Management methods and systems for restricted credentials in multi-security-domain computing instances
By adopting a credential management method that separates the primary authentication context, issues protected key issuance, and verifies policy enforcement points in multi-security domain computing scenarios, the problems of easy horizontal reuse of credentials, inaccurate revocation mechanisms, and difficult auditing are solved, thereby improving security and flexibility.
Patent Information
- Application Number
- CN202611512301.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-28
- Publication Date
- 2026-11-17
AI Technical Summary
In existing technologies, vouchers are easily reused horizontally, the revocation mechanism is inaccurate and has a wide impact, and auditing is difficult. In particular, in multi-security domain computing scenarios, voucher management has security risks and lacks management flexibility.
By separating the master authentication context from the access credentials and using a protected, independently issued key to issue instance-domain restricted credentials, which contain multi-dimensional constraint information, are rigorously verified at policy enforcement points, and maintain revocation status for different instance-domain credentials, a traceable audit chain is formed.
It effectively prevents lateral movement of credentials, improves management flexibility, ensures that credentials become invalid immediately when permissions change, provides reliable audit support, reduces security risks, and enhances user experience.
Smart Images

Figure CN123122991A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of network security technology, specifically relating to a method and system for managing restricted credentials for multi-security domain computing instances. Background Technology
[0002] In modern cloud computing, remote workstations and other centralized computing systems, single sign-on or federated identity authentication mechanisms are usually used to improve user experience. After a user successfully authenticates once, the system will issue an access token or service ticket for subsequent resource access.
[0003] However, in complex scenarios where a single user needs to access multiple compute instances simultaneously, and these instances may belong to different security domains, existing mechanisms reveal significant shortcomings. First, directly reusing a common single sign-on token across instances poses a serious security risk. If the credentials for one instance are leaked, attackers could use them to access resources on another instance or in another security domain, creating a lateral movement attack. Second, while existing token mechanisms can restrict resources, they cannot effectively prove that the use of the credential is completely consistent with the dynamic authorization relationship of "user-compute instance-security domain" currently maintained by the management platform. Furthermore, existing revocation mechanisms are typically coarse-grained, unable to precisely and independently revoke a single issued token, often resulting in the interruption of all sessions associated with the same master authentication, impacting flexibility. Additionally, relying solely on the fixed validity period of credentials cannot promptly block unauthorized access when user permissions are changed immediately. Finally, logs of key events in the credential lifecycle, such as issuance, use, and revocation, are scattered across different system components, making it difficult to link them into a unified and traceable audit evidence chain, posing significant challenges to security incident tracing and compliance review.
[0004] Therefore, existing technologies have technical shortcomings such as easy horizontal reuse of vouchers, inaccurate revocation mechanisms with a wide impact, and difficulties in auditing. A new voucher management solution is urgently needed to solve these problems. Summary of the Invention
[0005] The purpose of this application is to provide a management method and system for restricted credentials in multi-security domain computing instances, aiming to solve the technical problems of existing technologies, such as easy horizontal reuse of credentials, inaccurate revocation mechanisms with a wide impact, and difficulties in auditing.
[0006] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a method for managing restricted credentials for multi-security-domain computing instances, including: The management and control platform performs primary authentication on the user and forms a primary authentication context, which is separate from the credentials used to access computing instance resources; In response to a user's request to access a target computing instance, after verifying that the main authentication context is valid and the request conforms to the preset instance domain authorization relationship, the credential issuance component uses an issuance key that is separate from the main authentication context and held by the protected key component to issue an instance domain restricted credential. The instance domain restricted credential incorporates constraint information, including at least the target audience, the permitted scope of operation, the instance identifier of the target computing instance, the security domain identifier of the security domain to which the target computing instance belongs, the validity conditions, and the revocation control value, into a verifiable integrity relationship through a cryptographic mechanism. Before establishing resource access to the target computing instance, a policy enforcement point deployed on the resource access path verifies the integrity of the instance domain restricted credentials, verifies that the target audience, allowed operation scope, instance identifier, security domain identifier, and validity conditions match the current access request and the target computing instance, and compares the revocation control value with the current revocation status maintained by the management control platform; if any verification fails or the comparison result does not match, the resource access is denied. The management and control platform maintains the revocation status of different instance domain restricted credentials issued under the same primary authentication context. When it is necessary to revoke a certain instance domain restricted credential, only the revocation status corresponding to that instance domain restricted credential is updated, without causing other instance domain restricted credentials to become invalid automatically. The issuance, verification, use, and revocation events of the instance domain restricted credentials are recorded through a common associated identifier to form a traceable audit chain.
[0007] Secondly, this application provides a management system for restricted credentials for multi-security-domain computing instances, including: The primary authentication component is used to complete primary authentication for users and form a primary authentication context that is separate from resource access credentials; A credential issuance component communicates with a protected key component. In response to a request to access a target computing instance, the credential issuance component, after verifying the primary authentication context and the preset instance domain authorization relationship, invokes the issuance key held by the protected key component to issue an instance domain restricted credential. The instance domain restricted credential incorporates constraint information, including at least the target audience, permitted operation scope, instance identifier of the target computing instance, security domain identifier of the security domain to which the target computing instance belongs, validity conditions, and revocation control values, into a verifiable integrity relationship. The revocation control component is used to maintain the revocation status for different instance domain restricted credentials, so as to achieve independent revocation of the target credential; The policy enforcement point is used to verify the integrity of the instance domain restricted credentials before establishing resource access, verify that the target audience, allowed operation scope, instance identifier, security domain identifier, and validity conditions in the constraint information match the current access request and target computing instance, and compare the revocation control value with the current revocation status obtained from the revocation control component; if any verification fails or the comparison result does not match, resource access is denied. The associated audit component is used to associate and record events related to the issuance, verification, use, and revocation of credentials throughout their lifecycle.
[0008] According to the specific embodiments provided in this application, the following technical effects are disclosed: 1. By separating the primary authentication context from access credentials and issuing them using a protected, independently issued key, the authenticity of the credentials is guaranteed. More importantly, by forcibly binding the instance identifier and security domain identifier of the target computing instance to the instance-domain restricted credentials, and having the policy enforcement point perform a strict comparison during access, it is possible to effectively prevent the credentials from being illegally used to attack other instances or security domains after being stolen in one instance or security domain, significantly reducing the risk of lateral movement.
[0009] 2. By maintaining separate revocation statuses for restricted credentials issued under the same primary authentication context for different instance domains, independent revocation of authorization relationships for individual credentials or individual instance domains is achieved. When it is necessary to terminate a user's access to a specific computing instance, only the corresponding credential can be revoked without affecting the user's normal working sessions on other computing instances, greatly improving the system's management flexibility and user experience.
[0010] 3. By introducing a revocation control value and comparing it on each access, it ensures that even within the fixed validity period of the credential, if the corresponding authorization relationship changes, the credential will immediately become invalid due to the mismatch of the revocation control value, thus solving the access control delay problem caused by relying solely on the validity period.
[0011] 4. By recording all key events such as the issuance, verification, use, and revocation of credentials through a common associated identifier, a complete lifecycle audit chain is formed, providing reliable and consistent evidence support for subsequent security incident investigations, abnormal behavior analysis, and compliance reviews. Attached Figure Description
[0012] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0013] Figure 1 This is a schematic diagram of the architecture of a management system for calculating restricted credentials for multiple security domain instances, according to one embodiment of the present invention.
[0014] Figure 2 This is a schematic diagram illustrating the data relationship between primary authentication, authorization, and restricted credentials in one embodiment of the present invention.
[0015] Figure 3 This is a flowchart illustrating a restricted credential issuance and verification method in one embodiment of the present invention.
[0016] Figure 4 This is a schematic diagram illustrating the logical relationship between independent cancellation and cascading failure in one embodiment of the present invention.
[0017] Figure 5 This is a schematic diagram of the signaling interaction timing involving holder certification in one embodiment of the present invention. Detailed Implementation
[0018] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0019] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.
[0020] The method for managing restricted credentials for multi-security domain computing instances provided by the present invention specifically includes steps S110 to S150.
[0021] In step S110, the management control platform completes primary authentication for the user and forms a primary authentication context, which is separate from the credentials used to access computing instance resources.
[0022] In step S120, in response to a user's request to access the target computing instance, after verifying that the main authentication context is valid and the request conforms to the preset instance domain authorization relationship, the credential issuance component uses the issuance key, which is separate from the main authentication context and held by the protected key component, to issue an instance domain restricted credential. The instance domain restricted credential incorporates constraint information, including at least the target audience, the permitted scope of operation, the instance identifier of the target computing instance, the security domain identifier of the security domain to which the target computing instance belongs, and the revocation control value, into a verifiable integrity relationship through a cryptographic mechanism.
[0023] Step S130: Before establishing resource access to the target computing instance, the policy enforcement point deployed on the resource access path verifies the integrity of the instance domain restricted credentials and compares the revocation control value contained in the instance domain restricted credentials with the current revocation status maintained by the management control platform. If the comparison result does not match, resource access is rejected.
[0024] In step S140, the management control platform maintains the revocation status for different instance domain restricted credentials issued under the same primary authentication context. When it is necessary to revoke a certain instance domain restricted credential, only the revocation status corresponding to that instance domain restricted credential is updated, without automatically invalidating other instance domain restricted credentials.
[0025] Step S150: The issuance, verification, use and revocation events of instance domain restricted credentials are recorded through a common association identifier to form a traceable audit chain.
[0026] Reference Figure 1 Specifically, in the above technical solution: First, by the management and control platform (such as Figure 1 The primary authentication component (210) performs primary authentication on the user and forms a primary authentication context. This step decouples the user's initial authentication from subsequent resource access authorization. The primary authentication context serves only as the basis for authorization decisions within the management and control platform and does not have the power to directly access any computing instance resources. This design fundamentally eliminates the risk of directly exposing high-privilege primary credentials (such as login session cookies) to downstream applications, avoiding widespread security problems caused by credential theft.
[0027] Secondly, in response to a user's request to access the target computing instance, the system verifies that the primary authentication context is valid and the request conforms to the preset instance domain authorization relationship (e.g., through...). Figure 1 After the instance domain relationship management component 220 determines whether the access request conforms to the preset instance domain authorization relationship, the credential issuance component 230 issues an instance domain-restricted credential using an issuance key that is separate from the master authentication context and held by the protected key component 240. The independence and protection of the issuance key are crucial to ensuring that the credential cannot be forged. By using an independent key protected by specialized hardware such as an HSM, rather than a key derived from the user's master credential, it is ensured that even if the master credential is compromised, attackers cannot forge legitimate instance domain-restricted credentials, thus building a solid foundation of trust.
[0028] Furthermore, instance-domain restricted credentials incorporate constraint information—including at least the target audience, permitted operation scope, instance identifier of the target computing instance, security domain identifier of the security domain to which the target computing instance belongs, and revocation control value—into a verifiable integrity relationship through cryptographic mechanisms. This means that the credential itself is a "secure container," not only proving the holder's identity but, more importantly, strictly defining what the credential "can do" (permitted operation scope), "can do it to whom" (target audience), "on which instance" (instance identifier), and "within which security domain" (security domain identifier). This multi-dimensional binding mechanism solves the problem of insufficient constraints on general tokens in the background technology, strictly limiting the permissions of the credential and providing technical protection against lateral movement attacks.
[0029] In the resource access phase, before establishing resource access to the target computing instance, this invention uses a policy enforcement point deployed on the resource access path to verify the integrity of the instance domain-restricted credentials (e.g., ...). Figure 1 In this process, policy enforcement point 310 verifies the integrity of restricted credential C1, and policy point 320 verifies the integrity of restricted credential C2. It also compares the revocation control value in the constraint information contained in the restricted credential of the instance domain with the current revocation status maintained by the management control platform. This step is the core of implementing real-time access control. As a mandatory security component, the policy enforcement point not only checks the "authenticity" (integrity) of the credential but also its "timeliness" (revocation status). By comparing the revocation control value inherent in the credential with the latest status in the center, it can immediately detect and prevent the use of revoked credentials, solving the problem of revocation delay caused by relying solely on the validity period of the credential itself.
[0030] Furthermore, this invention allows the management and control platform to maintain the revocation status of restricted credentials for different instance domains issued under the same primary authentication context. When it is necessary to revoke a restricted credential for a specific instance domain, only the revocation status corresponding to that restricted credential is updated, without automatically invalidating other restricted credentials for other instance domains. This is key to achieving "independent revocation" and ensuring user experience. In scenarios where users access multiple instances simultaneously, when access to a specific instance needs to be terminated, this mechanism can precisely "target and clear" the corresponding credential without "accidentally" affecting the user's normal sessions on other instances, thus solving the problem of traditional coarse-grained revocation mechanisms uniformly handling all sessions.
[0031] Finally, this embodiment of the application records the issuance, verification, use, and revocation events of instance domain-restricted credentials through a common associated identifier to form a traceable audit chain. By assigning a unified "tag" (such as the master authentication context ID or credential ID) to each key node in the credential lifecycle (issuance, use, revocation, etc.), logs that were originally scattered across various components can be linked together to form a complete chain of evidence. This provides indispensable and reliable technical support for subsequent security investigations and compliance reviews, solving the auditing difficulties in the background technology.
[0032] In a preferred embodiment, to further enhance the security of the issuance key, the issuance key is held by at least one of a Hardware Security Module (HSM), a Key Management Service (KMS), or a Protected Execution Environment (TEE). Managing the issuance key within such specialized secure hardware or services ensures that the generation, storage, and use of the key occur within a highly trusted environment, and that the key itself does not leave the security boundary in plaintext. This approach significantly increases the difficulty of key theft, thereby establishing the trust foundation of the entire credential system on hardware-level security, effectively resisting various attacks from the software layer.
[0033] In another preferred embodiment, the revocation control value can be specifically implemented as a credential status version corresponding to the credential's unique identifier, or a relationship version corresponding to the instance domain authorization relationship. These two methods correspond to two mainstream revocation implementation strategies. Based on the "credential status version," it is equivalent to establishing an independent "blacklist" or state machine for each credential, directly changing the credential's status upon revocation, achieving the finest control. Based on the "relationship version," revocation is bound to the authorization relationship. When the authorization relationship between the user and the instance changes, the relationship version number is updated, and all credentials issued based on the old version relationship naturally become invalid. Providing these two specific implementation methods allows the present invention to flexibly adapt to different business needs and system architectures, enhancing the universality and operability of the solution.
[0034] Furthermore, the target audience can be determined by at least one of the target compute instance identifier, the instance access gateway identifier, or the service identifier within the instance. This further refines the scope of credential usage. By limiting the "audience" of a credential to a specific gateway or service, more granular access control can be achieved. For example, a credential might only be authorized to access the database service within a specific compute instance, but not the file service on that instance. This fine-grained audience limitation makes the permission model richer and more flexible, able to meet the principle of least privilege in complex application scenarios, and further tighten the attack surface.
[0035] In one alternative implementation, the scope of operations is limited to session access operations or in-session resource access operations, and the method does not include directly performing compute instance creation, shutdown, or resource configuration state changes based on instance domain-restricted credentials. This limitation clarifies the boundary of responsibility for the credentials of this invention, namely, it is an "access credential," not a "management credential." By strictly controlling its permissions at the user access level and separating it from high-privilege infrastructure management operations (such as power on / off and configuration adjustments), a separation of rights and responsibilities is achieved. This design not only conforms to security best practices but also enables the technical solution of this invention to be seamlessly integrated with existing cloud management platforms (such as OpenStack and Kubernetes), focusing on solving their core, user-level access control issues.
[0036] In a preferred embodiment, the instance-domain restricted credential is also bound to a holder-of-key public key; furthermore, the step of verifying the instance-domain restricted credential at the policy enforcement point further includes verifying a signature generated based on the private key corresponding to the holder-of-key public key. This introduces a "holder-of-key" mechanism, a powerful means of resisting credential replay attacks. Its principle is to cryptographically bind the credential to the client using it. Even if the credential is stolen during network transmission, the attacker, lacking the corresponding client private key, cannot generate a legitimate request signature and therefore cannot impersonate the credential. By adding this verification step, the present invention elevates security from "proving you know a secret (the credential itself)" to "proving you possess a secret (the private key)," significantly enhancing the overall security of the system.
[0037] The steps for updating the revocation status of a restricted credential for a specific instance domain may include: setting the unique identifier of the credential corresponding to the restricted credential to a revocation status, or updating the version of the instance domain authorization relationship corresponding to the restricted credential. This provides two specific and operable implementation paths for the core feature of "independent revocation." The former is the "blacklist" mode, which is direct and simple, suitable for precise revocation of specific credentials. The latter is the "version iteration" mode, which combines revocation with changes in authorization policies, suitable for scenarios where a batch of credentials become invalid due to permission adjustments. By clarifying these two implementation methods, the scope of protection of this invention covers different engineering implementation options, enhancing the stability and coverage of the rights.
[0038] In another preferred embodiment, the policy enforcement point can check the revocation status through at least one of the following methods: accessing the centralized status query interface provided by the management and control platform; verifying the revocation status certificate issued by the management and control platform with integrity protection and validity period; or relying on the short validity period of the instance domain-restricted credential. This provides three mainstream engineering solutions for how the policy enforcement point obtains the latest revocation status. The online query interface mode has the best real-time performance but is highly dependent on the central node; the offline proof mode (such as a variant of CRL / OCSP-stapling) reduces the pressure on the central node through signed status snapshots, balancing real-time performance and performance; the short validity period mode is a strategy that approximates revocation by frequently rotating credentials. Covering these three solutions allows the present invention to adapt to deployment scenarios with different network environments, performance requirements, and availability needs.
[0039] In one optional implementation, the method further includes: when the primary authentication context fails or a predetermined global security event occurs, invalidating the restricted credentials of multiple instance domains associated with the primary authentication context according to a cascading failure policy. This is an important supplement to the "independent revocation" scenario, constituting a complete revocation policy. Under normal circumstances, the system pursues granular "independent revocation"; however, in emergency situations, such as when user accounts are stolen or a large-scale attack is detected, the system needs to be able to perform rapid, global blocking. This "cascading failure" mechanism provides this "emergency braking" capability, making the access control policy both flexible and robust, capable of responding to various security postures.
[0040] To ensure the credibility of audit data, in a preferred embodiment, the step of linking records includes: the event production component applying a digital signature or message authentication code to the event record to form an event record with integrity protection; and the linking audit component verifying the source and integrity of the event record before it is stored. This measure extends security to the audit system itself. By signing each audit log, it is ensured that the log is not tampered with during transmission and storage, and that its source is trustworthy. This solves the problem of log forgery or deletion that may be faced by traditional log systems, enabling the audit chain generated by this invention to have a high level of non-repudiation, meeting the stringent requirements of high-level security compliance (such as in the financial and government sectors).
[0041] This invention also provides a parallel technical solution, the core of which lies in employing opaque instance-domain restricted credentials. Specifically, after the management and control platform completes primary authentication for the user and forms a primary authentication context, in response to an access request, the credential issuance component issues an opaque instance-domain restricted credential. This credential itself may simply be a randomly generated string without any business meaning. All its corresponding constraint information, such as target audience, scope of operation, instance identifier, etc., as well as its current validity status, are centrally stored on the management and control platform side.
[0042] Accordingly, during resource access, the policy enforcement point deployed on the resource access path no longer parses credentials locally. Instead, it initiates a query to the centralized verification interface provided by the management and control platform through a secure internal interface (which can be called the "introspection" interface), using the received opaque credential as a parameter. The management and control platform, based on the credential handle, retrieves the corresponding complete constraint information and validity status, and returns it to the policy enforcement point. The policy enforcement point then compares the obtained constraint information with the context of the current access request and checks whether the returned validity status is "valid." If the validity status is invalid or the comparison result does not match, the policy enforcement point will reject the resource access. This scheme achieves the strongest control over credentials and the most real-time status updates by centrally managing all states and information, but at the cost of increased dependence of the policy enforcement point on the central platform. This scheme effectively complements the aforementioned self-contained credential scheme.
[0043] Please see Figure 1 The present invention also provides a management system for restricted credentials for multi-security domain computing instances. This system serves as the hardware carrier and functional implementation of the aforementioned method. The system includes a master authentication component 210, a credential issuance component 230, a protected key component 240, a revocation control component 250, policy enforcement points 310 and 320, and an associated audit component 260. These components are typically deployed logically or physically on a management control platform and resource access paths.
[0044] Specifically, the master authentication component 210 is used to complete master authentication for the user and form a master authentication context separate from the resource access credentials. The credential issuance component 230 communicates with a protected key component 240 and, in response to a request to access a target computing instance, after verifying the master authentication context and the preset instance domain authorization relationship, invokes the issuance key held by the protected key component 240 to issue an instance domain-restricted credential. The revocation control component 250 is used to maintain revocation status for different instance domain-restricted credentials to achieve independent revocation of specific instance domain-restricted credentials. The policy enforcement point 310 is used to verify the integrity of the instance domain-restricted credential before establishing resource access and compare the revocation control value in the credential with the current revocation status obtained from the revocation control component 250. The associated audit component 260 is used to associate and record all key events in the credential lifecycle. These components work together to fully implement the technical solution proposed in this invention.
[0045] In a preferred embodiment, policy enforcement points 310 and 320 are set at at least one of the essential resource access points among the instance access gateway, host-side connection proxy, service mesh ingress, or protected ingress proxy within the instance. Deploying policy enforcement points at these "traffic arteries" is crucial to ensuring that all access requests are forcibly checked. Regardless of which ingress an attacker attempts to penetrate from, they cannot bypass the verification of the policy enforcement point, thereby guaranteeing the enforceability and coverage of the access control policy of this invention and avoiding the risk of the security policy being bypassed and rendered ineffective.
[0046] In a typical application scenario, the system of this invention is configured to manage at least two computing instances I1 and I2 associated with the same user, and to issue different instance domain-restricted credentials C1 and C2 for them respectively. When a user attempts to access instance I2 using credential C1 issued for accessing instance I1, the request will be explicitly rejected by the policy enforcement point 320 corresponding to instance I2 because the instance identifier (instance_id=I1) or security domain identifier (security_domain_id=SD1) contained in credential C1 does not match the instance identifier (I2) of the target computing instance I2 or the security domain identifier (SD2) of its security domain. This scenario vividly demonstrates how this invention effectively prevents the horizontal reuse of credentials across instances and domains through multi-dimensional credential binding and strict verification of policy enforcement points, thereby achieving its core technical effect.
[0047] The following exemplary embodiment will provide a more detailed description of the management method and system for restricted credentials of multi-security domain computing instances provided by the present invention.
[0048] In one exemplary embodiment, a complete process for the issuance, verification, and independent revocation of restricted credentials based on a self-contained format (e.g., JWT) is provided. This process is the fundamental core implementation of the present invention.
[0049] like Figure 1 As shown, the management system architecture for restricted credentials for multi-security-domain computing instances mainly consists of user terminal 110, management control platform, and policy enforcement points 310 and 320 deployed on the access paths of computing instances I1 and I2. The management control platform is the centralized control core of the entire system, responsible for centralized authentication, authorization, and auditing. Internally, it includes a master authentication component 210, an instance-domain relationship management component 220, a credential issuance component 230, a protected key component 240, a revocation control component 250, and an associated audit component 260. Computing instances I1 and I2 can belong to different security domains SD1 and SD2.
[0050] The primary authentication component 210 is responsible for authenticating users from user terminal 110, for example, through username and password, multi-factor authentication, or biometrics. Upon successful authentication, it generates a primary authentication context AC0, as shown in the attached diagram. Figure 2 As shown. This AC0 is for internal use; it proves that a user has passed through the first gate, but it is not a pass for any computing instance in itself.
[0051] The instance domain relationship management component 220 maintains a dynamic authorization policy database, recording which user has the right to access which compute instance, which security domain the instance belongs to, and the scope of access permissions, etc. For example... Figure 2 As shown, user U1 may have authorization to access both instance I1 and instance I2, which correspond to two instance domain authorization relationships B1 and B2, respectively.
[0052] The credential issuance component 230 is the unit responsible for issuing credentials. Upon receiving an access request, it first verifies the validity of the primary authentication context AC0 with the primary authentication component 210, and then queries the instance domain relationship management component 220 to determine if a corresponding authorization relationship exists. Only after both checks pass will it proceed to the next step.
[0053] A crucial step involves the credential issuance component 230 invoking the protected key component 240. This component 240 can be a hardware security module (HSM) that securely holds the issuance key KID-7 internally. The credential issuance component 230 sends the information required to bind the credential (such as user identifier, instance I1 identifier, security domain SD1 identifier, target audience, scope of operation, revocation control value, etc.) to the protected key component 240, which then completes the signing operation and returns the signed credential. Throughout this process, the issuance key KID-7 never leaves the security boundary of the protected key component 240, ensuring the reliability of the signature.
[0054] like Figure 2 As shown, based on the same primary authentication context AC0, due to different user access targets, the system will issue two different, non-interchangeable restricted credentials C1 and C2 using the same issuance key KID-7 according to different authorization relationships B1 and B2. Credential C1 is strictly bound to the information of instance I1 and security domain SD1, while credential C2 is bound to the information of instance I2 and security domain SD2.
[0055] Policy enforcement points 310 and 320 are enforcement components deployed on the resource access path. They can be instance access gateways or agents embedded in the host operating system. Their core responsibility is to forcibly intercept all requests to access the compute instance and verify the credentials carried in the requests.
[0056] The revocation control component 250 is responsible for maintaining the real-time status of all issued credentials. This can be an in-memory database (such as Redis) that records the version number of each credential or authorization relationship. When a credential needs to be revoked, simply update the corresponding status record in this component.
[0057] The associated audit component 260 acts as a recording component, faithfully recording key information at every stage from voucher issuance to use and revocation, and connecting this information through associated identifiers to form a complete chain of evidence.
[0058] Reference Figure 3 The complete workflow of this embodiment is as follows: In step S101, user U1 initiates an authentication request to the main authentication component 210 of the management and control platform through user terminal 110. After successful authentication, the system generates a main authentication context AC0.
[0059] In step S102, the user wishes to access computing instance I1 located in security domain SD1, so he sends a request to the management control platform 200 through user terminal 110 to obtain credentials to access instance I1.
[0060] In step S103, the instance domain relationship management component 220 of the management control platform 200 intervenes. It first verifies whether AC0 is valid, and then queries the database to confirm that a valid authorization relationship B1 does exist between user U1 and instance I1, and that the requested operation scope is within the authorized range.
[0061] In step S104, the revocation control component 250 provides an initial, valid revocation control value for the new credential to be issued. For example, if a relation versioning mechanism is used, it provides the current latest version number RV1 for relation B1.
[0062] In step S105, the credential issuance component 230 collects all necessary information (user ID, instance ID='I1', security domain ID='SD1', audience, scope, validity period, revocation control value=RV1, etc.), and calls the protected key component 240 to sign using the issuance key, finally generating the instance domain restricted credential C1, and returning it to the user terminal 110.
[0063] In step S106, user terminal 110 sends a resource access request to the access point of computing instance I1 carrying credential C1. This request is intercepted by policy enforcement point 310 deployed there.
[0064] In step S107, policy execution point 310 performs a series of verifications. First, it verifies the digital signature of credential C1 using a pre-set issuer public key to ensure that the credential has not been tampered with and was indeed issued by a legitimate management control platform. Next, it compares the constraint information in the credential with the context of the current request: Is the instance identifier in the credential "I1"? Is the security domain identifier "SD1"? Does the requested resource match the target audience in the credential? Is the credential valid? Most importantly, it queries the revocation control component 250 (or its cache) to obtain the latest version number of the current relationship B1 and compares it with the version number RV1 carried in credential C1. Verification is considered successful only if all checks match perfectly.
[0065] In step S108, after successful verification, policy enforcement point 310 allows the request to proceed, establishing a secure access session between the user and computing instance I1. If, during this process, the user attempts to access instance I2 with credential C1, policy enforcement point 320 of instance I2, when executing step S107, will find that the instance identifier "I1" in the credential does not match the current target "I2," and will immediately deny the access, effectively preventing lateral reuse of credentials.
[0066] In step S109, the policy execution point 310 reports the events of this verification and use (including success or failure) and related credential information to the associated audit component 260 for recording.
[0067] Through the above process, this embodiment constructs a closed-loop access control system based on strong cryptography and multi-dimensional binding, solving the core problems in the background technology.
[0068] Based on the above embodiments, in order to address the risk of credentials being illegally replayed after being eavesdropped on in the network, this embodiment provides an enhanced implementation method.
[0069] Please see Figure 5 , Figure 5The interaction flow after the introduction of the "Holder-of-Key" mechanism is shown in the form of a sequence diagram.
[0070] In this implementation, the issuance and verification process of credentials involves additional steps. Figure 5 Between steps 3 and 4, that is, after user terminal 110 requests credentials and before management control platform 200 issues credentials, user terminal 110 first generates a temporary asymmetric key pair (public key and private key). Then, when requesting credentials, the temporary public key is submitted to management control platform 200.
[0071] When issuing instance-domain restricted credential C1, the credential issuance component 230 of the management control platform 200 binds the temporary public key (or its digest) as a new field (e.g., cnf.jkt) to the payload of credential C1 and signs it together. In this way, the issued credential C1 establishes a cryptographic binding relationship with the client's temporary public key.
[0072] When user terminal 110 accesses resources with this credential C1, as shown in step 5, it can no longer simply submit C1 as before. It must perform an additional operation: using the private key corresponding to the public key bound to the credential, it must sign the key parts of the current request (such as the HTTP request method, path, host headers, and a random number or timestamp provided by the server) to generate a "holder signature". This signature will be sent to policy enforcement point 310 along with credential C1.
[0073] The verification process at policy execution point 310 also becomes more complex. After completing the basic verifications of the integrity, validity, and audience of credential C1 in step 6, and after completing the revocation status check in step 7, it executes an additional crucial step: step 9, verifying the holder's signature. It first extracts the bound holder's public key from the verified and valid credential C1, and then uses this public key to verify the "holder's signature" attached to the request.
[0074] Only when this signature can be successfully verified can policy execution point 310 confirm that the client currently initiating the request is indeed the legitimate holder of the credential C1 (because it possesses the corresponding private key). At this point, access will be finally granted (step 10). If an attacker eavesdrops on credential C1 over the network, they will be unable to generate a legitimate holder signature because they lack the corresponding private key. Therefore, even if they replay the request with the genuine credential C1, the verification in step 9 will fail. Finally, all relevant audit events, including the verification results of the holder's proof, will be reported to the associated audit component of the management control platform 200 in step 11.
[0075] By introducing a holder proof mechanism, this embodiment greatly improves the security of the system, so that even if the credential is leaked, its harm is greatly reduced, thereby effectively resisting credential replay attacks and making it suitable for scenarios with extremely high security requirements.
[0076] This embodiment also provides an alternative implementation method that runs parallel to the aforementioned self-contained credential scheme. This method is based on opaque credentials and a centralized verification mechanism. In this mode, the credential itself does not contain any business information, and its security and state management logic are significantly different from the aforementioned embodiments.
[0077] In this implementation, when a user requests access to instance I1, the credential issuance component 230 of the management control platform no longer generates a JWT containing rich information, but instead generates a high-entropy, meaningless random string, such as v1.local.AbCDefg..., which is the "opaque credential". Simultaneously, the credential issuance component 230 creates a record in a backend security database (e.g., jointly maintained by the revocation control component 250 and the instance domain relationship management component 220) that associates this random string with all relevant constraint information (user ID, instance ID='I1', security domain ID='SD1', validity period, current validity status, etc.).
[0078] Subsequently, this opaque credential is returned to user terminal 110. When user terminal 110 uses this opaque credential to access resources, policy enforcement point 310 receives this random string. Since the string itself does not contain any information, policy enforcement point 310 cannot perform any verification locally.
[0079] Therefore, policy enforcement point 310 must invoke a dedicated "Introspection" API provided by the management control platform through a secure backend channel. It will then use the received opaque credentials as parameters to initiate a query request to the management control platform.
[0080] Upon receiving a request, the introspection interface of the management control platform searches for a corresponding record in the backend database using the opaque credential as the key. If the record is not found, it indicates that the credential has been forged or expired and cleaned up. If the record is found, it checks the validity status field in the record (e.g., whether it is marked as "revoked"). Then, it returns all constraint information associated with the credential, along with the current validity status (a boolean value active: true / false), to policy execution point 310.
[0081] Upon receiving the response from the introspection interface, policy enforcement point 310 first checks the `active` field. If it is false, access is immediately denied. If it is true, the constraint information (instance ID, security domain ID, etc.) contained in the response is then compared with the context of the current request. Access is only permitted if `active` is true and all constraints match.
[0082] The advantages of this opaque credential scheme are: first, the credential itself does not reveal any internal information, thus offering higher security; second, credential revocation takes effect immediately because the policy enforcement point queries the central database for the latest status each time. Once the administrator invalidates the credential status on the management platform, subsequent access will be immediately denied. Its disadvantage is that each access requires an additional real-time communication between the policy enforcement point and the central platform, placing higher demands on the availability and performance of the central platform. This implementation provides an important and alternative technical path for this invention.
[0083] Reference Figure 4 The independent revocation and cascading failure mechanism in this embodiment is described in further detail. This is the key to achieving refined and flexible dynamic access control in this invention.
[0084] In a typical multi-tasking scenario, user U1 may obtain credentials C1 for accessing instance I1 and C2 for accessing instance I2 through a single primary authentication. These two credentials share the same primary authentication context AC0, but correspond to different instance domain authorization relationships B1 and B2, respectively.
[0085] When it is necessary to independently revoke a user's access rights to instance I1, such as when a user finishes working on I1 and logs out voluntarily, or when an administrator needs to temporarily restrict access to I1 for security reasons, the revocation control component 250 will trigger the "single credential / single instance domain relationship revocation" process. The specific operation can be one of the following two: 1. Revocation based on credential identifier: In this mode, the revocation control component 250 adds the unique identifier of credential C1 to a "revocation list," which can be a table in the database or a collection in memory. When the policy execution point 310 next receives C1 and queries its status, it will find that the unique identifier of credential C1 is on the revocation list, thus denying access.
[0086] 2. Revocation based on relationship version: If this mode is used, the revocation control component 250 will find the instance domain authorization relationship B1 associated with credential C1 and update its version number from RV1 to RV1+1. Since the version number bound in credential C1 is still the old RV1, when the policy execution point 310 receives C1 again and compares its version number with the latest version number RV1+1 obtained from the center, it will find an inconsistency and thus deny access.
[0087] Importantly, in any of the above operations, all states related to instance I2 and its authorization relationship B2 remain unchanged. Therefore, the user's credentials C2 remain fully valid, allowing seamless continued access to instance I2. This achieves precise and independent revocation of access rights to a single instance, avoiding the poor user experience of all sessions being interrupted due to the revocation of one session, as is common in traditional solutions.
[0088] The counterpart to "independent revocation" is the "cascading failure" mechanism. When a higher-level security event occurs, such as the primary authentication context AC0 itself expiring or being actively revoked (e.g., a user changing their password or a session timeout after a long period of inactivity), or an administrator triggering a global lock on user U1 (e.g., the account being disabled), or the system detecting a major security threat that may endanger the entire platform, the system will activate the cascading failure policy.
[0089] At this point, the management control platform broadcasts an event or updates a global state, indicating that all credentials associated with the primary authentication context AC0 have expired. When verifying credentials, policy enforcement points 310 and 320 check not only the credential's own state but also the global state of its associated primary authentication context AC0. Once AC0 is found to be expired, revocation control values for C1 and C2 will be rejected regardless of their validity. This mechanism ensures that in emergency situations, all access paths under a user or session source can be quickly and thoroughly cut off, achieving a comprehensive, "bottom-up" global blockade.
[0090] All these reversal events, whether independent or cascading, will be recorded in detail by the associated audit component 260, including the reason for reversal, time, operator, and scope of impact, providing a basis for post-event analysis.
[0091] The technical solution provided by this invention has a wide range of applications, especially suitable for large organizations and complex IT environments with high requirements for security, compliance, and management flexibility. For example, in the financial industry, developers, testers, and operations personnel in banks or securities companies need to access development, testing, and production environments separately. These environments are network isolated and have different security levels. Using this invention, different instance domain restricted credentials can be issued for the same employee after a single login, targeting different environments. Credentials for accessing the production database are strictly bound to the ID of the production database instance and the ID of the production security domain, and may require the use of holder authentication and a very short validity period. Credentials for accessing the development environment, on the other hand, have relatively looser restrictions. In this way, even if the credentials for the development environment are inadvertently leaked, they cannot be used to access the production environment, effectively isolating risks. When the employee leaves the project team, the administrator only needs to update their authorization relationship on the management control platform, and all credentials related to that project under their name will automatically become invalid, achieving rapid and accurate revocation of permissions.
[0092] In government and military sectors, access control for classified information is of paramount importance. A user with access to classified information may need to access multiple business systems (computing instances) with different security levels simultaneously. This invention can issue restricted credentials matching the security level of each business system. The issuance and verification process for credentials accessing high-security systems can enforce stricter policies, such as requiring access from physically isolated, authenticated user terminals. Simultaneously, all access actions, including successful, failed, and denied accesses, are immutably recorded in the associated audit component, forming a complete behavioral trail. When a security incident requires tracing, the audit system can clearly answer "who, when, with what credentials, accessed what resources, and performed what operations," providing reliable evidence for incident investigation and liability determination.
[0093] In large enterprises, with the widespread adoption of remote work and hybrid cloud, employees need to access various applications distributed across private and public clouds from different locations and using different devices. This invention can serve as a unified access control hub. Regardless of whether the computing instance is located in the enterprise's self-built data center or hosted in a public cloud vendor's VPC, it can be brought under unified management by deploying policy enforcement points along the access path. The management control platform can integrate with the enterprise's existing identity directory (such as Active Directory) to achieve unified identity management. Through this invention, enterprises can formulate granular and dynamic access policies for employees accessing different cloud resources, ensuring both business flexibility and meeting the enterprise's stringent security and compliance requirements.
[0094] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0095] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.
Claims
1. A method for managing restricted credentials for multi-security-domain computation instances, characterized in that, include: The management and control platform performs primary authentication on the user and forms a primary authentication context, which is separate from the credentials used to access computing instance resources; In response to a user's request to access a target computing instance, after verifying that the main authentication context is valid and the request conforms to the preset instance domain authorization relationship, the credential issuance component uses an issuance key that is separate from the main authentication context and held by the protected key component to issue an instance domain restricted credential. The instance domain restricted credential incorporates constraint information, including at least the target audience, the permitted scope of operation, the instance identifier of the target computing instance, the security domain identifier of the security domain to which the target computing instance belongs, the validity conditions, and the revocation control value, into a verifiable integrity relationship through a cryptographic mechanism. Before establishing resource access to the target computing instance, a policy enforcement point deployed on the resource access path verifies the integrity of the instance domain restricted credentials, verifies that the target audience, allowed operation scope, instance identifier, security domain identifier, and validity conditions match the current access request and the target computing instance, and compares the revocation control value with the current revocation status maintained by the management control platform; if any verification fails or the comparison result does not match, the resource access is denied. The management and control platform maintains the revocation status of different instance domain restricted credentials issued under the same primary authentication context. When it is necessary to revoke a certain instance domain restricted credential, only the revocation status corresponding to that instance domain restricted credential is updated, without causing other instance domain restricted credentials to become invalid automatically. The issuance, verification, use, and revocation events of the instance domain restricted credentials are recorded through a common associated identifier to form a traceable audit chain.
2. The method for managing restricted credentials for multi-security-domain computation instances according to claim 1, characterized in that, The issuance key is held by at least one of a hardware security module, a key management service, or a protected execution environment.
3. The method for managing restricted credentials for multi-security-domain computation instances according to claim 1 or 2, characterized in that, The revocation control value includes: the credential status version corresponding to the credential unique identifier, or the relationship version corresponding to the instance domain authorization relationship.
4. The method for managing restricted credentials for multi-security-domain computation instances according to claim 1 or 2, characterized in that, The target audience is determined by at least one of the target computing instance identifier, instance access gateway identifier, or in-instance service identifier.
5. The method for managing restricted credentials for multi-security-domain computation instances according to claim 1 or 2, characterized in that, The permitted operation scope is limited to session access operations or in-session resource access operations, and the method does not include directly performing compute instance creation, stopping, or resource configuration status changes based on the instance domain restricted credentials.
6. The method for managing restricted credentials for multi-security-domain computation instances according to claim 1 or 2, characterized in that, The instance domain restricted credential is also bound to the holder's proof public key; and the step of the policy execution point verifying the instance domain restricted credential further includes: verifying the signature generated based on the private key corresponding to the holder's proof public key.
7. The method for managing restricted credentials for multi-security-domain computation instances according to claim 1 or 2, characterized in that, The steps for updating the revocation status corresponding to the restricted credential of the target instance domain specifically include: setting the unique identifier of the credential corresponding to the restricted credential of the target instance domain to the revocation status, or updating the version of the instance domain authorization relationship corresponding to the restricted credential of the target instance domain.
8. The method for managing restricted credentials for multi-security-domain computation instances according to claim 1, characterized in that, The policy execution point checks the revocation status through at least one of the following methods: Access the centralized status query interface provided by the management and control platform; Verify the revocation status certificate issued by the management and control platform, which has integrity protection and a validity period; This depends on the short validity period of the instance domain-restricted credentials.
9. The method for managing restricted credentials for multi-security-domain computation instances according to any one of claims 1, 2, or 8, characterized in that, Also includes: When the primary authentication context fails or a predetermined global security event occurs, the restricted credentials of multiple instance domains associated with the primary authentication context are invalidated according to the cascading failure strategy.
10. A management system for multi-security-domain computation of restricted credentials, characterized in that, include: The primary authentication component is used to complete primary authentication for users and form a primary authentication context that is separate from resource access credentials; A credential issuance component communicates with a protected key component. In response to a request to access a target computing instance, the credential issuance component, after verifying the primary authentication context and the preset instance domain authorization relationship, invokes the issuance key held by the protected key component to issue an instance domain restricted credential. The instance domain restricted credential incorporates constraint information, including at least the target audience, permitted operation scope, instance identifier of the target computing instance, security domain identifier of the security domain to which the target computing instance belongs, validity conditions, and revocation control values, into a verifiable integrity relationship. The revocation control component is used to maintain the revocation status for different instance domain restricted credentials, so as to achieve independent revocation of the target credential; The policy enforcement point is used to verify the integrity of the instance domain restricted credentials before establishing resource access, verify that the target audience, allowed operation scope, instance identifier, security domain identifier, and validity conditions in the constraint information match the current access request and target computing instance, and compare the revocation control value with the current revocation status obtained from the revocation control component; if any verification fails or the comparison result does not match, resource access is denied. The associated audit component is used to associate and record events related to the issuance, verification, use, and revocation of credentials throughout their lifecycle.
