| secevent | A. Backman |
| Internet-Draft | Amazon |
| Intended status: Standards Track | W. Denniss |
| Expires: June 2, 2018 | |
| M. Ansari | |
| Cisco | |
| M. Jones | |
| Microsoft | |
| November 29, 2017 |
Security Event Token (SET)
draft-backman-secevent-token-01
This specification defines the Security Event Token, which may be distributed via a protocol such as HTTP. The Security Event Token (SET) specification profiles the JSON Web Token (JWT), which can be optionally signed and/or encrypted. A SET describes a statement of fact from the perspective of an issuer that it intends to share with one or more receivers.
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 June 2, 2018.
Copyright (c) 2017 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 Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
This specification defines an extensible Security Event Token (SET) format which may be exchanged using protocols such as HTTP. The specification builds on the JSON Web Token (JWT) format [RFC7519] in order to provide a self-contained token that can be optionally signed using JSON Web Signature (JWS) [RFC7515] and/or encrypted using JSON Web Encryption (JWE) [RFC7516].
This specification profiles the use of JWT for the purpose of issuing security event tokens (SETs). This specification defines a base format upon which profiling specifications define actual events and their meanings. Unless otherwise specified, this specification uses non-normative example events intended to demonstrate how events may be constructed.
This specification is scoped to security and identity related events. While security event tokens may be used for other purposes, the specification only considers security and privacy concerns relevant to identity and personal information.
Security Events are not commands issued between parties. A security event is a statement of fact from the perspective of an issuer about the state of a security subject (e.g., a web resource, token, IP address, the issuer itself) that the issuer controls or is aware of, that has changed in some way (explicitly or implicitly). A security subject MAY be permanent (e.g., a user account) or temporary (e.g., an HTTP session) in nature. A state change could describe a direct change of entity state, an implicit change of state or other higher- level security statements such as:
While subject state changes are often triggered by a user-agent or security-subsystem, the issuance and transmission of an event often occurs asynchronously and in a back-channel to the action which caused the change that generated the security event. Subsequently, an Event Receiver, having received a SET, validates and interprets the received SET and takes its own independent actions, if any. For example, having been informed of a personal identifier being associated with a different security subject (e.g., an email address is being used by someone else), the Event Receiver may choose to ensure that the new user is not granted access to resources associated with the previous user. Or, the Event Receiver may not have any relationship with the subject, and no action is taken.
While Event Receivers will often take actions upon receiving SETs, security events cannot be assumed to be commands or requests. The intent of this specification is to define a way of exchanging statements of fact that Event Receivers may interpret for their own purposes. As such, SETs have no capability for error signaling other to ensure the validation of a received SET.
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 [RFC2119]. These keywords are capitalized when used to unambiguously specify requirements of the protocol or application features and behavior that affect the inter-operability and security of implementations. When these words are not capitalized, they are meant in their natural-language sense.
For purposes of readability, examples are not URL encoded. Implementers MUST percent encode URLs as described in Section 2.1 of [RFC3986].
Throughout this document, all figures MAY contain spaces and extra line-wrapping for readability and space limitations. Similarly, some URIs contained within examples have been shortened for space and readability reasons.
The following definitions are used with SETs:
A SET is a data structure that encodes an “event payload” describing a security event, wrapped in an “envelope” providing metadata and context for the security event. The SET envelope is a JWT Claims Set as defined in [RFC7519], consisting of a JSON object containing a set of claims. The event payload is a JSON object contained within the SET envelope, itself containing claims that express information about the event, e.g. the type of event, the subject of the event, and other information defined in a Profiling Specification.
This specification defines a core set of claims for use in SET envelopes and event payloads, however Profiling Specifications MAY define additional claims of both types. It is RECOMMENDED that Profiling Specifications define claims to be used in the event payload rather than the envelope. If a Profiling Specification does define envelope claims, those claims SHOULD be registered in the JWT Token Claims Registry [IANA.JWT.Claims] or have Public Claim Names as defined in Section 4.2 of [RFC7519].
This specification profiles the following claims defined in [RFC7519] for use in the SET envelope:
This specification defines the following new claims for use in the SET envelope:
This specification defines the following claims for use in event payloads:
Both the SET envelope and event payload MAY contain additional claims, such as those defined in a Profiling Specification. The format and meaning of these claims is out of scope of this specification. Implementations SHOULD ignore any claims in the SET envelope or event payload that they do not understand.
The following is a non-normative example showing a SET envelope expressing a hypothetical event with two additional claims in the event payload:
{
"jti": "3d0c3cf797584bd193bd0fb1bd4e7d30",
"iss": "https://transmitter.example.com",
"aud": [ "https://receiver.example.com" ],
"iat": 1458496025,
"event": {
"event_type": "https://secevent.example.com/example_event",
"event_subject": {
"identifier_type": "urn:ietf:params:secevent:subject:email",
"email": "user@example.com"
},
"event_time": 1458492425,
"claim_1": "foo",
"claim_2": "bar"
}
}
Figure 1: Example SET With Event Claims In Payload
The payload in this example contains the following:
The Subject Identifier provides a common syntax for expressing the subject of a security event. A Subject Identifier is a JSON object representing an instance of a Subject Identifier Type. A Subject Identifier Type defines a way of identifying the subject of an event. Typically this is done by defining a set of one or more claims about a subject that when taken together collectively identify that subject. Each Subject Identifier Type MUST have a name which MUST be registered in the IANA “SET Subject Identifier Types” registry established by Section 7.1.
A Subject Identifier MUST contain an “identifier_type” claim, whose value is a string containing the name of the Subject Identifier’s Subject Identifier Type. All other claims within the Subject Identifier MUST be defined by the Subject Identifier Type.
The names of the Subject Identifier Types defined below are registered in the IANA “SET Subject Identifier Types” registry established by Section 7.1.
The “Implicit” Subject Identifier Type indicates that the recipient is to be determined implicitly, either from other claims in the SET envelope or event payload, or through some other context. For example, there may be event types for which the only logical subject is the transmitter itself, in which case the subject is implicitly known from the “iss” claim in the SET envelope.
The Implicit Subject Identifier Type has the name “implicit”. This type contains no additional claims.
The following is a non-normative example of a Subject Identifier representing an instance of the Implicit Subject Identifier Type:
{
"identifier_type": "implicit"
}
Figure 2: An Instance of the Implicit Subject Identifier Type
The “Email” Subject Identifier Type identifies a subject by email address. It has the name “email”, and contains a single additional claim:
The following is a non-normative example of a Subject Identifier representing an instance of the Email Subject Identifier Type:
{
"identifier_type": "email",
"email": "user@example.com"
}
Figure 3: An Instance of the Email Subject Identifier Type
The “Phone Number” Subject Identifier Type identifies a subject by phone number. It has the name “phone_number”, and contains a single claim:
The following is a non-normative example of a Subject Identifier representing an instance of the Phone Number Subject Identifier Type:
{
"identifier_type": "phone_number",
"phone_number": "+1 206 555 0123"
}
Figure 4: An Instance of the Phone Number Subject Identifier Type
The “Issuer and Subject” Subject Identifier Type identifies a subject by an issuer and subject pair. It has the name “iss-sub”, and contains two claims:
The following is a non-normative example of a Subject Identifier representing an instance of the Issuer and Subject Subject Identifier Type:
{
"identifier_type": "iss-sub",
"iss": "http://id.example.com",
"sub": "example.user.1234"
}
Figure 5: An Instance of the Issuer and Subject Subject Identifier Type
This specification registers the “application/secevent+jwt” media type. SETs MAY include this media type in the “typ” header parameter of the JWT containing the SET to explicitly declare that the JWT is a SET. This MUST be included if the SET could be used in an application context in which it could be confused with other kinds of JWTs. Profiling Specifications MAY declare that this is REQUIRED for SETs containing events defined by the Profiling Specification.
Per the definition of “typ” in Section 4.1.9 of [RFC7515], it is RECOMMENDED that the “application/” prefix be omitted. Therefore, the “typ” value used SHOULD be “secevent+jwt”.
A SET is a JWT, and therefore it’s construction follows that described in [RFC7519].
While this specification uses JWT to convey a SET, implementers SHALL NOT use SETs to convey authentication or authorization assertions.
The following is the example JWT Claims Set from Figure 1, expressed as an unsigned JWT. The JOSE Header is:
{"typ":"secevent+jwt","alg":"none"}
Base64url encoding of the octets of the UTF-8 representation of the JOSE Header yields:
eyJ0eXAiOiJzZWNldmVudCtqd3QiLCJhbGciOiJub25lIn0
The example JWT Claims Set is encoded as follows:
ew0KICAgImp0aSI6ICIzZDBjM2NmNzk3NTg0YmQxOTNiZDBmYjFiZDRlN2QzMCIsDQog ICAiaXNzIjogImh0dHBzOi8vdHJhbnNtaXR0ZXIuZXhhbXBsZS5jb20iLA0KICAgImF1 ZCI6IFsgImh0dHBzOi8vcmVjZWl2ZXIuZXhhbXBsZS5jb20iIF0sDQogICAiaWF0Ijog MTQ1ODQ5NjAyNSwNCiAgICJldmVudCI6IHsNCiAgICAgImV2ZW50X3R5cGUiOiAiaHR0 cHM6Ly9zZWNldmVudC5leGFtcGxlLmNvbS9leGFtcGxlX2V2ZW50IiwNCiAgICAgImV2 ZW50X3N1YmplY3QiOiB7DQogICAgICAgImlkZW50aWZpZXJfdHlwZSI6ICJlbWFpbCIs DQogICAgICAgImVtYWlsIjogInVzZXJAZXhhbXBsZS5jb20iDQogICAgIH0sDQogICAg ICJldmVudF90aW1lIjogMTQ1ODQ5MjQyNSwNCiAgICAgImNsYWltXzEiOiAiZm9vIiwN CiAgICAgImNsYWltXzIiOiAiYmFyIg0KICAgfQ0KIH0=
The encoded JWS signature is the empty string. Concatenating the parts yields the following complete JWT:
eyJ0eXAiOiJzZWNldmVudCtqd3QiLCJhbGciOiJub25lIn0. ew0KICAgImp0aSI6ICIzZDBjM2NmNzk3NTg0YmQxOTNiZDBmYjFiZDRlN2QzMCIsDQog ICAiaXNzIjogImh0dHBzOi8vdHJhbnNtaXR0ZXIuZXhhbXBsZS5jb20iLA0KICAgImF1 ZCI6IFsgImh0dHBzOi8vcmVjZWl2ZXIuZXhhbXBsZS5jb20iIF0sDQogICAiaWF0Ijog MTQ1ODQ5NjAyNSwNCiAgICJldmVudCI6IHsNCiAgICAgImV2ZW50X3R5cGUiOiAiaHR0 cHM6Ly9zZWNldmVudC5leGFtcGxlLmNvbS9leGFtcGxlX2V2ZW50IiwNCiAgICAgImV2 ZW50X3N1YmplY3QiOiB7DQogICAgICAgImlkZW50aWZpZXJfdHlwZSI6ICJlbWFpbCIs DQogICAgICAgImVtYWlsIjogInVzZXJAZXhhbXBsZS5jb20iDQogICAgIH0sDQogICAg ICJldmVudF90aW1lIjogMTQ1ODQ5MjQyNSwNCiAgICAgImNsYWltXzEiOiAiZm9vIiwN CiAgICAgImNsYWltXzIiOiAiYmFyIg0KICAgfQ0KIH0=.
Figure 6: Example Unsecured Security Event Token
For the purpose of a simpler example in Figure 5, an unsecured token was shown. When SETs are not signed or encrypted, the Event Receiver MUST employ other mechanisms such as TLS and HTTP to provide integrity, confidentiality, and issuer validation, as needed by the application.
When validation (i.e. auditing), or additional transmission security is required, JWS signing and/or JWE encryption MAY be used. To create and or validate a signed and/or encrypted SET, follow the instructions in Section 7 of [RFC7519].
In order to accommodate use cases that require communicating multiple related security events to an Event Receiver, this section defines the “Related Events” event type. A Related Events event is essentially a container for two or more events that are related to one another, in that they represent or express different aspects of the same event or state change. The Related Events event SHOULD NOT be used to combine unrelated events into a single set, and MUST NOT be used as a general purpose batch transmission mechanism. Profiling Specifications that require an event grouping mechanism with these or other semantics are encouraged to define additional event types for their use cases.
The event type for the Related Events event is the URN “urn:ietf:secevents:related_events”.
The Related Events event has a single additional event payload claim:
Nested Events can inherit the “event_id”, “event_subject”, and “event_time” claims from the Related Events payload. Transmitters MAY omit some, all, or none of these claims from a Nested Event. Transmitters MAY omit claims from some Nested Events and include them in others within the same Related Events event. When a claim is omitted, recipients MUST use the value of the corresponding claim in the Related Event event’s payload.
The following is a non-normative example of a SET containing a Related Events event:
{
"jti": "1c0038c2-02db-40de-ad50-122a64724166",
"iss": "https://transmitter.example.com",
"aud": [ "https://receiver.example.com" ],
"iat": 1510666261,
"event": {
"event_type": "urn:ietf:secevent:related_events",
"event_subject": {
"identifier_type": "email",
"email": "user@example.com"
},
"event_id": "container",
"event_time": 1510662661,
"events": [
{
"event_id": "nested_1",
"event_type": "http://specs.example.com/set_profile/event_1"
},
{
"event_id": "nested_2",
"event_type": "http://specs.example.com/set_profile/event_2",
"event_time": 151059061
}
]
}
}
Figure 7: Example SET Containing A Related Events Event
The following table demonstrates how Nested Events inherit values for omitted claims:
+-----------+------------+-------------------------------+
| Event ID | Event Time | Event Subject |
+-----------+------------+-------------------------------+
| container | 151062661 | { |
+-----------+ | "identifier_type": "email", |
| nested_1 | | "email": "user@example.com" |
+-----------+------------+ } |
| nested_2 | 151059061 | |
+-----------+------------+-------------------------------+
Figure 8: Example of Event Payloads Inheriting Values for Omitted Claims
Since the Nested Event with event ID “nested_1” omits the “event_time” claim, it inherits the event time from the Related Events event payload. Similarly, since both Nested Events “nested_1” and “nested_2” omit the “event_subject” claim, both inherit the event subject from the Related Events event payload.
Profiling Specifications for SETs define the syntax and semantics of SETs conforming to that SET profile and rules for validating those SETs. The syntax defined by Profiling Specifications includes what SET envelope and event payload claims are used by SETs expressing and event defined by the profile.
Defining the semantics of the SET contents for SETs utilizing the profile is equally important. Possibly most important is defining the procedures used to validate the SET issuer and to obtain the keys controlled by the issuer that were used for cryptographic operations used in the JWT representing the SET. For instance, some profiles may define an algorithm for retrieving the SET issuer’s keys that uses the “iss” claim value as its input. Likewise, if the profile allows (or requires) that the JWT be unsecured, the means by which the integrity of the JWT is ensured MUST be specified.
Profiling Specifications MUST define how the event Subject is identified in the SET, as well as how to differentiate between the event Subject’s Issuer and the SET Issuer, if applicable. It is NOT RECOMMENDED for Profiling Specifications to use the “sub” claim defined in [RFC7519] in cases in which the Subject is not globally unique and has a different Issuer from the SET itself.
Profiling Specifications MUST clearly specify the steps that a recipient of a SET utilizing that profile MUST perform to validate that the SET is both syntactically and semantically valid.
As needs change and new use cases develop, it may be desirable to augment existing event definitions with new claims. In order to avoid collisions, Profiling Specifications that extend existing events with additional event payload claims SHOULD use Collision-Resistant Names as defined in Section 2 of [RFC7519] for the names of the new claims.
SETs may often contain sensitive information. Therefore, methods for distribution of events SHOULD require the use of a transport-layer security mechanism when distributing events. Parties MUST support TLS 1.2 [RFC5246] and MAY support additional transport-layer mechanisms meeting its security requirements. When using TLS, the client MUST perform a TLS/SSL server certificate check, per [RFC6125]. Implementation security considerations for TLS can be found in “Recommendations for Secure Use of TLS and DTLS” [RFC7525].
Security Events distributed through third-parties or that carry personally identifiable information, SHOULD be encrypted using JWE [RFC7516] or secured for confidentiality by other means.
Unless integrity of the JWT is ensured by other means, it MUST be signed using JWS [RFC7515] so that individual events can be authenticated and validated by the Event Receiver.
This specification does not define a delivery mechanism by itself. In addition to confidentiality and integrity (discussed above), implementers and Profiling Specifications MUST consider the consequences of delivery mechanisms that are not secure and/or not assured. For example, while a SET may be end-to-end secured using JWE encrypted SETs, without TLS there is no assurance that the correct endpoint received the SET and that it could be successfully processed.
As defined in this specification, there is no defined way to order multiple SETs in a sequence. Depending on the type and nature of SET event, order may or may not matter. For example, in provisioning, event order is critical – an object could not be modified before it was created. In other SET types, such as a token revocation, the order of SETs for revoked tokens does not matter. If however, the event was described as a log-in or logged-out status for a user subject, then order becomes important.
Profiling Specifications and implementers SHOULD take caution when using timestamps such as “iat” to define order. Distributed systems will have some amount of clock-skew and thus time by itself will not guarantee order.
Specifications profiling SET SHOULD define a mechanism for detecting order or sequence of events.
When SETs are delivered asynchronously and/or out-of-band with respect to the original action that incurred the security event, it is important to consider that a SET might be delivered to an Event Receiver in advance or well behind the process that caused the event. For example, a user having been required to logout and then log back in again, may cause a logout SET to be issued that may arrive at the same time as the user-agent accesses a web site having just logged- in. If timing is not handled properly, the effect would be to erroneously treat the new user session as logged out. Profiling Specifications SHOULD be careful to anticipate timing and subject selection information. For example, it might be more appropriate to cancel a “session” rather than a “user”. Alternatively, the specification could use timestamps that allows new sessions to be started immediately after a stated logout event time.
Because [RFC7519] states that “all claims that are not understood by implementations MUST be ignored”, there is a consideration that a SET token might be confused with ID Token [OpenID.Core] if a SET is mistakenly or intentionally used in a context requiring an ID Token. If a SET could otherwise be interpreted as a valid ID Token (because it includes the required claims for an ID Token and valid issuer and audience claim values for an ID Token) then that SET profile MUST require that the “exp” claim not be present in the SET. Because “exp” is a required claim in ID Tokens, valid ID Token implementations will reject such a SET if presented as if it were an ID Token.
Excluding “exp” from SETs that could otherwise be confused with ID Tokens is actually defense in depth. In any OpenID Connect contexts in which an attacker could attempt to substitute a SET for an ID Token, the SET would actually already be rejected as an ID Token because it would not contain the correct “nonce” claim value for the ID Token to be accepted in contexts for which substitution is possible.
Note that the use of explicit typing, as described in Section 2.2, will not achieve disambiguation between ID Tokens and SETs, as the ID Token validation rules do not use the “typ” header parameter value.
OAuth 2.0 [RFC6749] defines access tokens as being opaque. Nonetheless, some implementations implement access tokens as JWTs. Because the structure of these JWTs is implementation-specific, ensuring that a SET cannot be confused with such an access token is therefore likewise, in general, implementation specific. Nonetheless, it is recommended that SET profiles employ the following strategies to prevent possible substitutions of SETs for access tokens in contexts in which that might be possible:
JWTs are now being used in application areas beyond the identity applications in which they first appeared. For instance, the Session Initiation Protocol (SIP) Via Header Field [RFC8055] and Personal Assertion Token (PASSporT) [I-D.ietf-stir-passport] specifications both define JWT profiles that use mostly or completely different sets of claims than are used by ID Tokens. If it would otherwise be possible for an attacker to substitute a SET for one of these (or other) kinds of JWTs, then the SET profile must be defined in such a way that any substituted SET will result in its rejection when validated as the intended kind of JWT.
The most direct way to prevent confusion is to employ explicit typing, as described in Section 2.2, and modify applicable token validation systems to use the “typ” header parameter value. This approach can be employed for new systems but may not be applicable to existing systems.
Another way to ensure that a SET is not confused with another kind of JWT is to have the JWT validation logic reject JWTs containing an “events” claim unless the JWT is intended to be a SET. This approach can be employed for new systems but may not be applicable to existing systems.
For many use cases, the simplest way to prevent substitution is requiring that the SET not include claims that are required for the kind of JWT that might be the target of an attack. For example, for [RFC8055], the “sip_callid” claim could be omitted and for [I-D.ietf-stir-passport], the “orig” claim could be omitted.
In many contexts, simple measures such as these will accomplish the task, should confusion otherwise even be possible. Note that this topic is being explored in a more general fashion in JSON Web Token Best Current Practices [I-D.sheffer-oauth-jwt-bcp]. The proposed best practices in that draft may also be applicable for particular SET profiles and use cases.
If a SET needs to be retained for audit purposes, JWS MAY be used to provide verification of its authenticity.
Event Transmitters SHOULD attempt to specialize feeds so that the content is targeted to the specific business and protocol needs of an Event Receiver.
When sharing personally identifiable information or information that is otherwise considered confidential to affected users, Event Transmitters and Receivers MUST have the appropriate legal agreements and user consent or terms of service in place.
The propagation of subject identifiers can be perceived as personally identifiable information. Where possible, Event Transmitters and Receivers SHOULD devise approaches that prevent propagation – for example, the passing of a hash value that requires the Event Receiver to know the subject.
This section establishes the IANA “SET Subject Identifier Types” registry // TODO
This specification registers the “event” claim in the IANA “JSON Web Token Claims” registry [IANA.JWT.Claims] established by [RFC7519].
This section registers the “application/secevent+jwt” media type [RFC2046] in the “Media Types” registry [IANA.MediaTypes] in the manner described in [RFC6838], which can be used to indicate that the content is a SET.
| [I-D.ietf-stir-passport] | Wendt, C. and J. Peterson, "Personal Assertion Token (PASSporT)", Internet-Draft draft-ietf-stir-passport-11, February 2017. |
| [I-D.sheffer-oauth-jwt-bcp] | Sheffer, Y., Hardt, D. and M. Jones, "JSON Web Token Best Current Practices", Internet-Draft draft-sheffer-oauth-jwt-bcp-01, July 2017. |
| [OpenID.Core] | "OpenID Connect Core 1.0", n.d.. |
| [RFC2046] | Freed, N. and N. Borenstein, "Multipurpose Internet Mail Extensions (MIME) Part Two: Media Types", RFC 2046, DOI 10.17487/RFC2046, November 1996. |
| [RFC6838] | Freed, N., Klensin, J. and T. Hansen, "Media Type Specifications and Registration Procedures", BCP 13, RFC 6838, DOI 10.17487/RFC6838, January 2013. |
| [RFC7515] | Jones, M., Bradley, J. and N. Sakimura, "JSON Web Signature (JWS)", RFC 7515, DOI 10.17487/RFC7515, May 2015. |
| [RFC7516] | Jones, M. and J. Hildebrand, "JSON Web Encryption (JWE)", RFC 7516, DOI 10.17487/RFC7516, May 2015. |
| [RFC8055] | Holmberg, C. and Y. Jiang, "Session Initiation Protocol (SIP) Via Header Field Parameter to Indicate Received Realm", RFC 8055, DOI 10.17487/RFC8055, January 2017. |
The editors would like to thank Phil Hunt for his SET draft – on which much of this specification is based – and his continuing contributions to this draft.
The editors would like to thank the participants on the IETF secevent mailing list and related working groups for their support of this specification.