Internet-Draft | Workload Identity | November 2024 |
Hofmann, et al. | Expires 17 May 2025 | [Page] |
- Workgroup:
- WIMSE
- Internet-Draft:
- draft-ietf-wimse-workload-identity-bcp-02
- Published:
- Intended Status:
- Informational
- Expires:
OAuth 2.0 Client Assertion in Workload Environments
Abstract
The use of the OAuth 2.0 framework for container orchestration systems poses a challenge as managing secrets, such as client_id and client_secret, can be complex and error-prone. Instead of manual provisioning these credentials the industry has moved to a federation-based approach where credentials of the underlying workload platform are used as assertions towards an OAuth authorization server leveraging the Client Assertion Flow [RFC7521], in particular [RFC7523].¶
This specification describes a meta flow in Section 3.1, gives security recommendations in Section 4 and outlines concrete patterns in Appendix A.¶
Status of This Memo
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶
This Internet-Draft will expire on 17 May 2025.¶
Copyright Notice
Copyright (c) 2024 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
1. Introduction
Workloads often require access to external resources to perform their tasks. For example, access to a database, a web server or another workload. These resources are protected by an authorization server and can only be accessed with an access token. The challenge for workloads is to get this access token issued.¶
Traditionally, workloads were provisioned with client credentials and use the corresponding client credential flow (Section 1.3.4 [RFC6749]) to retrieve an access token. This model comes with a set of challenges that make it insecure and high-maintenance. Secret materials need to be provisioned and rotated, which often happens manually. It also can be stolen and used by attackers to impersonate the workload.¶
A solution to this problem is to not provision secret material to the workload and use the platform the workload runs on to attest for it. Many workload platforms offer a credential, in most cases a JWT token. Signed by a platform-internal authorization server, the credential attests the workload and its attributes. Based on [RFC7521] and its JWT profile [RFC7523], this credential can then be used as a client assertion towards a different authorization server.¶
2. Terminology
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
For the scope of this specification, the term authorization server is only used for the authorization server of the external authorization domain as highlighted in Figure 1.¶
Even though, technically, the platform credential is also issued by an authorization server within the workload platform, this specification only refers to it as "platform issuer" or just "platform".¶
3. OAuth Assertion in Workload Environments
3.1. Overview
Figure 1 illustrates a generic pattern that applies across all of the patterns described in Appendix A:¶
The figure outlines the following steps which are applicable in any pattern.¶
-
1) retrieve credential issued by platform. The way this is achieved and whether this is workload (pull) or platform (push) initiated differs based on the platform.¶
-
2) present credential as an assertion towards an authorization server in an external authorization domain. This step uses the assertion_grant flow defined in [RFC7521] and, in case of JWT format, [RFC7523].¶
-
3) On success, an access token is returned to the workload to access the protected resource.¶
-
4) The access token is used to access a protected resource in the external authorization domain. For instance by making a HTTP call.¶
Accessing different protected resources may require steps 2) to 4) again with different scope parameters. Accessing a protected resource in an entirely different authorization domain often requires the entire flow to be followed again, to retrieve a new platform-issued credential with an audience for the external authorization server. This, however, differs based on the platform and implementation.¶
3.2. Credential format
For the scope of this document we focus on JSON Web Token credential formats. Other formats such as X.509 certificates are possible but not as widely seen as JSON Web Tokens.¶
The claims in the present assertion vary greatly based on use case and actual platform, but a minimum set of claims are seen across all of them. [RFC7523] describes them in detail and according to it, all MUST be present.¶
{ "iss": "https://example.org", "sub": "my-workload", "aud": "target-audience", "exp": 1729248124 }¶
For the scope of this specification, the claims can be described the following. Everything from [RFC7523] applies.¶
- iss
-
The issuer of the workload platform. While this can have any format, it is important to highlight that many authorization servers leverage OpenID Connect [OIDC] and/or OAuth 2.0 Authorization Server Metadata [RFC8414] to retrieve JSON Web Keys [RFC7517] for validation purposes.¶
- sub
-
Subject which identifies the workload within the domain/workload platform instance.¶
- audience
-
One or many audiences the platform issued credential is eligible for. This is crucial when presenting the credential as an assertion towards the external authorization server which MUST identify itself as an audience present in the assertion.¶
4. Security Recommendations
All security considerations in section 8 of [RFC7521] apply.¶
4.1. Issuer, subject and audience validation
Any authorization server that validates and accepts platform-issued credentials MUST take the following claims into account before accepting assertions:¶
Failure to verify any of these properties can result in impersonation or accepting an assertion that was issued for a different purpose.¶
4.2. Token type validation
Authorization servers MUST validate the token type of the assertion. For example, OAuth Refresh or ID tokens MUST NOT be accepted.¶
4.3. Custom claims are important for context
Some platform issued credentials have custom claims that are vital for context and are required to be validated. For example in a continuous integration and deployment platform where a workload is scheduled for a GIT repository, the branch is crucial. A 'main' branch may be protected and considered trusted to federate to external authorization servers, other branches may not be and are not allowed to access protected resources.¶
Authorization servers that validate assertions SHOULD make use of these claims. Platform issuers SHOULD allow differentiation based on the subject claim alone.¶
4.4. Token lifetime
Tokens SHOULD NOT exceed the lifetime of the workloads they represent. For example, a workload that has an expected lifetime of an hour should not receive a token valid for 2 hours or more.¶
For the scope of this specification, where a platform issued credential is used to authenticate to retrieve an access token for an external authorization domain, a short-lived credential is recommended.¶
4.5. Workload lifecycle and invalidation
Platform issuers SHOULD invalidate those when the workload stops, pauses or ceases to exist. How these credentials are invalidated is not in scope of this specification.¶
4.6. Proof of possession
Credentials SHOULD be bound to workloads and proof of possession SHOULD be performed when these credentials are used. This mitigates token theft. This proof of possession applies to the platform credential and the access token of the external authorization domains.¶
5. IANA Considerations
This document does not require actions by IANA.¶
6. Acknowledgements
Add your name here.¶
7. References
7.1. Normative References
- [RFC2119]
- Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, , <https://www.rfc-editor.org/rfc/rfc2119>.
- [RFC6749]
- Hardt, D., Ed., "The OAuth 2.0 Authorization Framework", RFC 6749, DOI 10.17487/RFC6749, , <https://www.rfc-editor.org/rfc/rfc6749>.
- [RFC7517]
- Jones, M., "JSON Web Key (JWK)", RFC 7517, DOI 10.17487/RFC7517, , <https://www.rfc-editor.org/rfc/rfc7517>.
- [RFC7519]
- Jones, M., Bradley, J., and N. Sakimura, "JSON Web Token (JWT)", RFC 7519, DOI 10.17487/RFC7519, , <https://www.rfc-editor.org/rfc/rfc7519>.
- [RFC7521]
- Campbell, B., Mortimore, C., Jones, M., and Y. Goland, "Assertion Framework for OAuth 2.0 Client Authentication and Authorization Grants", RFC 7521, DOI 10.17487/RFC7521, , <https://www.rfc-editor.org/rfc/rfc7521>.
- [RFC7523]
- Jones, M., Campbell, B., and C. Mortimore, "JSON Web Token (JWT) Profile for OAuth 2.0 Client Authentication and Authorization Grants", RFC 7523, DOI 10.17487/RFC7523, , <https://www.rfc-editor.org/rfc/rfc7523>.
- [RFC8174]
- Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, , <https://www.rfc-editor.org/rfc/rfc8174>.
- [RFC8414]
- Jones, M., Sakimura, N., and J. Bradley, "OAuth 2.0 Authorization Server Metadata", RFC 8414, DOI 10.17487/RFC8414, , <https://www.rfc-editor.org/rfc/rfc8414>.
- [RFC8693]
- Jones, M., Nadalin, A., Campbell, B., Ed., Bradley, J., and C. Mortimore, "OAuth 2.0 Token Exchange", RFC 8693, DOI 10.17487/RFC8693, , <https://www.rfc-editor.org/rfc/rfc8693>.
7.2. Informative References
- [KubernetesServiceAccount]
- "Kubernetes Service Account", , <https://kubernetes.io/docs/concepts/security/service-accounts/>.
- [OIDC]
- Sakimura, N., Bradley, J., Jones, M., de Medeiros, B., and C. Mortimore, "OpenID Connect Core 1.0 incorporating errata set 1", , <https://openid.net/specs/openid-connect-core-1_0.html>.
- [TokenRequestV1]
- "Kubernetes Token Request API V1", , <https://kubernetes.io/docs/reference/kubernetes-api/authentication-resources/token-request-v1/>.
- [TokenReviewV1]
- "Kubernetes Token Review API V1", , <https://kubernetes.io/docs/reference/kubernetes-api/authentication-resources/token-review-v1/>.
Appendix A. Patterns
A.1. Kubernetes
In Kubernetes, the primary concept of machine identity is implemented through "service accounts" [KubernetesServiceAccount]. These accounts can be explicitly created or a default one is automatically assigned. Service accounts utilize JSON Web Tokens (JWTs) [RFC7519] as their credential format, with these tokens being cryptographically signed by the Kubernetes Control Plane.¶
Service accounts serve multiple authentication purposes within the Kubernetes ecosystem. They are used to authenticate to Kubernetes APIs, between different workloads and to access external resources (which is particularly relevant to the scope of this document).¶
To programatically use service accounts, workloads can:¶
-
use the Token Request API [TokenReviewV1] of the control plane¶
-
have the token projected into the file system of the workload. This is commonly referred to as "projected service accout token".¶
Both options allow workloads to:¶
-
specify a custom audience. Possible audiences can be restricted based on policy.¶
-
specify a custom lifetime. Maximum lifetime can be restricted by policy.¶
-
bind the token lifetime to an object lifecycle. This allows the token to be invalidated when the object is deleted. For example, when a Kubernetes Deployment is removed from the server. It is important to highlight, that invalidation is only in effect if the Token Review API [TokenReviewV1] of Kubernetes is used to validate the token.¶
To validate service account tokens, Kubernetes offers workloads to:¶
-
make us of the Token Review API [TokenReviewV1]. This API introspects the token, makes sure it hasn't been invalidated and returns the claims.¶
-
mount the public keys used to sign the tokens into the file system of the workload. This allows workloads to decentrally validate the tokens signature.¶
-
Optionally, a JSON Web Key Set [RFC7517] is exposed via a webserver. This allows the Service Account Token to be validated outside of the cluster and without line of sight towards the actual Kubernetes Control Plane API.¶
The steps shown in Figure 2 are:¶
-
1) The Kubernetes Control Plane schedules the workload. This is much simplified and technically happens asynchronously.¶
-
2) The Kubernetes Control Plane projects the service account token into the workload. This step is also much simplified and technically happens alongside the scheduling with step 1.¶
-
3) Workloads present the projected service account token as a client assertion towards an external authorization server according to [RFC7523].¶
-
4) On success, an access token is returned to the workload to access the protected resource.¶
-
5) The access token is used to access the protected resource in the external authorization domain.¶
As an example, the following JSON showcases the claims a Kubernetes Service Account token carries.¶
A.2. Secure Production Identity Framework For Everyone (SPIFFE)
Secure Production Identity Framework For Everyone, also known as SPIFFE, is a cloud native compute foundation (CNCF) adopted project which defines an API defined called "Workload API" to delivery machine identity to workloads. Workloads can retrieve either X509 based or JWT credentials without the need to authenticate making it very easy to use. How workloads authenticate on the API is not part of the specification. It is common to use platform metadata from the operating system and the workload platform for authentication on the Workload API.¶
For the scope of this document, the JWT formatted credential is the most relevant one. SPIFFE referres to it as "JWT-SVID" (JWT - SPIFFE Verifiable Identity Document).¶
Workloads are required to specify at least one audience when requesting a JWT-SVID from the Workload API.¶
To allow validation, SPIFFE offers¶
-
to download a set JWK encoded public keys that can be used to validate JWT signatures. In SPIFFE this is referred to as the "JWT trust bundle".¶
-
invoke a validation method on the Workload API to validate JWT-SVIDs¶
Additionally, many SPIFFE deployments choose to separately publish the signing keys as a JSON Web Key Set on a web server to allow validation where the Workload API is not available.¶
The following figure illustrates how a workload can use its JWT-SVID to access a protected resource outside of SPIFFE.¶
The steps shown in Figure 4 are:¶
-
1) The workload request a JWT-SVID from the SPIFFE Workload API with an audience that identifies the external authorization server.¶
-
2) The workload presents the JWT-SVID as a client assertion in the assertion flow based on [RFC7523].¶
-
3) On success, an access token is returned to the workload to access the protected resource.¶
-
4) The access token is used to access the protected resource in the external authorization domain.¶
The claims of a JWT-SVID for example look like this:¶
{ "aud": [ "external-authorization-server" ], "exp": 1729087175, "iat": 1729086875, "sub": "spiffe://example.org/myservice" }¶
TODO: write about "iss" in JWT-SVID.¶
A.3. Cloud Providers
Workload in cloud platforms can have any shape or form. Historically, virtual machines were the most common, with the introduction of containerization, hosted container environment or Kubernetes clusters were introduced, and lately, serverless
functions are offered. Regardless of the actual workload packaging, distribution and runtime platform, all are in need of identity.¶
To create a common identity interface across cloud services and offerings, the pattern of an Instance Metadata Endpoint
has been established by the biggest cloud providers. Next to the option for workloads to get metadata about themselves, it also allows them to receive identity. The credential types offered can vary. JWT, however, is the one that is common across all of them. The issued credential allows proof to anyone it is being presented to, that the workload platform has attested the workload and it can be considered authenticated.¶
Within a cloud provider the issued credential can often directly be used to access resources of any kind across the platform making integration between the services easy and credential less
. While the term is technically misleading, from a user perspective, no credential needs to be issued, provisioned, rotated or revoked, as everything is handled internally by the platform.¶
Resources outside of the platform, for example resources or workloads in other clouds, generic web servers or on-premise resources, are most of the time, however, protected by different domains and authorization servers and deny the platform issued credential. In this scenario, the pattern of using the platform issued credential as an assertion in the context of [RFC7521], for JWT particularly [RFC7523] towards the authorization server that protected the resource to get an access token.¶
The steps shown in Figure 5 are:¶
-
1) The workload retrieves identity from the Instance Metadata Endpoint.¶
In case the workload needs to access a resource within the cloud (protected by the same authorization server that issued the workload identity)¶
-
A1) The workload directly access the protected resource with the credential issued in step 1.¶
In case the workload needs to access a resource outside of the cloud (protected by a different authorization server). This can also be the same cloud but different context (tenant, account).¶
-
B1) The workload presents cloud-issued credential as an assertion towards the external authorization server using [RFC7523].¶
-
B2) On success, an access token is returned to the workload to access the protected resource.¶
-
B3) Using the access token, the workload is able to access the protected resource in the external authorization domain.¶
A.4. Continuoues integration/deployment systems
Continuous integration and deployment systems allow their pipelines/workflows to receive identity every time they run. Particularly in situations where build outputs need to be uploaded to resources protected by other authorization server, deployments need to be made, or more generally, protected resources to be accessed, [RFC7523] is used to federate the pipeline/workflow identity to an identity of the other authorization server.¶
The steps shown in Figure 6 are:¶
-
1) The continuous integration / deployment platform (CI-CD platform) schedules a task (considered a workload) to be performed.¶
-
2) The workload is able to retrieve identity from the CI-CD platform. This can differ based on the platform and potentially is already supplied during scheduling phase in step 1.¶
-
3) The workload presents the CI-CD issued credential as an assertion towards the authorization server in the external authorization domain based on [RFC7521]. In case of JWT also [RFC7523].¶
-
4) On success, an access token is returned to the workload to access the protected resource.¶
-
5) Using the access token, the workload is able to access the protected resource in the external authorization domain.¶
Tokens of different providers look different, but all of them contain claims carrying the basic context of the executed tasks such as source code management data (e.g. git branch), initiation and more.¶
Appendix B. Variations
B.1. Direct access to protected resources
Resource servers that protect resources may choose to trust multiple authorization servers, including the one that issues the platform identities. Instead of using the platform issued identity to receive an access token of a different authorization domain, workloads can directly use the platform issued identity to access a protected resource.¶
In this case, technically, the protected resource and workload are part of the same authorization domain.¶
B.2. Custom assertion flows
While [RFC7521] and [RFC7523] are the proposed standards for this pattern, some authorization servers use [RFC8693] or a custom API for the issuance of an access token based on an existing platform identity credentials. These pattern are not recommended and prevent interoperability.¶