresponding private key) and verification failed (i.e., the digital signature is forged, the content is tampered with, or the private key does not match).

[0059] State information can be understood as the core fields in a decentralized identifier document used to identify the attributes related to the identity of a requesting agent. It includes not only the identity lifecycle state (such as activated, suspended, deactivated, expired, etc., representing the stage of identity existence) but also the identity validity state (such as normally valid, abnormally invalid, frozen, etc., representing whether the identity currently has the legal right to interact). Both together determine the usability of the decentralized identifier. For example, if the requesting agent's identity is in the "activated" lifecycle state and the "normally valid" validity state, it is legally interactive; if it is in the "deactivated" lifecycle state or the "frozen" validity state, it is illegal and cannot interact.

[0060] The identifier verification result can be understood as the conclusion obtained after verifying the validity and legality of the decentralized identifier, including verification passed (both the lifecycle and the validity state are legal) and verification failed (either the lifecycle or the validity state is illegal).

[0061] In practice, the intelligent agent interaction gateway first obtains the corresponding decentralized identifier document from the blockchain based on the decentralized identifier of the requesting intelligent agent; then, it extracts the public key from the decentralized identifier document and uses the public key to verify the digital signature, obtaining a signature verification result; at the same time, it verifies the validity of the decentralized identifier itself based on the state information in the decentralized identifier document, and also verifies the validity of the trust anchor endorsement certificate in the decentralized identifier document, obtaining an identifier verification result; only when both the signature verification result and the identifier verification result are verified successfully, does the intelligent agent interaction gateway determine that the requesting intelligent agent has been successfully authenticated.

[0062] In practical applications, to improve verification efficiency and intercept invalid requests in advance, decentralized identifiers can be verified first: extract state information from the obtained decentralized identifier document, verify the decentralized identifier's lifecycle status (e.g., whether it is activated or not expired) and validity status (e.g., whether it is normal or not frozen), and obtain the identifier verification result; if the identifier verification fails, the identity verification of the requesting agent can be directly determined to have failed, without the need for subsequent digital signature verification. If the decentralized identifier verification passes, then digital signature verification is performed: extract the requesting agent's public key from the decentralized identifier document, use this public key to verify the digital signature in the agent's interaction request, determine whether the digital signature is genuine and whether the request content has been tampered with, and obtain the signature verification result. When the signature verification result is also successful, the identity verification of the requesting agent can be determined to have succeeded.

[0063] The data processing system provided in one or more embodiments of this specification completes identity verification by obtaining decentralized identifier documents through blockchain, relying on the immutability of blockchain to ensure the trustworthiness and reliability of identity information. Using public key verification of digital signatures ensures that intelligent agent interaction requests have not been tampered with and are of genuine origin, effectively resisting forgery and tampering attacks. By verifying identity status information, it can identify deregistered or invalid intelligent agent identities in real time, preventing unauthorized identities from participating in interactions. This achieves decentralized, cross-domain universal identity authentication, does not rely on a centralized certification authority, solves the identity trust problem in multi-agent interactions, and improves the security and stability of identity verification.

[0064] Specifically, the intelligent agent interaction gateway is also used for: Using a trust assessment engine, the requesting agent is calculated to obtain its identity trust score (identity trust score), behavior trust score (behavior trust score), reputation trust score (reputation trust score), and environment trust score (environment trust score). The trust score of the requesting agent is obtained based on the identity trust score, the behavior trust score, the reputation trust score, and the environment trust score. The trust score is compared with a preset score threshold to verify the trust level of the requesting agent and obtain the trust verification result. The step of determining that the trust verification of the requesting agent was successful based on the trust verification result includes: If the trust verification result is determined to be that the trust score is greater than or equal to the preset score threshold, the trust verification of the requesting agent is determined to be successful.

[0065] The trust assessment engine can be understood as a functional module used to calculate and evaluate the trust program of the requesting agent. It can automatically collect multi-dimensional relevant data of the requesting agent and complete the calculation of the requesting agent's scores in each dimension and the overall trust score.

[0066] The identity credibility dimension can be understood as a perspective for assessing the credibility of the requesting agent's identity, such as the completeness of identity information and the success rate of identity verification history.

[0067] Identity credibility score can be understood as a quantitative score calculated based on the identity credibility dimension, used to characterize the credibility of the requesting agent's identity. The higher the score, the more credible the identity.

[0068] The behavioral credibility dimension can be understood as an assessment of the compliance and standardization of the historical interaction behavior of the requesting agent, such as whether the historical interaction behavior has exceeded authority, abnormal operation, or task execution success rate.

[0069] Behavioral credibility score can be understood as a quantitative score calculated based on the behavioral credibility dimension, used to characterize the credibility of the requesting agent's behavior. For example, if the requesting agent has no abnormal behavior, the behavioral credibility score is high.

[0070] The reputation credibility dimension can be understood as a perspective for evaluating the overall reputation of the requesting agent in a multi-agent interaction ecosystem, such as the evaluations of the requesting agent by other agents and historical dispute records.

[0071] Reputation credibility score can be understood as a quantitative score calculated based on the reputation credibility dimension, used to characterize the overall reputation level of the requesting agent. For example, if the requesting agent has no negative reviews, its reputation credibility score is high.

[0072] The environmental credibility dimension can be understood as an assessment of the security and compliance of the environment in which the requesting agent initiates the interaction request, such as the device, network environment, and time scenario in which the interaction request is initiated.

[0073] The environmental credibility score can be understood as a quantitative score calculated based on the environmental credibility dimension. It is used to characterize the security and credibility of the interaction environment of the requesting agent. For example, the environmental credibility score will be higher in a secure network environment.

[0074] The preset score threshold can be understood as a pre-set score standard used to determine whether the trust level of the requesting agent is qualified. It can be set by the data processing system according to the interaction security requirements. For example, the preset score threshold can be set to 80 points or 90 points.

[0075] In practice, after completing the authentication of the requesting agent, the intelligent agent interaction gateway invokes the trust assessment engine to verify the trust level of the requesting agent. First, the trust assessment engine can collect relevant data on the requesting agent across four dimensions from trusted data sources such as the DID registry, historical interaction records, and reputation systems, calculating identity trust score, behavior trust score, reputation trust score, and environment trust score. Then, the trust assessment engine combines these four dimension scores and calculates the overall trust score of the requesting agent using a preset algorithm (such as averaging) or a preset trust assessment model. Next, the intelligent agent interaction gateway compares this overall trust score with a preset score threshold to perform trust verification. If the trust score is greater than or equal to the preset score threshold, the trust verification is considered successful; if the trust score is less than the preset score threshold, the trust verification is considered unsuccessful, and the subsequent interaction process is terminated.

[0076] In practical applications, during the scoring calculation phase for each dimension, the trust assessment engine can base its scores on preset scoring rules (fixed scoring conditions + corresponding score ranges). For example, in the identity credibility dimension, 90-100 points are awarded for all successful identity verifications, 60-80 points for occasional verification failures that have been resolved, and 0-50 points for identity anomalies or multiple verification failures. In the behavior credibility dimension, 90-100 points are awarded for no historical anomalies or unauthorized operations, 70-80 points for one minor anomaly without security risks, and 0-60 points for multiple anomalies or unauthorized records. In the reputation credibility dimension, 90-100 points are awarded for good cross-domain evaluations and no complaint records. The system assigns scores as follows: 70-80 points for no negative or positive feedback, and 0-60 points for complaints or negative interaction records. For the environmental trustworthiness dimension, 90-100 points are awarded for network security and interactions initiated on frequently used devices; 70-80 points are awarded for a general environment with no obvious risk characteristics; and 0-60 points are awarded for interactions initiated on unfamiliar networks or in unusual locations. The basic data is then converted into dimensional scores from 0-100. For example, identity trustworthiness is scored based on identity stability and verification success rate; behavioral trustworthiness is scored based on historical compliance frequency and anomaly frequency; reputation trustworthiness is scored based on evaluations from other agents; and environmental trustworthiness is scored based on network and device security status. Subsequently, the trust assessment engine uses weighted summation to synthesize a comprehensive trust score, assigning weights to each dimension's score, multiplying them separately, and then summing the results to obtain the final comprehensive trust score.

[0077] See Figure 2 , Figure 2 This is a flowchart of a multidimensional trust assessment in a data processing system provided in one embodiment of this specification, which specifically includes the following steps.

[0078] Step 202: Request the agent to send an agent interaction request to the agent interaction gateway.

[0079] Step 204: The intelligent agent interaction gateway sends a trust assessment request to the trust assessment engine, triggering the multi-dimensional trust assessment process.

[0080] Step 206: The trust assessment engine evaluates the credibility of the identity.

[0081] Step 208: The trust assessment engine performs a verification of the identity trustworthiness dimension.

[0082] Specifically, the trust assessment engine verifies the trust anchor (the on-chain identity validity of the decentralized identifier), checks the identity validity period (identity lifecycle and validity status), and verifies historical authentication records (historical identity verification pass rate and abnormal records) in terms of identity credibility.

[0083] Step 210: The trust assessment engine calculates the identity credibility score.

[0084] Specifically, after verifying the identity credibility dimension, the trust assessment engine calculates the identity credibility score of the requesting agent based on preset scoring rules.

[0085] Step 212: The trust assessment engine evaluates the credibility of the behavior.

[0086] Step 214: Verify the credibility dimension of the trust assessment engine's execution behavior.

[0087] Specifically, the trust assessment engine analyzes historical behavior records (compliance of historical interactions, unauthorized access records), detects abnormal patterns (frequency of abnormal operations, atypical behaviors), and verifies intent in the dimension of behavioral credibility. Figure 1 Consistency (the degree to which the stated intent matches the actual action).

[0088] Step 216: The trust assessment engine calculates the credibility score of the behavior.

[0089] Specifically, after completing the verification of the behavioral credibility dimension, the trust assessment engine calculates the behavioral credibility score of the requesting agent based on preset scoring rules.

[0090] Step 218: The trust assessment engine evaluates reputation credibility.

[0091] Step 220: The trust assessment engine performs a verification of the reputation credibility dimension.

[0092] Specifically, the trust assessment engine, in terms of reputation credibility, counts the success rate of interactions (completion rate of historical interaction tasks), queries user reviews (feedback and evaluation from other intelligent agents / users), and checks dispute records (historical disputes, complaints, and violation records).

[0093] Step 222: The trust assessment engine calculates the reputation credibility score.

[0094] Specifically, after completing the verification of the reputation credibility dimension, the trust assessment engine calculates the reputation credibility score of the requesting agent based on preset scoring rules.

[0095] Step 224: The trust assessment engine evaluates the credibility of the environment.

[0096] Step 226: The trust assessment engine performs a verification of the trustworthiness dimension of the execution environment.

[0097] Specifically, the trust assessment engine checks the network environment (network security level, whether it is a trusted network), verifies time and space constraints (whether the interaction time and location conform to the normal scenario), and assesses the security of the operating environment (device security, system environment risks) in terms of environmental trustworthiness.

[0098] Step 228: The trust assessment engine calculates the environment's trustworthiness score.

[0099] Specifically, after completing the verification of the environmental trustworthiness dimension, the trust assessment engine calculates the environmental trustworthiness score of the requesting agent based on preset scoring rules.

[0100] Step 230: The trust assessment engine calculates the trust score.

[0101] Specifically, after all four dimensions of scores are calculated, the trust assessment engine can use a weighted average method or preset comprehensive rules to sum the credibility scores of the four dimensions of identity, behavior, reputation, and environment to calculate the comprehensive trust score of the requesting agent.

[0102] Step 232: The trust assessment engine returns a trust score to the agent interaction gateway.

[0103] Step 234: If the trust score is greater than or equal to the preset score threshold, the intelligent agent interaction gateway determines that the trust verification of the requesting intelligent agent is successful and allows the interaction to continue.

[0104] Step 236: If the trust score is less than the preset score threshold, the intelligent agent interaction gateway determines that the trust verification of the requesting intelligent agent has failed and rejects or requests enhanced authentication.

[0105] Specifically, if the trust score is less than the preset score threshold, the trust verification of the requesting agent is deemed to have failed. The agent interaction gateway rejects the interaction request or requires the requesting agent to complete enhanced authentication (such as secondary identity verification, supplementary authorization credentials, etc.) before continuing the interaction.

[0106] The data processing system provided in one or more embodiments of this specification calculates multi-dimensional trust scores through a trust assessment engine, covering four core dimensions: identity, behavior, reputation, and environment. Compared to single-dimensional assessments, this system more comprehensively and objectively reflects the trustworthiness of the requesting agent. The four dimensions complement each other, accurately capturing potential risks in the agent's identity, behavior, and interaction environment. By setting preset score thresholds, trust levels are quantified, with clear standards and strong operability, avoiding subjective judgment bias. This method dynamically assesses the real-time trust status of agents, identifies low-trust agents in advance, effectively avoids interaction risks caused by the uncertainty of agent autonomous behavior, further enhances the security of multi-agent interactions, and provides a reliable trust basis for subsequent interactions.

[0107] Specifically, the intelligent agent interaction gateway is also used for: The intent proof is analyzed to obtain the intent analysis result; Based on the intent parsing result, the reasonableness of the requesting agent is verified to obtain the reasonableness verification result. Based on the intent parsing result, the requesting agent is subjected to intent analysis. Figure 1 Consistency verification, obtaining the intention Figure 1 Consistency verification results; Based on the results of the intent rationality verification and the intent Figure 1 The consistency verification result is used to obtain the intent verification result; The step of determining that the intent verification of the requesting agent was successful based on the intent verification result includes: After determining the reasonableness verification results of the stated intent and the stated intent Figure 1 If all consistency verification results pass, it is determined that the intent verification of the requesting agent is successful.

[0108] The process of parsing the intent proof can be understood as the intelligent agent interaction gateway parsing the intent proof, extracting key information such as operation type, parameters, and constraints, and forming a structured intent parsing result. For example, the operation type can be parsed from the intent proof as a purchase order, the amount as 4,500 yuan, and the project number as XXX.

[0109] Intent resolution results can be understood as structured data obtained after parsing, which represents the purpose and content of this interaction, including operational behavior, key parameters, business objects, etc.

[0110] Intent rationality verification can be understood as judging whether the parsed interaction intent is legal, compliant, and in line with normal business logic based on system policies, agent roles, and business context. For example, it is reasonable for a financial agent to initiate a normal expense reimbursement query, but it is unreasonable to initiate the deletion of accounting data.

[0111] meaning Figure 1 Consistency verification can be understood as checking whether the parameters, behaviors, task execution data, delegation credential constraints, historical behavior patterns, etc. in the intent parsing results are consistent with each other. For example, if the intent declaration amount is 4,500 yuan, but the task data is 9,000 yuan, it is determined to be inconsistent.

[0112] The intent verification result can be understood as a comprehensive verification of the reasonableness of the intent and the intent itself. Figure 1 The final verification conclusion obtained from the consistency verification only includes the verification of the reasonableness of the intent and the intent itself. Figure 1 Only when all consistency verifications pass can the intent verification result be considered successful.

[0113] In practice, after the trust verification of the requesting agent is passed, the intelligent agent interaction gateway first parses the intent proof carried by the requesting agent, extracting information such as operation type, business parameters, and constraints to form an intent parsing result. Subsequently, the intelligent agent interaction gateway performs intent rationality verification based on system policies and business rules, determining whether the interaction intent of the requesting agent conforms to its role permissions and business scenario. Simultaneously, the intelligent agent interaction gateway compares the intent parsing result with the defined operation scope, such as task execution data and delegation credentials, to complete the intent resolution process. Figure 1 Consistency verification. Finally, the intelligent agent interaction gateway integrates the results of the two verifications (intention rationality verification and intention verification). Figure 1 (Conformity verification), only when the intent is reasonable and the intention is... Figure 1 Only if all consistency verifications pass will the requesting agent's intent be deemed successfully verified, and the process will proceed to the subsequent authorization verification stage.

[0114] The data processing system provided in one or more embodiments of this specification structures interaction intents through intent parsing, facilitating accurate identification of the agent's interaction purpose; it effectively identifies unauthorized, illegal, and abnormal malicious intents through intent rationality verification, blocking high-risk behaviors in advance; and it further... Figure 1 Consistency verification can prevent behaviors such as discrepancies between declared intent and actual business data, and exceeding delegated authority, thus preventing data tampering and fraudulent requests. By employing dual intent verification in conjunction, the authenticity and compliance of interactive behaviors can be comprehensively guaranteed, effectively addressing the risk of uncontrollable intent arising from autonomous agent behavior, enhancing the security and reliability of interactions between agents, and laying a trustworthy foundation for subsequent authorization verification and task execution.

[0115] Specifically, the intelligent agent interaction gateway is also used for: The validity of the requesting agent is verified based on the authorization certificate, and the validity verification result is obtained. If the validity verification result of the credential is determined to be successful, the permission information of the authorization credential is matched with the preset permission policy to obtain the credential permission verification result. If the verification result of the credential permission is determined to be successful, the target permission of the requesting agent is calculated and obtained based on the permission information of the entrustment credential and the trust score. Based on the target permissions and the task execution data, the requesting agent is dynamically authorized to obtain the authorization verification result. The step of determining that the authorization verification of the requesting agent was successful based on the authorization verification result includes: If the authorization verification result is determined to be successful, the authorization verification of the requesting agent is determined to be successful.

[0116] Among them, the validity verification of credentials can be understood as the intelligent agent interaction gateway verifying the entrustment credentials carried by the requesting intelligent agent, and determining whether the entrustment credentials themselves are legal, not expired, not tampered with, not cancelled, etc. For example, verifying whether the signature of the entrustment credentials is valid and whether the validity period of the credentials is within the current time. If the entrustment credentials have expired or the signature is forged, the validity verification result is considered to be verification failure.

[0117] The validity verification result of the document can be understood as the verification conclusion output by the validity verification process of the entrustment document, which is divided into two states: verification passed (the entrustment document is legal and valid) and verification failed (the entrustment document is expired, forged, cancelled, etc.).

[0118] Permission information can be understood as the scope of operational permissions granted by the user to the requesting agent, as clearly stated in the authorization certificate. This includes the types of tasks that can be executed, the resources that can be accessed, and the operational limits. For example, permission information could be stated in the authorization certificate as "Allow the purchasing agent to create purchase orders with a value of ≤5000 yuan".

[0119] Preset permission policies can be understood as pre-configured, unified rules used to regulate the use of permissions by intelligent agents, clarifying the permission boundaries and usage norms for different types of intelligent agents and in different scenarios. For example, preset permission policies might include rules such as "the purchasing agent can create no more than 10 orders per day" or "the finance agent can only query the accounts of its own department."

[0120] The verification result of the credentials can be understood as comparing the permission information in the authorization credentials with the preset permission policy to determine whether the permission information conforms to the preset permission policy. The verification result is divided into two states: verification passed (permissions are compliant) and verification failed (permissions exceed the preset permission policy).

[0121] The target permission can be understood as the final permission that the requesting agent can actually exercise in this interaction, calculated by combining the permission information of the entrustment certificate and the trust score of the requesting agent. It is a dynamic combination of permission information and trust level. For example, if the entrustment certificate permission is "purchase amount ≤ 5000 yuan" and the trust score is 90 points (high trust), the target permission can retain the 5000 yuan limit; if the trust score is 60 points (low trust), the target permission can be adjusted to ≤ 3000 yuan, etc.

[0122] Dynamic permission verification can be understood as the process of verifying whether the target task and parameters recorded in the task execution data are within the scope of the target permission, based on the calculated target permission, thus achieving dynamic adaptation and control of permissions. For example, if the target permission is "purchase amount ≤ 3000 yuan", if the order amount in the task execution data is 2800 yuan, the authorization verification result is successful; if it is 3500 yuan, the authorization verification result is unsuccessful.

[0123] In practice, after the intelligent agent interaction gateway verifies the requesting agent's intent, it can invoke the authorization engine to initiate the authorization verification process. First, the gateway calls the authorization engine to verify the validity of the authorization credential carried by the requesting agent, checking its legality, validity period, etc. If the verification fails, the process terminates. If the verification passes, the gateway compares the permission information in the authorization credential with a preset permission policy to verify if it conforms. If the permission verification fails, the process terminates. If the permission verification passes, the gateway combines the permission information in the authorization credential with the requesting agent's trust score to calculate the target permission for this interaction. The higher the trust score, the closer the target permission is to the upper limit of permissions recorded in the authorization credential. Finally, the gateway compares the target permission with the task execution data, performing dynamic permission verification to determine if the task execution parameters are within the target permission range. If the authorization verification passes, the authorization verification is successful, and the request forwarding phase begins; if the authorization verification fails, the interaction terminates.

[0124] The data processing system provided in one or more embodiments of this specification can intercept expired or forged illegal authorization credentials by verifying the validity of authorization credentials, thus ensuring the legality of authorization delegation. By comparing the credential permissions with preset permission policies, it can ensure that the granting of permissions complies with system specifications and prevent unauthorized permissions from participating in interactions. By dynamically calculating target permissions in conjunction with trust scores, permissions are bound to the trust level of the intelligent agent. High-trust intelligent agents enjoy more lenient permissions, while low-trust intelligent agents are restricted, adapting to the uncertainty of autonomous behavior of intelligent agents. Dynamic permission verification can check the matching degree between task execution behavior and target permissions in real time to prevent unauthorized operations. The entire authorization verification process is progressive and dynamically adapted, realizing both secure delegation and scope control of user permissions, and dynamically adjusting permissions according to the trust status of intelligent agents, effectively solving security risks in authorization delegation and improving the security and flexibility of permission control in multi-agent interactions.

[0125] Specifically, the interactive execution subsystem 102 also includes an audit log module. The intelligent agent interaction gateway is further configured to generate intelligent agent interaction logs between the requesting intelligent agent and the target intelligent agent in response to the intelligent agent interaction request based on the task execution result, generate a corresponding interaction log identifier for the intelligent agent interaction logs, and send the intelligent agent interaction logs carrying the interaction log identifiers to the audit log module. The audit log module is used to extract key interaction information from the smart agent interaction log, submit the key interaction information and the interaction log identifier to the blockchain for on-chain evidence storage, and return a log recording completion notification to the smart agent interaction gateway upon receiving the evidence storage completion notification returned by the blockchain. The intelligent agent interaction gateway is also used to return the task execution result to the requesting intelligent agent.

[0126] The audit log module can be understood as a functional module in the interaction execution subsystem responsible for receiving, processing, and storing the interaction logs of intelligent agents. It is used to retain interaction records and support subsequent auditing and traceability. For example, the audit log module is specifically designed to record the interaction process and task execution results between the procurement agent and the e-commerce agent, and between the financial agent and the bank agent.

[0127] Agent interaction logs can be understood as log data generated by the agent interaction gateway based on task execution results, recording key information throughout the entire process of the interaction between the requesting agent and the target agent in response to the agent's interaction request. For example, an agent interaction log might record the following interaction log: "Purchase Agent (DID: xxx) initiates a purchase order request, the target e-commerce Agent (DID: xxx) completes the execution, the order amount is 2800 yuan, and the execution result is successful."

[0128] An interaction log identifier can be understood as a unique identifier assigned to each agent interaction log, used to distinguish different interaction records and facilitate subsequent querying, tracing, and correlation. For example, an interaction log identifier with the generation format "Log-20260402-xxx" corresponds to a specific agent procurement interaction log.

[0129] Key interaction information can be understood as the core information extracted from the agent's interaction logs for auditing and traceability, including but not limited to the interaction subject (DID of the requesting agent / target agent), interaction time, task content, task execution result, verification record, etc. For example, key interaction information could be extracted as "requesting agent DID, target agent DID, order amount, execution result, and interaction time".

[0130] On-chain evidence storage can be understood as the process by which the audit log module submits key interaction information and interaction log identifiers to the blockchain. The blockchain then stores this key interaction information and interaction log identifiers in an immutable manner, ensuring the authenticity and integrity of the interaction records. For example, on-chain evidence storage involves submitting key interaction information of a procurement interaction and its corresponding interaction log identifiers to the blockchain, where they are permanently retained and cannot be tampered with.

[0131] The notification of completion of evidence storage can be understood as the confirmation information returned to the audit log module after the blockchain has completed the storage of key interaction information and its corresponding interaction log identifier, indicating that the interaction record has been successfully stored on the blockchain.

[0132] The log recording completion notification can be understood as a confirmation message returned by the audit log module to the smart agent interaction gateway after receiving the evidence storage completion notification, indicating that the interaction log has been recorded and stored on the blockchain.

[0133] In practice, after forwarding the agent interaction request to the target agent and receiving the task execution result returned by the target agent, the agent interaction gateway generates an agent interaction log containing information about the entire interaction process based on the task execution result. A unique interaction log identifier is assigned to this log, and then the agent interaction log carrying this identifier is sent to the audit log module. Upon receiving the agent interaction log and its corresponding identifier, the audit log module extracts key interaction information such as the requesting agent's DID, the target agent's DID, the task content, verification records, and the task execution result. This key interaction information, along with its corresponding interaction log identifier, is then submitted to the blockchain for on-chain notarization. After the blockchain completes notarization and returns a notarization completion notification, the audit log module sends a log recording completion notification to the agent interaction gateway. Upon receiving this notification, the agent interaction gateway formally returns the previously received task execution result to the requesting agent, completing the entire interaction process.

[0134] The data processing system provided in one or more embodiments of this specification combines an audit log module with blockchain evidence storage to achieve full-process retention and immutability of intelligent agent interaction records. Specifically, the intelligent agent interaction log can record the entire interaction process in detail, and the interaction log identifiers facilitate quick querying and tracing; key interaction information is stored on the blockchain for evidence storage, which, relying on the characteristics of blockchain, ensures that the records are authentic, traceable, and immutable, providing a reliable basis for subsequent auditing and liability determination. Furthermore, returning the task execution result after the log recording is completed ensures the integrity of the interaction record retention, avoids log loss or tampering, effectively solves the problems of untraceable behavior and easily tampered records in multi-agent interactions, improves the interaction closed loop, strengthens the auditability of interaction behavior, and further enhances the security and credibility of multi-agent interactions.

[0135] See Figure 3 , Figure 3 This is a flowchart of a complete verification process for inter-agent interaction executed by an interactive execution subsystem in a data processing system, provided by one embodiment of this specification. The flowchart specifically includes the following steps.

[0136] Step 302: Request the Agent to send an agent interaction request to the agent interaction gateway.

[0137] Step 304: The intelligent agent interaction gateway initiates authentication.

[0138] Step 306: The intelligent agent interaction gateway authenticates the requesting agent.

[0139] Step 308: The intelligent agent interaction gateway determines the authentication result.

[0140] Step 310: If the intelligent agent interaction gateway determines that the authentication result is an authentication failure, it rejects the current interaction request and the process terminates.

[0141] Step 312: If the intelligent agent interaction gateway determines that the authentication result is successful, it initiates a trust verification.

[0142] Step 314: The intelligent agent interaction gateway requests the Agent to perform trust verification.

[0143] Step 316: The intelligent agent interaction gateway determines the trust verification result.

[0144] Step 318: If the intelligent agent interaction gateway determines that the trust verification result is a trust verification failure, it will reject the interaction request and the process will terminate.

[0145] Step 320: If the intelligent agent interaction gateway determines that the trust verification result is successful, it initiates intent verification.

[0146] Step 322: The intelligent agent interaction gateway verifies the intent of the requesting Agent.

[0147] Step 324: The intelligent agent interaction gateway determines the intent verification result.

[0148] Step 326: If the intelligent agent interaction gateway determines that the intent verification result is intent verification failure, it rejects the current interaction request and the process terminates.

[0149] Step 328: If the intelligent agent interaction gateway determines that the intent verification result is successful, it initiates authorization verification.

[0150] Step 330: The intelligent agent interaction gateway performs authorization verification on the requesting Agent.

[0151] Step 332: The intelligent agent interaction gateway determines the authorization verification result.

[0152] Step 334: If the intelligent agent interaction gateway determines that the authorization verification result is an authorization verification failure, it will reject the current interaction request and the process will terminate.

[0153] Step 336: If the authorization verification result is successful, the intelligent agent interaction gateway forwards the intelligent agent interaction request to the target Agent.

[0154] Step 338: The target agent executes the target task according to the task execution parameters in the task execution data carried in the intelligent agent interaction request, and obtains the task execution result.

[0155] Step 340: The target agent returns the task execution result to the intelligent agent interaction gateway.

[0156] Step 342: Based on the task execution result, the intelligent agent interaction gateway generates intelligent agent interaction logs for the requesting agent and the target agent regarding the intelligent agent interaction request, generates a corresponding interaction log identifier for the intelligent agent interaction logs, and sends the intelligent agent interaction logs carrying the interaction log identifiers to the audit log module.

[0157] Step 344: The audit log module extracts key interaction information from the agent's interaction log and submits the key interaction information and interaction log identifier to the blockchain for on-chain storage.

[0158] Step 346: The blockchain returns a notification of evidence storage completion to the audit log module.

[0159] Step 348: The audit log module returns a log recording completion notification to the intelligent agent interaction gateway.

[0160] Step 350: The intelligent agent interaction gateway returns the task execution result to the requesting Agent.

[0161] The data processing system provided in one or more embodiments of this specification constructs a progressive intelligent agent interaction verification system covering the entire chain of identity, trust, intent, and authorization, forming a complete security closed loop and fully adapting to the security requirements of Agent-to-Agent autonomous interaction scenarios. Specifically, through the identity verification stage, cross-domain trusted identity authentication is achieved by relying on decentralized identifiers and blockchain, intercepting identity attacks such as forgery and impersonation at the source; multi-dimensional trust assessment dynamically quantifies the trustworthiness of intelligent agents, identifies low-trust subjects in advance, and effectively addresses the uncertainty of autonomous behavior of intelligent agents; dual intent verification accurately verifies the rationality and consistency of the interaction purpose, blocking malicious and abnormal requests; dynamic authorization decision-making combines delegation credentials and trust scores to achieve secure delegation and dynamic control of permissions, preventing unauthorized operations. After the entire process verification is passed, the interaction log is simultaneously stored on the blockchain, ensuring that the interaction records are tamper-proof and fully traceable, providing a reliable basis for responsibility auditing and dispute tracing. Through layered and multi-dimensional protection, it comprehensively addresses the five core challenges in agent interaction: identity trustworthiness, cross-domain mutual trust, dynamic risk assessment, access control, and traceability. This significantly improves the security, standardization, and reliability of intelligent agent interaction, providing a solid security guarantee for agent autonomous interaction to replace the traditional human interaction mode.

[0162] Before implementing the full-process verification of the A2A interaction described above, to ensure that the participating agents have a legitimate, unique, and traceable identity, an agent identity registration mechanism must first be established. This mechanism provides a prerequisite identity basis and underlying support for subsequent identity verification, trust assessment, and authorization decisions. The specific implementation method is as follows: The data processing system further includes an identity management subsystem 104, which comprises an identity registration module, an identity generator, a trust anchor manager, and a decentralized registry module. The identity registration module is used to receive an identity registration application sent by the applicant for the requesting agent, verify the identity of the applicant based on the identity information contained in the identity registration application, and send an identity generation instruction for the requesting agent to the identity generator if the identity verification is successful. The identity generator is used to generate and return to the identity registration module a decentralized identifier and a decentralized identifier document of the requesting agent. The identity registration module is also used to send a trust endorsement request for the requesting agent to the trust anchor manager; The trust anchor manager is used to generate and return the endorsement credentials of the requesting agent to the identity registration module; The identity registration module is also used to store the decentralized identifier document and the endorsement certificate in the decentralized registry module; The decentralized registry module is used to submit the decentralized identifier document to the blockchain for on-chain notarization, and upon receiving a notarization success notification from the blockchain, it returns a notarization completion notification to the identity registration module. The identity registration module is also used to return the intelligent agent identity credentials of the requesting intelligent agent to the applicant.

[0163] The identity management subsystem can be understood as a collection of functional modules in the data processing system responsible for the full lifecycle management of intelligent agent identities. It is used for the registration, generation, endorsement, and storage of intelligent agent identities, providing an identity foundation for subsequent interactions.

[0164] The identity registration module can be understood as the entry point module of the identity management subsystem. It is responsible for receiving identity registration applications, verifying the applicant's identity, scheduling the identity generation and endorsement process, storing identity information, and returning identity credentials. For example, the identity registration module receives identity registration applications submitted by enterprises for their financial agents.

[0165] The applicant can be understood as the entity that initiates the identity registration request for the intelligent agent (such as the requesting intelligent agent mentioned above). It is generally the user, enterprise or organization to which the intelligent agent belongs. For example, the applicant could be an enterprise administrator applying for identity for a procurement agent.

[0166] An identity registration application can be understood as a request submitted by the applicant to apply for a legitimate identity for the target intelligent agent (such as the requesting intelligent agent mentioned above), which includes the applicant's identity information, the intelligent agent's basic information, etc.

[0167] An identity generator can be understood as a functional module that generates a unique decentralized identifier (DID) and a corresponding DID document for an intelligent agent, i.e., the unit for generating the identity of an intelligent agent.

[0168] The Trust Anchor Manager can be understood as a module responsible for generating trust endorsement certificates for the identity of an intelligent agent, which is used to provide a reliable guarantee for the identity of the intelligent agent and improve the credibility of the identity.

[0169] A trust endorsement request can be understood as a request initiated by the identity registration module to the trust anchor manager to apply for trust endorsement for the identity of the target intelligent agent.

[0170] An endorsement certificate can be understood as a certificate generated by a trust anchor manager to prove the legitimacy and trustworthiness of an agent's identity. For example, an endorsement certificate is an identity endorsement issued by a corporate trust anchor for its financial agent.

[0171] The decentralized registry module can be understood as a module that stores the agent's DID document and endorsement certificate and submits them to the blockchain for on-chain storage, that is, the on-chain storage unit of the agent's identity.

[0172] The agent identity credential can be understood as a credential returned by the identity registration module to the applicant, proving the agent's legitimate identity. This agent identity credential includes, but is not limited to, core identity information such as DID, DID document, and endorsement certificate, and serves as the agent's identity credential for participating in subsequent A2A interactions.

[0173] In practice, the applicant submits an identity registration application for the requesting agent to the identity registration module of the identity management subsystem. This application must include at least the applicant's identity information. The identity registration module first verifies the applicant's identity based on this information. Upon successful verification, it sends an identity generation instruction to the identity generator. The identity generator generates a decentralized identifier for the requesting agent and a corresponding decentralized identifier document, which is then returned to the identity registration module. Subsequently, the identity registration module sends a trust endorsement request to the trust anchor manager, which generates and returns an endorsement certificate. The identity registration module stores the decentralized identifier document and the endorsement certificate in the decentralized registry module, which then submits the decentralized identifier document to the blockchain for on-chain notarization. After the blockchain returns a successful notarization notification, the decentralized registry module returns a notarization completion notification to the identity registration module. Finally, the identity registration module returns the requesting agent's identity certificate to the applicant, completing the entire identity registration process.

[0174] The data processing system provided in one or more embodiments of this specification constructs a complete intelligent agent identity registration system through an identity management subsystem, providing a legitimate and reliable identity foundation for A2A interaction. The identity registration module strictly verifies the applicant's identity, ensuring the legality of the identity application from the source; the identity generator generates a unique decentralized identifier and DID document, achieving decentralized and cross-domain identifiable intelligent agent identity. The trust anchor manager generates endorsement certificates, providing a reliable guarantee for the intelligent agent's identity and enhancing its credibility. The decentralized registry module, combined with blockchain notarization, ensures that the DID document is tamper-proof and traceable, guaranteeing the authenticity and integrity of identity information. This process achieves full-process trusted management of intelligent agent identity, providing reliable support for subsequent identity verification, trust assessment, and other interactive links, solving the problem of the lack of a unified trusted identifier for intelligent agent identity, and adapting to multi-agent autonomous interaction scenarios.

[0175] Specifically, the identity generator is also used for: Based on a preset encryption algorithm, an asymmetric encryption key including a public-private key pair is generated for the requesting agent. Based on the asymmetric encryption key and the preset decentralized identifier generation rules, a decentralized identifier for the requesting agent is generated. Based on the decentralized identifier of the requesting agent, the asymmetric encryption key, and the agent attribute information of the requesting agent, construct the decentralized identifier document of the requesting agent; The decentralized identifier of the requesting agent and the decentralized identifier document of the requesting agent are returned to the identity registration module.

[0176] The preset encryption algorithm can be understood as a pre-configured encryption standard used to generate keys.

[0177] Asymmetric encryption keys can be understood as key pairs consisting of a public key and a private key. The private key is kept by the agent itself and used for signing, while the public key is made public and used for verification. For example, the private key is used to sign interaction requests, and the public key is used by the agent to verify its identity with the gateway.

[0178] The preset decentralized identifier generation rules can be understood as predefined format specifications and concatenation rules for generating DIDs. For example, the preset decentralized identifier generation rules include the prefix, hash method, encoding format, etc. for generating decentralized identifiers.

[0179] Agent attribute information can be understood as information describing the agent's own characteristics, such as agent type, affiliated entity, functional scope, and version information. For example, an agent's agent attribute information might be: purchasing agent, affiliated company XYZ, capable of performing order creation operations, etc.

[0180] In practice, upon receiving an identity generation instruction, the identity generator first generates an asymmetric encryption key pair containing a public and private key for the requesting agent based on a preset encryption algorithm. Then, following preset decentralized identifier generation rules, it generates a corresponding decentralized identifier using this asymmetric encryption key. Next, using this decentralized identifier as the core, and combining it with the public key from the asymmetric encryption key and the agent's attribute information, a complete decentralized identifier document is constructed. Finally, the identity generator returns the generated decentralized identifier and the decentralized identifier document to the identity registration module, completing the identity generation process.

[0181] The data processing system provided in one or more embodiments of this specification generates asymmetric key pairs through a preset encryption algorithm, ensuring the security of the agent's identity key and providing reliable support for subsequent signature verification and data encryption. Decentralized identifiers are generated according to unified rules, ensuring unique and cross-domain identifiable agent identities and avoiding identity conflicts. A DID document is constructed by combining the key and agent attributes, making the identity information structurally complete and its content trustworthy, facilitating rapid verification during interaction. The overall identity generation process is standardized and automated, eliminating the need for centralized control. This improves identity registration efficiency while ensuring the security and uniqueness of identity information, providing a stable and reliable foundation for identity verification and trust assessment in subsequent interactions, and effectively enhancing the security of identity management in multi-agent systems.

[0182] Specifically, the trust anchor manager is also used for: According to the preset trust policy, a corresponding trust anchor point is selected for the requesting agent; The decentralized identifier and the decentralized identifier document of the requesting agent are trusted and endorsed by the trust anchor, generating the endorsement certificate of the requesting agent, and the endorsement certificate of the requesting agent is returned to the identity registration module.

[0183] The preset trust policy can be understood as a pre-configured set of rules for selecting trust anchor points, which can match corresponding anchor points according to the type of intelligent agent, its affiliated organization, and the application scenario. For example, the preset trust policy selects enterprise qualification nodes as trust anchor points for enterprise-type intelligent agents, and regulatory service nodes as trust anchor points for government-type intelligent agents, etc.

[0184] Trust anchors can be understood as credible nodes or entities within a data processing system that possess public credibility and can provide identity guarantees, serving to legitimize the identity of intelligent agents. For example, trust anchors can be certification authorities, platform operators, enterprise management nodes, regulatory nodes, etc.

[0185] Trust endorsement can be understood as the process by which a trust anchor uses its own trusted identity to verify and sign the DID and DID document of an intelligent agent in order to prove its authenticity and validity.

[0186] In practice, upon receiving a trust endorsement request, the trust anchor manager matches and selects a suitable trust anchor for the requesting agent based on a preset trust policy. Then, through the selected trust anchor, it provides a trusted endorsement of the requesting agent's decentralized identifier and its document, generates a corresponding endorsement credential, and returns this credential to the identity registration module, thus completing the trust endorsement process.

[0187] The data processing system provided in one or more embodiments of this specification dynamically matches trust anchors for intelligent agents through a preset trust policy. It can flexibly select a trusted guarantor based on the scenario, improving endorsement adaptability. Endorsement of the intelligent agent's DID and DID document by the trusted anchor enhances the credibility and trustworthiness of the intelligent agent's identity, providing a reliable trust foundation for cross-domain interaction. The generation of endorsement credentials enables verifiable and traceable identity guarantees, effectively preventing false and forged identities from accessing the system. This mechanism does not rely on a single centralized institution, ensuring the security of identity registration and improving the overall trust level of the multi-agent system. It provides strong support for identity verification and trust assessment in subsequent interactions, further ensuring the credibility and stability of A2A interactions.

[0188] See Figure 4 , Figure 4 This is a flowchart of an intelligent agent identity registration process performed by an identity management subsystem in a data processing system, as provided in one embodiment of this specification, specifically including the following steps.

[0189] Step 402: The applicant submits an identity registration application for the target intelligent agent to the identity registration module.

[0190] Step 404: The identity registration module verifies the applicant's identity.

[0191] Step 406: Once the identity registration module confirms that the applicant's identity verification is successful, it sends an identity generation instruction to the identity generator for the target intelligent agent.

[0192] Step 408: The identity generator generates a decentralized identifier for the target intelligent agent and a decentralized identifier document.

[0193] Step 410: The identity generator returns the decentralized identifier of the target agent and the decentralized identifier document to the identity registration module.

[0194] Step 412: The identity registration module sends a trust endorsement request for the target intelligent agent to the trust anchor manager.

[0195] Step 414: The Trust Anchor Manager selects a trust anchor and generates endorsement credentials for the target agent.

[0196] Step 416: The Trust Anchor Manager returns the endorsement credentials to the Identity Registration Module.

[0197] Step 418: The identity registration module stores the decentralized identifier document and endorsement certificate to the decentralized registry module.

[0198] Step 420: The decentralized registry module submits the decentralized identifier document to the blockchain for on-chain storage.

[0199] Step 422: The blockchain returns a notification of successful evidence storage to the decentralized registry module.

[0200] Step 424: Upon receiving the certificate of authenticity completion notification from the decentralized registry module, the identity registration module returns the target smart agent's smart agent identity certificate to the applicant.

[0201] After completing the registration and trusted endorsement of the intelligent agent's identity, in order to achieve controllable and secure delegation of user permissions to the intelligent agent, a user delegation authorization process needs to be further established to provide a legal basis for permission verification and dynamic authorization in subsequent A2A interactions. The specific implementation method is as follows: The data processing system further includes a delegation and authorization subsystem 106, which includes a delegation credential manager. The delegation credential manager is used to send an authorization confirmation request to the requesting agent based on the delegation permission application containing task operation permissions sent by the requesting agent. This allows the requesting agent to display a permission authorization confirmation page to the user based on the authorization confirmation request. Upon receiving a confirmation authorization instruction and task execution constraints submitted by the user based on the permission authorization confirmation page, the manager synchronizes the confirmation authorization instruction and the task execution constraints to the delegation credential manager. The delegation permission request is sent by the requesting agent after receiving the target task sent by the user and determining the task operation permission to execute the target task based on the target task. The delegation credential manager is further configured to generate the delegation credential based on the confirmation authorization instruction and the task execution constraints, and return the delegation credential to the requesting agent.

[0202] The delegation and authorization subsystem can be understood as a set of functional modules in the data processing system responsible for managing the delegation of user permissions to intelligent agents. It is used to generate and manage delegation credentials to enable users to authorize intelligent agents in a controllable manner.

[0203] The entrustment certificate manager is responsible for processing entrustment permission applications, issuing authorization confirmations, and generating and issuing entrustment certificates.

[0204] A delegated permission request can be understood as a permission request initiated by the requesting agent to the delegated credential manager after determining the required operation permissions to execute the task based on the target task issued by the user. For example, a delegated permission request might be a purchasing agent requesting the operation permission of "placing a purchase order with an amount ≤ 5000 yuan" based on the target task of "creating a purchase order within 5000 yuan".

[0205] An authorization confirmation request can be understood as a request issued by the delegated credential manager to the requesting agent to confirm the user's authorization permissions, thereby triggering the user authorization confirmation process.

[0206] The permission authorization confirmation page can be understood as an interactive page that the requesting agent displays to the user to confirm authorization. It contains information such as the operation permissions to be authorized and the constraints of task execution, for the user to verify and confirm.

[0207] A confirmation authorization instruction can be understood as an instruction submitted by the user on the permission authorization confirmation page, agreeing to grant the requested smart agent the corresponding operation permission.

[0208] Task execution constraints can be understood as the limitations set by the user for the requesting agent to perform a target task, such as operation limits, validity periods, and applicable scenarios. For example, task execution constraints could be set as "purchase amount ≤ 5000 yuan, validity period 7 days, only for XX project procurement," etc.

[0209] A delegation credential can be understood as a credential generated by the delegation credential manager based on user authorization instructions and constraints, proving that the requesting agent has obtained legitimate permissions. It includes information such as the scope of permissions, constraints, and user signature, and serves as the basis for requesting the agent to execute a task.

[0210] In practice, after a user issues a target task to the requesting agent, the agent determines the necessary task operation permissions based on the task content and sends a delegation permission request containing these permissions to the delegation credential manager. Upon receiving the delegation permission request, the delegation credential manager sends an authorization confirmation request to the requesting agent, which then displays a permission authorization confirmation page to the user for verification of the permission scope. The user submits a confirmation instruction on this page and sets task execution constraints. The requesting agent then synchronizes the confirmation instruction and the constraints to the delegation credential manager. Based on the confirmation instruction and constraints, the delegation credential manager generates the corresponding delegation credential and returns it to the requesting agent, completing the delegation authorization process.

[0211] The data processing system provided in one or more embodiments of this specification constructs a user-controllable intelligent agent permission delegation mechanism through a delegation and authorization subsystem, realizing on-demand granting and dynamic constraints of permissions. The delegation credential manager oversees the entire authorization process, ensuring the authenticity and legality of permission grants through user confirmation, preventing unauthorized authorization. The setting of task execution constraints enables fine-grained control of permissions, adapting to the security requirements of different task scenarios. The generation of delegation credentials ensures the verifiability and traceability of permissions, providing a legal basis for authorization decisions in subsequent A2A interactions. This mechanism not only guarantees the ability of intelligent agents to autonomously execute tasks but also ensures user control over permissions, effectively solving the problem of uncontrolled intelligent agent permissions and improving the security and controllability of permissions in multi-agent interactions.

[0212] See Figure 5 , Figure 5 This is a flowchart illustrating user delegation authorization performed by a delegation and authorization subsystem in a data processing system, as provided in one embodiment of this specification. The flowchart specifically includes the following steps.

[0213] Step 502: The user issues an office supplies procurement task to the procurement agent.

[0214] Specifically, the user issues a target task to the purchasing agent (requesting agent), such as "help me purchase office supplies", which triggers the subsequent delegation and authorization process.

[0215] Step 504: Purchase the necessary permissions for the agent analysis task.

[0216] Specifically, the procurement agent analyzes and determines the operational permissions required to execute the target task issued by the user, such as "creating a purchase order" and "payment permission".

[0217] Step 506: The purchasing agent requests an authorization certificate from the authorization certificate manager.

[0218] Specifically, the procurement agent proactively initiates a delegation permission request to the delegation credential manager, requesting the operational permissions required to execute the target task, rather than passively waiting for authorization.

[0219] Step 508: The entrustment certificate manager sends an authorization confirmation request to the purchasing agent.

[0220] Specifically, after receiving the authorization request, the authorization certificate manager sends an authorization confirmation request to the procurement agent, requiring the user to be shown the list of required permissions, thus triggering the user confirmation process.

[0221] Step 510: The user confirms the authorization and sets constraints.

[0222] Specifically, the procurement agent displays a permission authorization confirmation page to the user. After verifying the permissions, the user submits a confirmation authorization command and sets task execution constraints, such as "budget ≤ 5000 yuan".

[0223] Step 512: The entrustment certificate manager generates entrustment certificates.

[0224] Specifically, the authorization certificate manager generates an authorization certificate containing the scope of permissions, constraints, and validity period based on the user's confirmation authorization instructions and task execution constraints.

[0225] Step 514: The entrustment document manager returns the entrustment document to the purchasing agent.

[0226] Step 516: The purchasing agent, carrying the authorization certificate, initiates a purchase request to the supplier agent.

[0227] Step 518: After receiving the purchase request, the supplier agent submits the authorization certificate to the authorization engine for verification.

[0228] Step 520: The authorization engine verifies the validity of the delegated credentials.

[0229] Step 522: The authorization engine checks the scope of permissions of the delegated credentials.

[0230] Step 524: The authorization engine verifies the constraints of the delegated credentials.

[0231] Step 526: Exceeding constraints triggers manual approval.

[0232] Specifically, if the purchase amount exceeds 5,000 yuan (exceeding the user-defined constraints), the authorization engine will trigger a manual approval process.

[0233] Step 528: The user approves or rejects the request.

[0234] Step 530: The authorization engine outputs the authorization decision results.

[0235] Specifically, the authorization engine outputs the final authorization decision based on the verification results and manual approval: authorization approved or authorization denied.

[0236] Step 532: If authorization is granted, the supplier agent executes the procurement.

[0237] Step 534: The supplier agent returns the order result to the purchasing agent.

[0238] Step 536: If authorization is denied, the supplier agent returns a denial of access to the purchasing agent.

[0239] Step 538: The purchasing agent returns the purchasing results to the user.

[0240] Specifically, the procurement agent will return the final procurement result (order success / rejection reason) to the user, and the entire delegation authorization and interaction process will be completed.

[0241] As mentioned above, Figure 3 Complete verification process for inter-agent interaction in China Figure 5 The user delegation and authorization process is an interconnected and collaborative whole. The user delegation and authorization process serves as the preliminary authorization preparation stage, where the user issues a delegation credential to the Agent using a secure digital signature, clearly defining the scope, constraints, and validity period of the authorization. The inter-agent interaction complete verification process is the subsequent runtime execution stage. When an Agent uses this delegation credential to interact with other Agents, the interaction gateway must perform comprehensive verification of the requester, including identity, trust level, intent, and authorization. Without a valid delegation credential generated by the user delegation and authorization process, the authorization decision-making stage of the inter-agent interaction complete verification process will fail. Conversely, the interaction verification result of the inter-agent interaction complete verification process supports the rationality of the user delegation and authorization process, forming a complete closed loop of "authorization-interaction-verification," jointly ensuring the security and standardization of agent interactions.

[0242] The difference lies in the fact that, firstly, the objects being verified are different. Figure 5 Focusing on the principal-agent relationship between the user and the agent, only the authenticity, completeness, and validity of the authorization certificate need to be verified; while Figure 3 Focusing on cross-domain interactions between two independent agents requires simultaneously verifying the other party's identity, behavioral credibility, intent legitimacy, and authorized scope of operations. Secondly, the foundations of trust differ, namely... Figure 5 The trust anchor is singular (the user is the sole source of authorization), and the credibility of the credential can be guaranteed through digital signatures, thus requiring only credential verification; while Figure 3 Involving diverse organizations, dynamic behaviors, and various potential risks, a multi-dimensional assessment system encompassing identity, trust level, intent, and authorization is needed. Furthermore, the complexity of verification varies. Figure 5 It is a relatively simple linear process that focuses on checking the signature, validity period, and scope of permissions. Figure 3 This is a complete process with multiple stages and levels, requiring sequential execution of steps such as identity verification, trust scoring, intent verification, and authorization decision-making. Failure at any stage results in rejection of interaction. In other words, the user delegation authorization process addresses the single trust issue of "whether the user allows this agent to handle the transaction on their behalf," while the complete verification process for inter-agent interaction addresses the complex trust issue of "whether two agents from different trust domains can cooperate securely."

[0243] Furthermore, after the entire A2A interaction process is verified and the intelligent agent's interaction logs are stored on the blockchain, in order to trace responsibility and restore the facts in dispute scenarios, a dispute tracing process can be initiated based on the blockchain's immutable on-chain interaction records. The specific implementation method is as follows: The data processing system further includes a dispute tracing subsystem 108, wherein, The dispute tracing subsystem 108 is used to receive a dispute tracing request sent by the arbitrator, wherein the dispute tracing request includes the interaction log identifier; Based on the interaction log identifier, the key interaction information is obtained from the blockchain, and the key interaction information is parsed to determine the requesting agent and the target agent; Based on the interaction log identifier, obtain the first local behavior log from the requesting agent and the second local behavior log from the target agent; The key interaction information, the first local behavior log, and the second local behavior log are cross-validated to generate a dispute tracing report, which is then returned to the arbitrator.

[0244] The dispute tracing subsystem can be understood as a functional module in the data processing system responsible for handling disputes and restoring the facts of interactions. It is used to receive tracing requests, retrieve on-chain records and local logs, and generate reports.

[0245] A dispute tracing request can be understood as a request initiated by an arbitrator (such as a user) to inquire about the truth of an agent's interaction in order to resolve a dispute. It includes an interaction log identifier used to locate the agent's interaction logs. For example, a company might initiate a dispute tracing request to its data processing system due to a dispute over the amount of a purchase order.

[0246] Local behavior logs can be understood as records of interaction behaviors stored locally by the requesting agent and the target agent, containing detailed request details, response data, operation sequence, and other supporting evidence. For example, the purchasing agent's local record might state "Request initiated at 10:00, carrying credential ID: Log-123," while the supplier agent's local record might state "Request received at 10:01, order created ID: OD-456."

[0247] Cross-validation can be understood as comparing and verifying the immutable key interaction information on the blockchain with the local behavior logs of two smart agents to confirm data consistency. For example, checking whether the amount on the blockchain matches the amount in the local log.

[0248] A dispute tracing report can be understood as a document generated after the dispute tracing subsystem completes cross-validation. It includes factual findings, evidence chain analysis, and liability determination conclusions, and serves as the core basis for the arbitrator to handle disputes.

[0249] In practice, after receiving a dispute tracing request containing an interaction log identifier from the arbitrator, the dispute tracing subsystem first extracts and parses the corresponding key interaction information from the blockchain based on the interaction log identifier to clarify the identity information of the requesting agent and the target agent. Then, it retrieves the first local behavior log from the requesting agent and the second local behavior log from the target agent. Finally, the dispute tracing subsystem performs multi-faceted cross-verification of the key interaction information on the chain, the requester's first local behavior log, and the target's second local behavior log, comparing data consistency to generate a dispute tracing report containing factual reconstruction and liability determination conclusions, and then sends this dispute tracing report back to the arbitrator.

[0250] The data processing system provided in one or more embodiments of this specification, through a dispute tracing subsystem, relies on the cross-verification of the immutable key interaction information of the blockchain and the local behavior logs of the intelligent agents to construct an authoritative and multi-dimensional dispute resolution mechanism. On-chain information, as core evidence, ensures the authenticity and non-repudiation of the facts, while local behavior logs provide rich supplementary details. The combination of the two enables precise tracing of the entire interaction process. The generated dispute tracing report provides the arbitrator with objective and comprehensive grounds for adjudication, effectively resolving disputes, avoiding unfair judgments due to missing or tampered evidence, significantly improving the efficiency and credibility of multi-agent interactive dispute resolution, and providing a solid guarantee for the secure operation and cross-domain trust of the data processing system.

[0251] See Figure 6 , Figure 6 This is a flowchart of a dispute tracing process performed by a dispute tracing subsystem in a data processing system, as provided in one embodiment of this specification. The flowchart specifically includes the following steps.

[0252] Step 602: The arbitrator initiates a tracing request to the audit system (dispute tracing subsystem) and provides the corresponding dispute ID (i.e., the interaction log identifier).

[0253] Step 604: The audit system initiates a query request to the blockchain based on the dispute ID to retrieve the on-chain evidence data (key interaction information) corresponding to the dispute ID.

[0254] Step 606: The blockchain returns complete on-chain evidence data to the auditing system.

[0255] Step 608: The audit system analyzes the complete interaction chain.

[0256] Specifically, the auditing system analyzes the on-chain evidence data to reconstruct the complete interaction chain, including but not limited to core information such as identity authentication records, trust scoring records, intent proof, authorization decision records, and interaction details.

[0257] Step 610: The auditing system sends a request to Agent A, the requesting party in the intelligent agent interaction, asking it to provide local behavior logs of this interaction as supplementary evidence.

[0258] Step 612: Agent A submits its own retained first local behavior log to the auditing system.

[0259] Step 614: The auditing system sends a request to Agent B, the target party in the intelligent agent interaction, requesting it to provide local behavior logs of this interaction as supplementary evidence.

[0260] Step 616: Agent B submits its own retained second local behavior log to the auditing system.

[0261] Step 618: The auditing system will cross-verify the immutable key interaction information on the blockchain with the first and second local behavior logs of Agent A and Agent B from multiple dimensions to verify data consistency, restore the truth, and finally generate a complete dispute tracing report.

[0262] Step 620: The audit system returns a complete retrospective report to the arbitrator, which includes a complete timeline of interactions, liability determination conclusions, and a complete chain of evidence.

[0263] The data processing system provided in one or more embodiments of this specification constructs a fully trusted intelligent agent interaction system, and achieves comprehensive security protection for identity, intent, permissions and traceability through multiple core innovations. First, relying on a multi-dimensional trust chain model, a four-dimensional evaluation system covering identity, behavior, reputation, and environment is established to achieve comprehensive and real-time dynamic evaluation of agents, effectively solving the problems of single and lagging trust in traditional interactions. Second, combined with an intent proof mechanism, agents submit verifiable declarations containing behavioral intent, expected results, and constraints before interaction, blocking malicious and abnormal behavior at the source and improving interaction security. Third, the agent autonomous identity identification system, through a layered DID architecture and trust anchor endorsement, enables intelligent agents to achieve independent interaction and trusted identity authentication without the user's identity, laying the foundation for cross-domain interaction. Fourth, the dynamic authorization strategy engine dynamically allocates permissions based on trust scores, intent, and context, strictly implementing the principle of least privilege and risk perception control to ensure the accuracy and security of permissions. Fifth, automated management of delegation credentials realizes the full-process delegation of permissions from agent proactive application, user confirmation, and system verification, combined with fine-grained constraints and secure storage in TokenVault (lexicon library) to ensure controllable permissions and data security. Sixth, the layered trust anchor system constructs a multi-level trust transmission and cross-domain trust mechanism, reducing dependence on a single central institution and improving the overall trust and stability of the system. The synergistic effect of these multiple innovations comprehensively enhances the security, credibility, and controllability of multi-agent interactions, providing solid support for the large-scale implementation of autonomous agent interactions.

[0264] See Figure 7 , Figure 7 This is a flowchart of a data processing method provided in one embodiment of this specification. The data processing method is applied to an intelligent agent interaction gateway and specifically includes the following steps.

[0265] Step 702: Receive the agent interaction request sent by the requesting agent.

[0266] The intelligent agent interaction request includes the decentralized identifier, intent proof, digital signature, delegation credential, and task execution data of the requesting intelligent agent. The task execution data is used to indicate the target task to be executed by the target intelligent agent and the corresponding task execution parameters.

[0267] Step 704: Verify the requesting agent based on the decentralized identifier, the intent proof, the digital signature, the delegation credential, and the task execution data. If the verification is successful, forward the agent interaction request to the target agent so that the target agent can execute the target task according to the task execution parameters and return the task execution result to the agent interaction gateway.

[0268] Step 706: Receive and return the task execution result to the requesting agent.

[0269] Specifically, the implementation of this data processing method is the same as that of the intelligent agent interaction gateway in the aforementioned data processing system, and will not be repeated here.

[0270] The data processing method provided in one or more embodiments of this specification, through an intelligent agent interaction gateway, performs unified multi-dimensional verification of intelligent agent interaction requests carrying decentralized identifiers, intent proofs, digital signatures, delegation credentials, and task execution data. It can achieve decentralized and trusted authentication of intelligent agent identities based on decentralized identifiers, and effectively resist malicious agents from forging identities by combining digital signature verification, thus solving the identity trust issue. Through strict verification of delegation credentials and task execution data, it achieves secure delegation of user permissions to intelligent agents and boundary control, preventing unauthorized execution and solving the permission delegation problem. The decentralized identifier possesses universal trust characteristics across platforms, organizations, and trust domains, breaking down trust barriers between different trust domains and enabling cross-domain communication. Efficient mutual trust between agents solves cross-domain trust issues. During the verification process, the intelligent agent interaction gateway comprehensively analyzes intent proof, permission information, and interaction data in real time for dynamic discrimination. It can accurately identify abnormal requests and unauthorized behaviors, realize dynamic risk assessment of agent autonomous behavior, and address the uncertainty risk of their behavior. At the same time, intent proof fixes the content of interaction behavior, and digital signatures ensure that the content is tamper-proof and the subject is non-repudiable, forming a credible behavior record. This provides a reliable basis for liability determination and behavior auditing, and enables traceability of interaction behavior. Thus, it comprehensively solves the problems of identity credibility, authorization delegation, cross-domain trust, dynamic risk assessment, and traceability in multi-agent autonomous interaction scenarios, significantly improving the security, standardization, and reliability of interactions between agents.

[0271] See Figure 8 , Figure 8 This is a structural block diagram of a computing device provided in one embodiment of this specification.

[0272] The computing device 800 includes, but is not limited to, a memory 810 and a processor 820. The processor 820 is connected to the memory 810 via a bus 830, and a database 850 is used to store data.

[0273] The computing device 800 also includes an access device 840, which enables the computing device 800 to communicate via one or more networks 860. Examples of these networks include Public Switched Telephone Network (PSTN), Local Area Network (LAN), Wide Area Network (WAN), Personal Area Network (PAN), or combinations of communication networks such as the Internet. The access device 840 may include one or more of any type of wired or wireless network interface (e.g., a network interface card (NIC)), such as an IEEE 802.11 Wireless Local Area Network (WLAN) wireless interface, a Wi-MAX (Worldwide Interoperability for Microwave Access) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, or a Near Field Communication (NFC) interface.

[0274] In one embodiment of this specification, the above-described components of the computing device 800 and Figure 8 Other components, not shown, can also be connected to each other, for example, via a bus. It should be understood that... Figure 8 The block diagram of the computing device shown is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0275] The computing device 800 can be any type of stationary or mobile computing device, including mobile computers or mobile computing devices (e.g., tablet computers, personal digital assistants, laptop computers, notebook computers, netbooks, etc.), mobile phones (e.g., smartphones), wearable computing devices (e.g., smartwatches, smart glasses, etc.) or other types of mobile devices, or stationary computing devices such as desktop computers or personal computers (PCs). The computing device 800 can also be a mobile or stationary server.

[0276] The processor 820 is configured to execute the following computer-executable instructions, which, when executed by the processor, implement the steps of the above-described data processing method.

[0277] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the computing device embodiments are basically similar to the data processing method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the data processing method embodiments.

[0278] An embodiment of this specification also provides a computer-readable storage medium storing a computer program / instructions that, when executed by a processor, implement the steps of the above-described data processing method.

[0279] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the computer-readable storage medium embodiments are basically similar to the data processing method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the data processing method embodiments.

[0280] An embodiment of this specification also provides a computer program product, including a computer program / instructions that, when executed by a processor, implement the steps of the above-described data processing method.

[0281] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on its differences from other embodiments. In particular, the computer program product embodiments are basically similar to the data processing method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the data processing method embodiments.

[0282] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0283] The computer instructions include computer program code, which may be in the form of source code, object code, executable file, or certain intermediate forms. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium may be appropriately added or removed according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media may not include electrical carrier signals and telecommunication signals.

[0284] It should be noted that the above description describes specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in a different order than that shown in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the drawings do not necessarily require a specific or sequential order to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous. Secondly, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to the embodiments of this specification.

[0285] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0286] The preferred embodiments disclosed above are merely illustrative of this specification. The optional embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the embodiments described herein. These embodiments are selected and specifically described in this specification to better explain the principles and practical applications of the embodiments, thereby enabling those skilled in the art to better understand and utilize this specification. This specification is limited only by the claims and their full scope and equivalents.

Claims

1. A data processing system, comprising an interactive execution subsystem, wherein the interactive execution subsystem includes an intelligent agent interactive gateway, wherein, The intelligent agent interaction gateway is used to receive intelligent agent interaction requests sent by a requesting intelligent agent. The intelligent agent interaction request includes the decentralized identifier, intent proof, digital signature, delegation credential, and task execution data of the requesting intelligent agent. The task execution data is used to indicate the target task to be executed by the target intelligent agent and the task execution parameters corresponding to the target task. The requesting agent is verified based on the decentralized identifier, the intent proof, the digital signature, the delegation credential, and the task execution data. If the verification is successful, the agent interaction request is forwarded to the target agent so that the target agent can execute the target task according to the task execution parameters and return the task execution result to the agent interaction gateway. Receive and return the task execution result to the requesting agent.

2. The data processing system according to claim 1, wherein the intelligent agent interaction gateway is further configured to: The requesting agent is authenticated based on the decentralized identifier and the digital signature to obtain the authentication result; If the authentication of the requesting agent is successful based on the authentication result, the trust score of the requesting agent on multiple evaluation dimensions is used to verify the trust level of the requesting agent and obtain the trust verification result. If the trust verification of the requesting agent is successful based on the trust verification result, the intent verification of the requesting agent is performed based on the intent proof to obtain the intent verification result. If the intent verification of the requesting agent is successful based on the intent verification result, the authorization verification of the requesting agent is performed based on the delegation credential and the task execution data to obtain the authorization verification result; If the authorization verification of the requesting agent is successful based on the authorization verification result, the agent interaction request is forwarded to the target agent.

3. The data processing system according to claim 2, wherein the intelligent agent interaction gateway is further configured to: Based on the decentralized identifier, obtain the corresponding decentralized identifier document from the blockchain; Extract the public key of the requesting agent from the decentralized identifier document, and use the public key to verify the digital signature to obtain the signature verification result; verifying the decentralized identifier according to the state information in the decentralized identifier document, obtaining an identifier verification result, wherein The status information is a status field recorded in the decentralized identifier document, used to identify the lifecycle and validity of the requesting agent's identity; The authentication result is obtained based on the signature verification result and the identifier verification result; The step of determining that the authentication of the requesting agent was successful based on the authentication result includes: If both the signature verification result and the identifier verification result are found to be successful, the identity verification of the requesting agent is determined to be successful.

4. The data processing system according to claim 2, wherein the intelligent agent interaction gateway is further configured to: Using a trust assessment engine, the requesting agent is calculated to obtain its identity trust score (identity trust score), behavior trust score (behavior trust score), reputation trust score (reputation trust score), and environment trust score (environment trust score). The trust score of the requesting agent is obtained based on the identity trust score, the behavior trust score, the reputation trust score, and the environment trust score. The trust score is compared with a preset score threshold to verify the trust level of the requesting agent and obtain the trust verification result. The step of determining that the trust verification of the requesting agent was successful based on the trust verification result includes: If the trust verification result is determined to be that the trust score is greater than or equal to the preset score threshold, the trust verification of the requesting agent is determined to be successful.

5. The data processing system according to claim 2, wherein the intelligent agent interaction gateway is further configured to: The intent proof is analyzed to obtain the intent analysis result; Based on the intent parsing result, the reasonableness of the requesting agent is verified to obtain the reasonableness verification result. Based on the intent parsing result, the intent consistency verification is performed on the requesting agent to obtain the intent consistency verification result; The intent verification result is obtained based on the intent rationality verification result and the intent consistency verification result; The step of determining that the intent verification of the requesting agent was successful based on the intent verification result includes: If both the intent rationality verification result and the intent consistency verification result are verified as passed, the intent verification of the requesting agent is determined to be successful.

6. The data processing system according to claim 4, wherein the intelligent agent interaction gateway is further configured to: The validity of the requesting agent is verified based on the authorization certificate, and the validity verification result is obtained. If the validity verification result of the credential is determined to be successful, the permission information of the authorization credential is matched with the preset permission policy to obtain the credential permission verification result. If the verification result of the credential permission is determined to be successful, the target permission of the requesting agent is calculated and obtained based on the permission information of the entrustment credential and the trust score. Based on the target permissions and the task execution data, the requesting agent is dynamically authorized to obtain the authorization verification result. The step of determining that the authorization verification of the requesting agent was successful based on the authorization verification result includes: If the authorization verification result is determined to be successful, the authorization verification of the requesting agent is determined to be successful.

7. The data processing system according to any one of claims 1-6, wherein the interactive execution subsystem further includes an audit log module. The intelligent agent interaction gateway is further configured to generate intelligent agent interaction logs between the requesting intelligent agent and the target intelligent agent in response to the intelligent agent interaction request based on the task execution result, generate a corresponding interaction log identifier for the intelligent agent interaction logs, and send the intelligent agent interaction logs carrying the interaction log identifiers to the audit log module. The audit log module is used to extract key interaction information from the smart agent interaction log, submit the key interaction information and the interaction log identifier to the blockchain for on-chain evidence storage, and return a log recording completion notification to the smart agent interaction gateway upon receiving the evidence storage completion notification returned by the blockchain. The intelligent agent interaction gateway is also used to return the task execution result to the requesting intelligent agent.

8. The data processing system according to any one of claims 1-6 further includes an identity management subsystem, wherein the identity management subsystem includes an identity registration module, an identity generator, a trust anchor manager, and a decentralized registry module, wherein, The identity registration module is used to receive an identity registration application sent by the applicant for the requesting agent, verify the identity of the applicant based on the identity information contained in the identity registration application, and send an identity generation instruction for the requesting agent to the identity generator if the identity verification is successful. The identity generator is used to generate and return to the identity registration module a decentralized identifier and a decentralized identifier document of the requesting agent. The identity registration module is also used to send a trust endorsement request for the requesting agent to the trust anchor manager; The trust anchor manager is used to generate and return the endorsement credentials of the requesting agent to the identity registration module; The identity registration module is also used to store the decentralized identifier document and the endorsement certificate in the decentralized registry module; The decentralized registry module is used to submit the decentralized identifier document to the blockchain for on-chain notarization, and upon receiving a notarization success notification from the blockchain, it returns a notarization completion notification to the identity registration module. The identity registration module is also used to return the intelligent agent identity credentials of the requesting intelligent agent to the applicant.

9. The data processing system according to claim 8, wherein the identity generator is further configured to: Based on a preset encryption algorithm, an asymmetric encryption key including a public-private key pair is generated for the requesting agent. Based on the asymmetric encryption key and the preset decentralized identifier generation rules, a decentralized identifier for the requesting agent is generated. Based on the decentralized identifier of the requesting agent, the asymmetric encryption key, and the agent attribute information of the requesting agent, construct the decentralized identifier document of the requesting agent; The decentralized identifier of the requesting agent and the decentralized identifier document of the requesting agent are returned to the identity registration module.

10. The data processing system according to claim 8, wherein the trust anchor manager is further configured to: According to the preset trust policy, a corresponding trust anchor point is selected for the requesting agent; The decentralized identifier and the decentralized identifier document of the requesting agent are trusted and endorsed by the trust anchor, generating the endorsement certificate of the requesting agent, and the endorsement certificate of the requesting agent is returned to the identity registration module.

11. The data processing system according to any one of claims 1-6, further comprising a delegation and authorization subsystem, wherein the delegation and authorization subsystem includes a delegation credential manager, wherein, The delegation credential manager is used to send an authorization confirmation request to the requesting agent based on the delegation permission application containing task operation permissions sent by the requesting agent. This allows the requesting agent to display a permission authorization confirmation page to the user based on the authorization confirmation request. Upon receiving a confirmation authorization instruction and task execution constraints submitted by the user based on the permission authorization confirmation page, the manager synchronizes the confirmation authorization instruction and the task execution constraints to the delegation credential manager. The delegation permission request is sent by the requesting agent after receiving the target task sent by the user and determining the task operation permission to execute the target task based on the target task. The delegation credential manager is further configured to generate the delegation credential based on the confirmation authorization instruction and the task execution constraints, and return the delegation credential to the requesting agent.

12. The data processing system according to claim 7 further includes a dispute tracing subsystem, wherein, The dispute tracing subsystem is used to receive a dispute tracing request sent by the arbitrator, wherein the dispute tracing request includes the interaction log identifier; Based on the interaction log identifier, the key interaction information is obtained from the blockchain, and the key interaction information is parsed to determine the requesting agent and the target agent; Based on the interaction log identifier, obtain the first local behavior log from the requesting agent and the second local behavior log from the target agent; The key interaction information, the first local behavior log, and the second local behavior log are cross-validated to generate a dispute tracing report, which is then returned to the arbitrator.

13. A data processing method applied to an intelligent agent interaction gateway, comprising: Receive an agent interaction request sent by a requesting agent, wherein the agent interaction request includes the requesting agent's decentralized identifier, intent proof, digital signature, delegation credential and task execution data, wherein the task execution data is used to indicate the target task to be executed by the target agent and the task execution parameters corresponding to the target task; The requesting agent is verified based on the decentralized identifier, the intent proof, the digital signature, the delegation credential, and the task execution data. If the verification is successful, the agent interaction request is forwarded to the target agent so that the target agent can execute the target task according to the task execution parameters and return the task execution result to the agent interaction gateway. Receive and return the task execution result to the requesting agent.

14. A computing device, comprising: Memory and processor; The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions, which, when executed by the processor, implement the steps of the data processing method of claim 13.

15. A computer-readable storage medium storing a computer program / instructions that, when executed by a processor, implement the steps of the data processing method of claim 13.

16. A computer program product comprising a computer program / instructions that, when executed by a processor, implement the steps of the data processing method of claim 13.

Images