Internet-Draft | WIMSE Credential Exchange | February 2025 |
Schwenkschuster | Expires 8 August 2025 | [Page] |
WIMSE defines Workload Identity and its representation through credentials. Typically, a credential is provisioned to the workload, allowing it to represent itself. The credential format is usually chosen by the platform. Common formats are JSON Web Tokens or X.509 certificates. However, workloads often encounter situations where a different identity or credential is required.¶
This document describes various situations where a workload requires another credential. It also outlines different ways this can be acchieved and compares them.¶
This note is to be removed before publishing as an RFC.¶
The latest revision of this draft can be found at https://ietf-wg-wimse.github.io/draft-ietf-wimse-s2s-protocol/draft-ietf-wimse-s2s-protocol.html. Status information for this document may be found at https://datatracker.ietf.org/doc/draft-schwenkschuster-wimse-credential-exchange/.¶
Discussion of this document takes place on the Workload Identity in Multi System Environments Working Group mailing list (mailto:wimse@ietf.org), which is archived at https://mailarchive.ietf.org/arch/browse/wimse/. Subscribe at https://www.ietf.org/mailman/listinfo/wimse/.¶
Source for this draft and an issue tracker can be found at https://github.com/arndt-s.¶
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 8 August 2025.¶
Copyright (c) 2025 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.¶
Workload Identity credentials come in all shapes and forms. JSON Web Tokens are popular but also X.509 certificates are commonly used. When a workload is provisioned it can be assumed that it gets all of the following¶
an identity in the form of an identifier.¶
one or multiple credentials that allow the workload to represent itself (as that identity). Multiple credentials are often different types but representing the same identity.¶
an indication of trust domain. (TODO not sure)¶
Identity, credential and trust domain enable the workload to interact within its environment, communicate to sibling workloads (same trust domain), access APIs inside that trust domain or provide an API itself.¶
Why a workload cannot use its existing credential can have lots of reasons and subsequent need. The following list highlights the most common ones and is certainly not complete.¶
Workloads may require a different format representing the same identity in the same trust domain. Some concrete examples are:¶
initial credential was an X.509 certificate but infrastructure requires application-level authentication such as JWT or Workload Identity Tokens as defined in (TODO).¶
initial credential was a JWT bound to a key, requiring to be presented along with proof of possession but the peer does not support it and requires a Bearer credential.¶
Credential format is dificult to define. Some formats are opague to the workload and should remain that way. For instance how an OAuth Bearer token is construct and whether it carries claims or not is not a concern of the workload. That a Bearer token is required, however, is. Hence, for example, a change in format between a Bearer token and an X.509 certificate is certainly a change in format the workload can require. A different encoding of a Bearer token on the other hand is not and this specification is not meant for this.¶
A credential in the same format representing the same identity but scoped differently. Examples are:¶
A JWT credential audienced to interact with the Workload platform but access to other workloads are required. The workload is in need of JWTs with different, dedicated audiences.¶
A X.509 credential allowing the workload to TODO based on X.509 Key Usage extension but the workload requires additonal usages.¶
Generally, scope should already be present and configured approperately with the workload platform only issuing narrowly scoped credentials to the workload.¶
In some situation the platform may only support the provisioning of a single credential and not support scoping it. If those cannot be requested by the platform itself an exchange may be necessary.¶
A workload may be known under multiple identities. For example:¶
A workload identity representing an exact physical instance may be eligable for a workload identity representing a logical unit that consists of many phyiscal instances. Another example is a workload running in a specific region being eligable for a more broader, geo scoped identity.¶
A workload that can act on behalf of other workloads. These workloads often are part of infrastructure such as API-Gateways, proxies or service meshes in container environments.¶
A provisioned workload identity is often part of a trust domain that is coupled to infrastructure or deployment. Workloads often require to intercept with other workloads or access outside resources located in other trust domains or reside in different trust domains. This requires the client workload to retrieve an identity of the other trust domain. Examples here include:¶
Federation (a workload identity federates to a identity in a different trust domain). In existing workload identity environment OAuth2 with Token Exchange (TODO) and Assertion framework (TODO) are popular.¶
A workload requires a credential of "higher trust" to interact with other workloads. This "higher trust" is facilitated by another trust domain. For instance a workload requiring a WebPKI certificate to offer a service to the world wide web.¶
A workload platform may not support the provisioning of credentials required by the workload. Technically, any of these would likely fall under the reasons above but it's a very common reason and often falls into multiple categories. As an example:¶
Workload platform provisions identity & credential in the form of a simple signed document that carries the attributes attested by the platform but gives not access in any way.¶
Reasons and needs to exchange credentials are often not binary. A change in trust domain effectively is a change in identity too. A change in format can require a change in trust domain because formats come with different trust structures and security promises. E.g. a trust domain issuing JSON Web Tokens may not be able to issue WebPKI certificates.¶
Once a workload requires a different credential it has multiple options to retrieve it. Based on the need the list of options shrinks. We differientate between the following 3 main options:¶
The initial credential(s) the workload is provisioned with are modified. This is often a matter of configuration within the platform. For instance a service account token in Kubernetes is projected with a different audience. Or a new credential is provisioned alongside the existing credential.¶
Workloads are able to request credentials without authentication. This can be in addition to an initial credential or the way the initial credential is delivered in the first place. This approach allows the workload to request credential in the exact format, scope and identity it requires.¶
Workload use a provisioned credential to authenticate a delivery of a different credential. The significant difference here is that this is an authenticated action, compared to the other above, which are unauthenticated.¶
If a workload is in need of a different format, identity or scope it is always reccomended to use the initial provisioning or re-provisioning instead of a credential exchange. These approaches have generally a higher assurance and are more secure than exchange an already provisioned credential.¶
The following table gives some guidance based on the need:¶
Need | Preferred mechanism | Other options (in order) |
---|---|---|
Change in trust domain | Credential exchange | None |
Change in identity | Provisioning | 1) Re-provisioning 2) credential exchange |
Change in scope | Provisioning | 1) Re-provisioning 2) credential exchange |
Change in format | Provisioning | 1) Re-provisioning 2) credential exchange |
Existing trust & identity framework often consist of a protocol or framework to exchange credentials. Leveraging this makes use of existing adoption and specific guidelines.¶
The following bullets give an overview of the existing patterns and when to use them based on the needs given above:¶
X.509 Certificate Management Protocol [RFC4210]:¶
Is this valid here? If yes, write about it.¶
SPIFFE federation:¶
Not a credential exchange but rather allowing for identities of different trust domains to trust each other and communicate securely. For SPIFFE the credential formats are JWT and X.509.¶
Workload environments can be highly dynamic and connected with a high variety of resources protected by different identity frameworks and formats. A format-agnostic, exchange component that exchanges credentials on behalf of the workload may be desired to remain control of credential issuance. For instance to enforce policy, collect audit trails or ease management.¶
TODO: describe steps¶
The author believes that a specific protocol that fits all credential formats and trust frameworks is not feasable while remaining the existing security promises of the format or framework. It rather believes that a profile for each scenario is the best way forward and welcomes them to profile this specificiation for their concrete use cases. As a general guidance it is reccommended * to narrowly scope the scenarios and don't build a one-fits-all exchange for a specific format. * to decouple authentication and access control from the actual exchange as best as possible. E.g. a credential of one profile should be allowed as a mean of authentication to exchange to a credential of a different profile, regardless if the profiles are aware of each other or not.¶
The "Credential Exchanger" shown in the figure may be the Workload Platform itself that offers this capability. Potentially also in a "re-provisioning" way without authentication.¶
TODO: add more.¶
A credential exchange is an authenticated way to retrieve credential(s). Thus, the issued credential cannot have higher trust than the credential that was used to authenticate the request. This is particularly relevant when a credential is required which format and frameworks is of a higher trust than the one that was used to authenticate the request. This includes exchanging credentials not requiring proof of key possession to credentials carrying it.¶
Generally, these situations are not reccommended and should be avoided. Workloads SHOULD be provisioned with the credential of the highest trust and only retrieve less-trusted credentials via credential exchange.¶
Alternatively, the authentication request should be enriched with additional identification that increases the level of authentication. E.g. authentication and additional proof of platform attestation.¶
Because credential exchange is authenticated it cannot replace provisioning. Without an initial credential a workload cannot facilitate credential exchange as there's no proof the workload is eligable for the requested credential.¶
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.¶
TODO Security¶
This document has no IANA actions.¶
RFC Editor: please remove before publication.¶
Initial individual draft & write up.¶
Big shoutout to the WIMSE token exchange design team (Dean Saxe, Yaroslav Rosomakho, Andrii Deinega, Dmitry Izumskiy, Ken McCracken and George Fletcher) that have done amazing groundlaying work in this area.¶