GB/Z 185.3-2026Artificial intelligence - Agent interconnection - Part 3: Identity management (English PDF)

Foreword

GB/Z 185.3-2026: Artificial intelligence - Agent interconnection - Part 3: Identity management

GB /Z 185.3-2026.Interconnection of Intelligent Agents through Artificial Intelligence - Part

3.Identity Management GB /Z 185.3-2026 English version. Artificial intelligence - Agent interconnection - Part

3.Identity management National Standardization Guiding Technical Documents of the People's Republic of China ICS 35.100CCS L 79 AI-powered intelligent agent interconnection Part

3 Identity Management Released on 2026-05-

22 State Administration for Market Regulation The State Administration for Standardization issued a statement.

1.Scope This document specifies the framework and requirements for identity management of intelligent agents in interconnected environments, covering intelligent agent identity registration and verification, identity... Account management, credential management, identity verification, etc. This document applies to the development, application, and evaluation of intelligent agent identity management functions/systems.

4.Abbreviations The following abbreviations apply to this document. HSM. Hardware Security Module TEE. Trusted Execution Environment

5.Intelligent Agent Identity Management Framework The intelligent agent identity management framework is shown in Figure 1. The framework includes the following main roles. --The agent entrusting party. entrusts and authorizes an intelligent agent to carry out activities on its behalf in specific business scenarios, and assumes responsibility for the actions of the intelligent agent. Ultimately, the responsibility usually falls on the human user or the organization. --Intelligent Agent. Accepts authorization from the principal agent and acts on behalf of the principal agent to carry out activities, and maintains and manages its own identity information. Information, including smart agent identity code and smart agent credentials, etc. --Tool provider. When an agent initiates a protocol-based access request to the tool, the provider authenticates the agent's identity and relies on... The system provides tools and services for identity verification results and access control management for intelligent agents. It also includes a complete set of tools for intelligent agent invocation. The service can be referenced in GB /Z 185.7-2026. --Intelligent agent identity registration service provider. In identity registration and verification, it processes intelligent agent identity registration requests and executes intelligent agent identity verification. The verification process involves creating an identity account for the agent and assigning a corresponding agent identity code and issuing agent credentials upon successful verification. After completing the registration of the agent's identity, maintain and manage the agent's identity account. --Agent credential issuer. Responsible for generating, issuing, and managing agent credentials. This party can be involved in agent identity registration. The service provider can belong to the same organization, or it can be an independent third-party organization. --Agent authentication party. During the identity authentication process, it is responsible for verifying the authenticity and completeness of the agent credentials presented by the agent. And validity. During the verification process, the agent's identity verification party may interact with the agent's credential issuer to complete the verification. Validation of agent credentials. The framework includes the following main processes. --Agent identity management can be further divided into. - Identity Registration and Verification. This activity describes the process of registering the identity of an agent, that is, creating an identity for the agent. The prerequisites for subsequent intelligent agent identity management and applications are detailed in Chapter 6; - Identity and Account Management. This activity describes the process of managing the identity and account of an agent, including identity updates and identity locks. For details on locking, unlocking, and identity cancellation, please refer to Chapter 7. --Credential Management. This activity describes the process of issuing and managing agent credentials associated with an agent's identity, including credential management. This includes certificate issuance, certificate renewal, certificate locking and unlocking, and certificate cancellation. This is in addition to issuance through smart agent identity registration service providers. In addition to managing agent credentials, agents can also directly apply for and manage agent credentials from the issuer of the agent credentials. See Chapter 8. --Identity Authentication. This activity describes how an agent proves its identity to other entities, including in scenarios involving agent interactions and... Scenarios for tool access, etc. See Chapter 9.

Note. In Figure 1, when agent A initiates an interaction request to agent B, agent B verifies agent A's identity and proceeds accordingly based on the verification result. Subsequent interconnection authentication is performed to determine whether to accept the interaction request from agent A and provide corresponding services or resources.

6 Identity Registration and Verification

6.1 General Process The general process for agent identity registration and verification is as follows:

a) Identity Registration Initiation. After authorizing the agent, the agent's principal initiates the registration process with the agent's identity registration service provider. The registration request submits registration information about the agent, such as a description of the agent's functions, a list of supported tasks, and the version. Number, etc.;

b) Risk Assessment and Evidence Requirements. The agent identity registration service provider analyzes the received registration information, assesses its potential risks, and The combination of identity verification documents required for this registration will be determined based on the assessment results.

c) Identity Evidence Submission and Verification. The agent identity registration service provider collects the necessary evidence from multiple sources, including the agent's client and the agent itself. Required supporting documents, and verification and validation of the supporting documents;

d) Identity Establishment and Credential Issuance. After identity verification is successful, the agent identity registration service provider creates an identity account for the agent. The system identifies users and assigns them intelligent agent identity codes, which can be done according to GB /Z 185.2-2026.Simultaneously, it assigns identity codes to the intelligent agent credential issuer. Initiate a request to generate the corresponding agent credentials for the agent;

e) Identity Account Activation. The agent identity registration service provider returns the agent identity code and associated agent credentials to the agent. The agent's identity account will be activated at that time to complete the identity registration and verification process.

Note. For details on the implementation of identity registration, verification, and credential issuance, please refer to Appendix A.

6.2 Identity verification documents Supporting documents that can be used for agent identity verification include, but are not limited to. supporting documents related to system structure and authorization documents. Documents, supporting evidence of the boundaries of behavior, and other supporting materials.

Note. See Appendix B for specific example materials.

7 Identity Account Management

7.1 General Rules Identity account management is as follows:

a) The intelligent agent identity registration service provider should support processes such as identity registration, identity update, and identity cancellation;

b) The intelligent agent identity registration service provider should support processes such as identity locking and identity unlocking;

c) The intelligent agent identity registration service provider should formulate clear rules to define whether an identity update is triggered when an intelligent agent changes or whether it is treated as a new identity. The agent is re-registered.

7.2 Identity Update Regarding identity updates.

a) The intelligent agent identity registration service provider should support identity update processes initiated by the intelligent agent, and the process should preferably include the intelligent agent's principal. Confirmation information;

b) During the identity update process, the agent's identity code should remain unchanged;

c) Any update operation to an identity account should be subject to appropriate authorization and authentication, and corresponding log records should be generated, with technical measures in place to prevent [further issues]. Prevent unauthorized modification of logs;

d) Intelligent agents should support identity update processes initiated proactively by the intelligent agent identity registration service provider;

e) When the agent changes (e.g., changes occur in core code, function list, permission scope, runtime environment, etc.), the agent should issue a warning. Start the identity update process.

7.3 Identity Locking and Unlocking Regarding identity locking and unlocking.

a) When an identity account is locked, the agent identity registration service provider and the agent credential issuer should refuse to interact with that account. For related requests such as voucher issuance and voucher updates, the voucher should be considered invalid during verification;

b) Locking and unlocking operations on identity accounts should be subject to appropriate authorization and authentication, generate corresponding log records, and employ technical measures to prevent [illegal activities]. Prevent unauthorized modification of logs;

c) The intelligent agent identity registration service provider should support identity locking and unlocking processes initiated by the intelligent agent, and the process should include intelligent agent identification features. Confirmation information from the client who commissioned the energy source;

8 Voucher Management

8.1 General Rules Voucher management is as follows:

a) The smart agent identity registration service provider and the smart agent credential issuer should support processes such as credential issuance, credential update, and credential cancellation;

b) Intelligent agent identity registration service providers and intelligent agent credential issuers should support processes such as credential locking and credential unlocking.

8.2 Issuance of Certificates Regarding the issuance of certificates.

a) The issuer of smart agent credentials should support the credential issuance process initiated by the smart agent identity registration service provider;

b) The issuer of smart agent credentials should support the credential issuance process initiated directly by smart agents to it;

c) Smart agent credentials should preferably be digitally signed by the smart agent credential issuer to ensure their authenticity and integrity, and should preferably use compliant... The algorithm and key length of the published national or industry standards;

d) Smart agent credentials should have a clearly defined expiration date, and their lifespan should not be longer than the lifespan of the identity account to which they are associated.

8.3 Credential Update For voucher updates.

a) The issuer of smart agent credentials and the smart agent identity registration service provider should support credentials triggered by updates to smart agent identity accounts. renew;

b) The issuer of the agent credential and the agent identity registration service provider should support credential update requests initiated directly by the agent during the process. It should include confirmation information from the agent's principal;

c) Issuers of smart agent credentials and smart agent identity registration service providers should provide automated, [services] for smart agent credentials with short lifecycles. Periodic update mechanism.

8.4 Credential Locking and Unlocking Regarding locking and unlocking credentials.

a) For each lock or unlock operation, the smart agent credential issuer shall record a clear reason code for auditing and status query;

9 Identity Verification

9.1 General Process Scenarios involving identity authentication during intelligent agent interconnection may include. --Agent A initiates an interconnection request to another agent B. Agent B needs to verify the identity claimed by Agent A. The system recognizes and relies on the results of identity verification to determine whether to continue the interconnection request and to provide related services or resources. --The intelligent agent initiates a protocol-based tool access request to the tool provider. The tool provider authenticates the intelligent agent's identity and... It provides corresponding tools and services based on the results of identity verification. For entities that rely on identity authentication results to determine subsequent interconnection activities, this document uniformly uses the "agent identity dependency party" method. describe. The general process for agent identity authentication is as follows:

a) Service request and triggering of identity authentication process. The agent initiates a service interaction request to the party that depends on the agent's identity, and the agent's identity... The dependent party authenticates the agent's identity according to its access control policy.

b) Credential Presentation. Based on the requirements of the party relying on the agent's identity, the agent uses the corresponding agent credentials to generate a verifiable process. The credentials package is submitted to the agent's identity-dependent party. If the agent's identity-dependent party cannot verify independently, it can submit the credentials to the agent. The identity verification party performs the verification.

c) Credential Verification. The agent's identity verification party verifies the received process credential packet, which can be done in conjunction with the agent's credential issuance process. The parties can interact (e.g., obtain the public key of the smart agent credential issuer or query the validity of the smart agent credential). This process can... Includes one or more of the following steps. 1) Integrity and Authenticity Verification. Verifying the agent's credentials through cryptographic means (such as digital signatures) to ensure they have not been tampered with and are indeed authentic. Issued by the issuer of the smart agent certificate as declared by it; 2) Validity verification. Check whether the agent's credentials are within their validity period; 3) Status verification. Check whether the agent's credentials have been cancelled or locked by its issuer; 4) Audience and Scope Verification. Verify whether the intended recipient of the agent's credentials is the agent's current identity dependency or the agent itself. The identity verification party, and whether the scope of the smart agent credentials granted is consistent with the requested operation; 5) Delegation Chain Verification. Verifies the authenticity of the delegation and authorization from the delegator to the agent contained in the agent's credentials. efficient.

d) Authentication Assertion Generation. After credential verification is completed, the agent authentication party generates an authentication assertion containing the authentication result. Assertions. Besides identity verification results, assertions may also contain other verified and trusted information, such as information about the agent's principal. Functional scope, etc.

e) Service Authorization and Session Establishment. The agent's authentication party passes the identity authentication assertion to the agent's identity-dependent party, and the agent... The identity-dependent party makes its final decision based on the identity authentication result in this assertion and in conjunction with its own authorization strategy. If the authorization... Once the authorization is granted, an authenticated and secure session is established between the agent and the agent's identity-dependent party, and services commence. Interaction.

Note. For implementation details of identity authentication, please refer to Appendix C.

9.2 Technical Requirements To support version 9.1, the process is as follows:

a) During the presentation of the voucher. 1) The agent should be able to select agent credentials and, based on the specific requirements of the agent identity dependency party or agent identity verification party, Generate a certificate package to prove their identity; 2) Intelligent agents should have the ability to generate dynamic process credentials, such as random numbers, timestamps, or challenge values. It is packaged into the process credential using a digital signature and used for subsequent verification by the smart agent's identity verification party; 3) Intelligent agents should be able to perceive the current interaction context (such as transaction risk level, the identity of the other party entity, historical interaction records, etc.). It also dynamically adjusts the type and combination of smart agent credentials to achieve risk-adaptive credential presentation.

b) During the credential verification process. 1) The agent authentication party should be able to verify the completeness and authenticity of the presented agent credentials, such as verifying the agent credentials themselves. Digital signatures can be traced back to one or more trusted smart agent credential issuer root certificates, forming a complete chain of trust for verification. The root certificate should be issued by a nationally recognized electronic certification service provider and should comply with the requirements of the national competent authority for intelligent systems. Regulations concerning key certificate management; 2) The agent authentication provider should have the ability to check the status of agent credentials. For agent credentials that have been locked or cancelled, the provider should be able to verify their validity. The certificate should be rejected. If the status of the smart agent's certificate cannot be verified in real time, it is advisable to decide whether to accept or reject it based on a preset security policy. reject; 3) The agent authentication party should verify the matching between the recipient of the agent's credentials and the party on which the current agent's identity depends, and should also... Verify that the scope of authorization declared in the agent's credentials is consistent with the intent of the currently requested operation to prevent abuse of permissions or overreach. Domination attack; 4) The verification party should verify the authorization and delegation information from the agent's delegator to the agent contained in the agent's credentials, and ensure that the agent... The behavior of the agent is carried out within a valid authorization context.

c) In the process of generating the identification assertion. 1) The authentication assertion generated by the agent authentication party for the agent's identity-dependent party should be tamper-proof and forgery-proof, such as verification... The authentication assertion generated by the party should be digitally signed by itself to ensure the tamper-proof nature and non-determinacy of its origin. Recognition; 2) The assertion should include a clear assertion result (e.g., success, failure, further verification required, etc.), and may include the verified result. Information such as the agent's identity code, key identity attribute digest, verification timestamp, and unique identifier for this authentication is used to support... The agent relies on the party holding the agent's identity to make fine-grained authorization decisions; 3) The agent authentication mechanism should dynamically assess the risk of the interaction and adaptively adjust the authentication severity based on the risk. The level of qualification can be adjusted, such as requiring the agent to provide additional agent credentials when a high risk is identified; 4) Assertions may include security policy recommendations based on the results of this identification, for the decision-making reference of the agent's identity-dependent parties.