]> Ping Identity

bcampbell@pingidentity.com

Independent

dfeldman.mn@gmail.com

Venafi

joe.salowey@gmail.com

Microsoft

arndts.ietf@gmail.com

Intuit

yaronf.ietf@gmail.com

Applications and Real-Time Workload Identity in Multi System Environments workload identity The WIMSE architecture defines authentication and authorization for software workloads in a variety of runtime environments, from the most basic ones up to complex multi-service, multi-cloud, multi-tenant deployments. This document defines the simplest, atomic unit of this architecture: the protocol between two workloads that need to verify each other's identity in order to communicate securely. The scope of this protocol is a single HTTP request-and-response pair. To address the needs of different setups, we propose two protocols, one at the application level and one that makes use of trusted TLS transport. These two protocols are compatible, in the sense that a single call chain can have some calls use one protocol and some use the other. Service A can call Service B with mutual TLS authentication, while the next call from Service B to Service C would be authenticated at the application level. About This Document The latest revision of this draft can be found at . Status information for this document may be found at . Discussion of this document takes place on the Workload Identity in Multi System Environments Working Group mailing list (), which is archived at . Subscribe at . Source for this draft and an issue tracker can be found at .

Introduction This document defines authentication and authorization in the context of interaction between two workloads. This is the core component of the WIMSE architecture . Assume that Service A needs to call Service B. For simplicity, this document focuses on HTTP-based services, and the service-to-service call consists of a single HTTP request and its response. We define the credentials that both services should possess and how they are used to protect the HTTP exchange. There are multiple deployment styles in use today, and they result in different security properties. We propose to address them differently.

• Many use cases have various middleboxes inserted between pairs of services, resulting in a transport layer that is not end-to-end encrypted. We propose to address these use cases by protecting the HTTP messages at the application level ().
• The other commonly deployed architecture has a mutual-TLS connection between each pair of services. This setup can be addressed by a simpler solution ().

It is an explicit goal of this protocol that a service deployment can include both architectures across a multi-chain call. In other words, Service A can call Service B with mutual TLS protection, while the next call to Service C is protected at the application level. For application-level protection we currently propose two alternative solutions, one inspired by DPoP in and one which is a profile of HTTP Message Signatures in . The design team believes that we need to pick one of these two alternatives for standardization, once we have understood their pros and cons.

Deployment Architecture and Message Flow Regardless of the transport between the workloads, we assume the following logical architecture: | | | | | +--------+ | | | | PEP | +------------+ +---+--------+ ^ ^ ^ | | | | +----------------------+ | | | | v v v +------------+ +------------+ | | | | | Identity | | PDP | | Server | | (optional) | | | | | +------------+ +------------+ ]]> The Identity Server provisions credentials to each of the workloads. At least Workload A (and possibly both) must be provisioned with a credential before the call can proceed. Details of communication with the Identity Server are out of scope of this document, however we do describe the credential received by the workload. PEP is a Policy Enforcement Point, the component that allows the call to go through or blocks it. PDP is an optional Policy Decision Point, which may be deployed in architectures where policy management is centralized. All details of policy management and message authorization are out of scope of this document. The high-level message flow is as follows:

• Workload A obtains a credential from the Identity Server. This happens periodically, e.g. once every 24 hours.
• Workload A makes an HTTP call into Workload B. This is a regular HTTP request, with the additional protection mechanisms defined below.
• Workload B now authenticates Workload A and decides whether to authorize the call. In certain architectures, Workload B may need to consult with an external server to decide whether to accept the call.
• Workload B returns a response to Workload A, which may be an error response or a regular one.

Conventions and Definitions This document uses "service" and "workload" interchangeably. Otherwise, all terms are as defined by . 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 when, and only when, they appear in all capitals, as shown here.

Workload Identity This document defines a workload identity as a URI . This URI is used in the subject fields in the certificates and tokens defined later in this document. This specification treats the URI as opaque. The format of the URI and the namespace for the URI are at the discretion of the deployment at large. Other specifications may define specific URI structures for particular use cases. An example of a defined identity format is the SPIFFE ID. A workload identity only has meaning within the scope of a specific issuer. Two identities of the same value issued by different issuers may or may not refer to the same workload. In order to avoid collisions identity URIs SHOULD specify, in the URI's "authority" field, the trust domain associated with an issuer that is selected from a global name space such as host domains. However, the validator of an identity credential MUST make sure that they are using the correct issuer credential to verify the identity credential and that the issuer is trusted to issue tokens for the defined trust domain.

Application Level Service To Service Authentication As noted in the Introduction, for many deployments communication between workloads cannot use end-to-end TLS. For these deployment styles, this document proposes application-level protections. The current version of the document includes two alternatives, both using the newly introduced Workload Identity Token (). The first alternative () is inspired by the OAuth DPoP specification. The second () is based on the HTTP Message Signatures RFC. We present both alternatives and expect the working group to select one of them as this document progresses towards IETF consensus.

The Workload Identity Token The Workload Identity Token (WIT) is a JWS signed JWT that represents the identity of a workload. It is issued by the Identity Server and binds a public key to the workload identity. A WIT MUST contain the following:

• in the JOSE header:
  • alg: An identifier for a JWS asymmetric digital signature algorithm (registered algorithm identifiers are listed in the IANA JOSE Algorithms registry ). The value none MUST NOT be used.
  • typ: the WIT is explicitly typed, as recommended in , using the wimse-id+jwt media type.
• in the JWT claims:
  • iss: The issuer of the token, which is the Identity Server, represented by a URI.
  • sub: The subject of the token, which is the identity of the workload, represented by a URI.
  • exp: The expiration time of the token (as defined in ). WITs should be refreshed regularly, e.g. on the order of hours.
  • jti: A unique identifier for the token.
  • cnf: A confirmation claim containing the public key of the workload using the jwk member as defined in . The workload MUST prove possession of the corresponding private key when presenting the WIT to another party, which can be accomplished by using it in conjunction with one of the methods in or . As such, it MUST NOT be used as a bearer token and is not intended for use in the Authorization header.

An example WIT might look like this (all examples, of course, are non-normative and with line breaks and extra space for readability):

An example Workload Identity Token (WIT)

The decoded JOSE header of the WIT from the example above is shown here:

Example WIT JOSE Header

The decoded JWT claims of the WIT from the example above are shown here:

Example WIT Claims

The claims indicate that the example WIT:

• was issued by an Identity Server known as wimse://example.com/trusted-central-authority.
• is valid until May 15, 2024 3:28:45 PM GMT-06:00 (represented as NumericDate value 1717612470).
• identifies the workload to which the token was issued as wimse://example.com/specific-workload.
• has a unique identifier of x-_1CTL2cca3CSE4cwb__.
• binds the public key represented by the jwk confirmation method to the workload wimse://example.com/specific-workload.

For elucidative purposes only, the workload's key, including the private part, is shown below in JWK format:

Example Workload's Key

The afore-exampled WIT is signed with the private key of the Identity Server. The public key(s) of the Identity Server need to be known to all workloads in order to verify the signature of the WIT. The Identity Server's public key from this example is shown below in JWK format:

Example Identity Server Key

The WIT HTTP Header A WIT is conveyed in an HTTP header field named Workload-Identity-Token. For those who celebrate, ABNF for the value of Workload-Identity-Token header field is provided in :

Workload-Identity-Token Header Field ABNF

The following shows the WIT from in an example of a Workload-Identity-Token header field:

An example Workload Identity Token HTTP Header Field

Note that per , header field names are case insensitive; thus, Workload-Identity-Token, workload-identity-token, WORKLOAD-IDENTITY-TOKEN, etc., are all valid and equivalent header field names. However, case is significant in the header field value.

Option 1: DPoP-Inspired Authentication This option, inspired by the OAuth DPoP specification , uses a DPoP-like mechanism to authenticate the calling workload in the context of the request. The WIMSE Identity Token () is sent in the request as described in . An additional JWT, the Workload Proof Token (WPT), is signed by the private key corresponding to the public key in the WIT. The WPT is sent in the Workload-Proof-Token header field of the request. A WPT contains the following:

• in the JOSE header:
  • alg: An identifier for an appropriate JWS asymmetric digital signature algorithm corresponding to the confirmation key in the associated WIT.
  • typ: the WPT is explicitly typed, as recommended in , using the application/wimse-proof+jwt media type.
• in the JWT claims:
  • iss: The issuer of the token, which is the calling workload, represented by the same value as the sub claim of the associated WIT.
  • aud: The audience of the token contains the HTTP target URI () of the request to which the WPT is attached, without query or fragment parts.
  • exp: The expiration time of the WIT (as defined in ). WPT lifetimes MUST be short, e.g., on the order of minutes or seconds.
  • jti: A unique identifier for the token.
  • ath: Hash of the OAuth access token, if present in the request, which might convey end-user identity and authorization context of the request. The value, as per , is the base64url encoding of the SHA-256 hash of the ASCII encoding of the access token's value.
  • tth: Hash of the Txn-Token , if present in the request, which might convey end-user identity and authorization context of the request. The value MUST be the result of a base64url encoding (as defined in ) of the SHA-256 hash of the ASCII encoding of the associated token's value.
  • oth: Hash of any other token in the request that might convey end-user identity and authorization context of the request. The value MUST be the result of a base64url encoding (as defined in ) of the SHA-256 hash of the ASCII encoding of the associated token's value. (note: this is less than ideal but seems we need something like this for extensibility)

An example WPT might look like the following:

Example Workload Proof Token (WPT)

The decoded JOSE header of the WPT from the example above is shown here:

Example WPT JOSE Header

The decoded JWT claims of the WPT from the example above are shown here:

Example WPT Claims

An example of an HTTP request with both the WIT and WPT from prior examples is shown below:

Example HTTP Request with WIT and WPT

To validate the WPT in the request, the recipient MUST ensure the following:

• There is exactly one Workload-Proof-Token header field in the request.
• The Workload-Proof-Token header field value is a single and well-formed JWT.
• The WPT signature is valid using the public key from the confirmation claim of the WIT.
• The typ JOSE header parameter of the WPT conveys a media type of wimse-proof+jwt.
• The iss claim of the WPT matches the sub claim of the WIT. (note: not sure iss in the WPT is useful or necessary)
• The aud claim of the WPT matches the target URI, or an acceptable alias or normalization thereof, of the HTTP request in which the WPT was received, ignoring any query and fragment parts.
• The exp claim is present and conveys a time that has not passed. WPTs with an expiration time unreasonably far in the future SHOULD be rejected.
• Optionally, check that the value of the jti claim has not been used before in the time window in which the respective WPT would be considered valid.
• If presented in conjunction with an OAuth access token, the value of the ath claim matches the hash of that token's value.
• If presented in conjunction with a Txn-Token, the value of the tth claim matches the hash of that token's value.
• If presented in conjunction with a token conveying end-user identity or authorization context, the value of the oth claim matches the hash of that token's value.

Option 2: Authentication Based on HTTP Message Signatures This option uses the WIMSE Identity Token () to sign the request and optionally, the response. This specification defines a profile of the Message Signatures specification . The request is signed as per . The following derived components MUST be signed:

• @method
• @request-target

In addition, the following request headers MUST be signed when they exist:

• Content-Type
• Content-Digest
• Authorization
• Txn-Token
• Workload-Identity-Token

If the response is signed, the following components MUST be signed:

• @status
• @method;req
• @request-target;req
• Content-Type if it exists
• Content-Digest if it exists
• Workload-Identity-Token

For both requests and responses, the following signature parameters MUST be included:

• created
• expires - expiration MUST be short, e.g. on the order of minutes. The WIMSE architecture will provide separate mechanisms in support of long-lived compute processes.
• nonce
• tag - the value for implementations of this specification is wimse-service-to-service

Since the signing key is sent along with the message, the keyid parameter SHOULD NOT be used. It is RECOMMENDED to include only one signature with the HTTP message. If multiple ones are included, then the signature label included in both the Signature-Input and Signature headers SHOULD be wimse. A sender MUST ensure that each nonce it generates is unique, at least among messages sent to the same recipient. To detect message replays, a recipient MAY reject a message (request or response) if a nonce is repeated. To promote interoperability, the ecdsa-p256-sha256 signing algorithm MUST be implemented by general purpose implementations of this spec. OPEN ISSUE: do we use the `Accept-Signature` field to signal that the response must be signed? Following is a non-normative example of a signed request and a signed response, where the caller is using the keys specified in .

Signed Request

Assuming that the workload being called has the following keypair:

Callee Private Key

A signed response would be:

Signed Response

Using Mutual TLS for Service To Service Authentication The WIMSE workload identity may be carried within an X.509 certificate. The WIMSE workload identity MUST be encoded in a SubjectAltName extension of type URI. There MUST be only one SubjectAltName extension of type URI in a WIMSE certificate. The WIMSE certificate may contain SubjectAltName extensions of other types such as DNSName. WIMSE identities may be used to validate server and client connections. When validating a WIMSE identity the relying party MUST validate that the CA issuer for the WIMSE identity is authorized to issue certificates for the trust domain of the WIMSE identity in the certificate. Other PKIX path validation rules apply. Servers wishing to use the WIMSE identity for authorizing the client MUST require client certificate authentication in the TLS handshake. Other methods of post handshake authentication are not specified by this document. WIMSE clients and servers MUST validate that the trust domain portion of the WIMSE certificate matches the expected trust domain for the other side of the connection.

Host Name Validation TODO: need to define trust root used to validate the certificate is appropriate for the trust domain. It is RECOMMENDED that the server certificate contain a DNS SAN that the client can use to perform standard host name validation (). The client SHOULD also extract the WIMSE identity from the certificate if it is present and validate that the WIMSE trust domain matches the intended trust domain for the server. The client MAY then further use the WIMSE identity in applying authorization policy to the server. If the client does not use the DNS SAN then the client MUST match the WIMSE identity in the certificate against the WIMSE identity of the workload of the intended server according to a locally defined policy. The host portion of the WIMSE URI is NOT treated as a host name as specified in section 6.4 of but rather as a trust domain. The server identity is encoded in the path portion of the WIMSE identity in a deployment specific way.

Authorization Using the WIMSE Identity The client or server application may retrieve the WIMSE identity from the TLS layer for use in authorization, accounting and auditing. For example, the full URI may be matched against ACLs and other policy constructs to authorize actions requested by the peer.

Security Considerations

WIMSE Identity The WIMSE ID is scoped within an issuer and therefore any sub-components (path portion of ID) are only unique within a trust domain defined by the issuer. Using a WIMSE ID without taking into account the trust domain could allow one domain to issue tokens to spoof identities in another domain. Additionally, the trust domain must be tied to an authorized issuer cryptographic trust root through some mechanism such as a JWKS or X.509 certificate chain. The association of an issuer, trust domain and a cryptographic trust root MUST be communicated securely out of band. TODO: Should there be a DNS name to Trust domain mapping defined or some other discovery mechanism?

Workload ID Token and Proof of Possession The Workload ID token (WIT) is bound to a secret cryptographic key and is always presented with a proof of possession as described in . The WIT is a general purpose token that can be presented in multiple contexts. The WIT and its PoP are only used in the application-level options, and both are not used in MTLS. The WIT MUST NOT be used as a bearer token. While this helps reduce the sensitivity of the token it is still possible that a token and its proof of possession may be captured and replayed within the PoP's lifetime. The following are some mitigations for the capture and reuse of the proof of possession (PoP):

• Preventing Eavesdropping and Interception with TLS

An attacker observing or intercepting the communication channel can view the token and its proof of possession and attempt to replay it to gain an advantage. In order to prevent this the token and proof of possession MUST be sent over a secure, server authenticated TLS connection unless a secure channel is provided by some other mechanisms. Host name validation according to MUST be performed. The WIT itself is not usable without a proof of possession.

• Limiting Proof of Possession Lifespan

The proof of possession MUST be time limited. A PoP should only be valid over the time necessary for it to be successfully used for the purpose it is needed. This will typically be on the order of minutes. PoPs received outside their validity time MUST be rejected.

• Limiting Proof of Possession Scope

In order to reduce the risk of theft and replay the PoP should have a limited scope. For example, a PoP may be targeted for use with a specific workload and even a specific transaction to reduce the impact of a stolen PoP. In some cases a workload may wish to reuse a PoP for a period of time or have it accepted by multiple target workloads. A careful analysis is warranted to understand the impacts to the system if a PoP is disclosed allowing it to be presented by an attacker along with a captured WIT.

• Binding to a Timestamp or Nonce

A proof of possession should include information that can be used to uniquely identify it such as a unique timestamp or nonce. This can be used by the receiver to perform basic replay protection against tokens it has already seen. Depending upon the design of the system it may be difficult to synchronize the replay cache across all token validators. In this case, if the PoP is not sufficiently scoped it may be usable with another workload. While a fresh nonce could be included in the PoP, a mechanism for distributing a fresh challenge nonce from the validator to provide single use properties of a PoP is outside the scope of this specification.

• Binding to TLS Endpoint

The POP MAY be bound to a transport layer sender such as the client identity of a TLS session or TLS channel binding parameters. The mechanisms for binding are outside the scope of this specification.

Middle Boxes In some deployments the WIMSE token and proof of possession may pass through multiple systems. The communication between the systems is over TLS, but the token and PoP are available in the clear at each intermediary. While the intermediary cannot modify the token or the information within the PoP they can attempt to capture and replay the token or modify the data not protected by the PoP. Mitigations listed in the previous section can be used to provide some protection from middle boxes. Deployments should perform analysis on their situation to determine if it is appropriate to trust and allow traffic to pass through a middle box.

Privacy Considerations WITs and the proofs of possession may contain private information such as user names or other identities. Care should be taken to prevent the disclosure of this information. The use of TLS helps protect the privacy of WITs and proofs of possession. WITs and certificates with WIMSE identifiers are typically associated with a workload and not a specific user, however in some deployments the workload may be associated directly to a user. While these are exceptional cases a deployment should evaluate if the disclosure of WITs or certificates can be used to track a user.

IANA Considerations TODO: maybe a URI Scheme registration of wimse in URI schemes per but it's only being used in an example right now and might not even be appropriate. Or maybe use an ietf URI scheme a la URN Namespace for IETF Use somehow. Or maybe nothing. Or maybe something else. TODO: tth and maybe oth claim in JSON Web Token Claims Registry

Media Type Registration TODO: application/wimse-id+jwt or appropriately bikeshedded media type name (despite my ongoing unease with using media types for typing JWTs) in Media Types. TODO: application/wimse-proof+jwt ...

Hypertext Transfer Protocol (HTTP) Field Name Registration TODO: Workload-Identity-Token from TODO: Workload-Proof-Token from

References Normative References Internet technical specifications often need to define a formal syntax. Over the years, a modified version of Backus-Naur Form (BNF), called Augmented BNF (ABNF), has been popular among many Internet specifications. The current specification documents ABNF. It balances compactness and simplicity with reasonable representational power. The differences between standard BNF and ABNF involve naming rules, repetition, alternatives, order-independence, and value ranges. This specification also supplies additional rule definitions and encoding for a core lexical analyzer of the type common to several Internet specifications. [STANDARDS-TRACK] JSON Web Signature (JWS) represents content secured with digital signatures or Message Authentication Codes (MACs) using JSON-based data structures. Cryptographic algorithms and identifiers for use with this specification are described in the separate JSON Web Algorithms (JWA) specification and an IANA registry defined by that specification. Related encryption capabilities are described in the separate JSON Web Encryption (JWE) specification. A JSON Web Key (JWK) is a JavaScript Object Notation (JSON) data structure that represents a cryptographic key. This specification also defines a JWK Set JSON data structure that represents a set of JWKs. Cryptographic algorithms and identifiers for use with this specification are described in the separate JSON Web Algorithms (JWA) specification and IANA registries established by that specification. JSON Web Token (JWT) is a compact, URL-safe means of representing claims to be transferred between two parties. The claims in a JWT are encoded as a JSON object that is used as the payload of a JSON Web Signature (JWS) structure or as the plaintext of a JSON Web Encryption (JWE) structure, enabling the claims to be digitally signed or integrity protected with a Message Authentication Code (MAC) and/or encrypted. This specification describes how to declare in a JSON Web Token (JWT) that the presenter of the JWT possesses a particular proof-of- possession key and how the recipient can cryptographically confirm proof of possession of the key by the presenter. Being able to prove possession of a key is also sometimes described as the presenter being a holder-of-key. JSON Web Tokens, also known as JWTs, are URL-safe JSON-based security tokens that contain a set of claims that can be signed and/or encrypted. JWTs are being widely used and deployed as a simple security token format in numerous protocols and applications, both in the area of digital identity and in other application areas. This Best Current Practices document updates RFC 7519 to provide actionable guidance leading to secure implementation and deployment of JWTs. The Hypertext Transfer Protocol (HTTP) is a stateless application-level protocol for distributed, collaborative, hypertext information systems. This document describes the overall architecture of HTTP, establishes common terminology, and defines aspects of the protocol that are shared by all versions. In this definition are core protocol elements, extensibility mechanisms, and the "http" and "https" Uniform Resource Identifier (URI) schemes. This document updates RFC 3864 and obsoletes RFCs 2818, 7231, 7232, 7233, 7235, 7538, 7615, 7694, and portions of 7230. This document describes a mechanism for creating, encoding, and verifying digital signatures or message authentication codes over components of an HTTP message. This mechanism supports use cases where the full HTTP message may not be known to the signer and where the message may be transformed (e.g., by intermediaries) before reaching the verifier. This document also describes a means for requesting that a signature be applied to a subsequent HTTP message in an ongoing HTTP exchange. In many standards track documents several words are used to signify the requirements in the specification. These words are often capitalized. This document defines these words as they should be interpreted in IETF documents. This document specifies an Internet Best Current Practices for the Internet Community, and requests discussion and suggestions for improvements. RFC 2119 specifies common key words that may be used in protocol specifications. This document aims to reduce the ambiguity by clarifying that only UPPERCASE usage of the key words have the defined special meanings. A Uniform Resource Identifier (URI) is a compact sequence of characters that identifies an abstract or physical resource. This specification defines the generic URI syntax and a process for resolving URI references that might be in relative form, along with guidelines and security considerations for the use of URIs on the Internet. The URI syntax defines a grammar that is a superset of all valid URIs, allowing an implementation to parse the common components of a URI reference without knowing the scheme-specific requirements of every possible identifier. This specification does not define a generative grammar for URIs; that task is performed by the individual specifications of each URI scheme. [STANDARDS-TRACK] Many application technologies enable secure communication between two entities by means of Transport Layer Security (TLS) with Internet Public Key Infrastructure using X.509 (PKIX) certificates. This document specifies procedures for representing and verifying the identity of application services in such interactions. This document obsoletes RFC 6125. Informative References IANA Venafi Zscaler The increasing prevalence of cloud computing and micro service architectures has led to the rise of complex software functions being built and deployed as workloads, where a workload is defined as a running instance of software executing for a specific purpose. This document discusses an architecture for designing and standardizing protocols and payloads for conveying workload identity and security context information. This document describes a mechanism for sender-constraining OAuth 2.0 tokens via a proof-of-possession mechanism on the application level. This mechanism allows for the detection of replay attacks with access and refresh tokens. SGNL Capital One Microsoft Transaction Tokens (Txn-Tokens) enable workloads in a trusted domain to ensure that user identity and authorization context of an external programmatic request, such as an API invocation, are preserved and available to all workloads that are invoked as part of processing such a request. Txn-Tokens also enable workloads within the trusted domain to optionally immutably assert to downstream workloads that they were invoked in the call chain of the request. This document updates the guidelines and recommendations, as well as the IANA registration processes, for the definition of Uniform Resource Identifier (URI) schemes. It obsoletes RFC 4395.

Acknowledgments TODO acknowledge.