]> Independent Identity Inc

phil.hunt@independentid.com

Cisco Systems

ncamwing@cisco.com

SCIM Identity Event SET Provisioning Token Internet-Draft SCIM This specification defines a set of Security Event types using the Security Event Token format (RFC8417). These events can be used across domains by SCIM Service Providers and receivers to exchange security event signals used for: request confirmations, replication, provisioning co-ordination, and security signals.

Introduction and Overview This specification defines Security Events for SCIM service providers and receivers as specified by the Security Event Tokens (SET) specificaiton. These events include: asynchronous transaction confirmations, replication, provisioning co-ordination, and security signals. In a typical HTTP client-server relationship a SCIM Protocol Client issues commands to a SCIM Server using HTTP methods such as GET, POST, PATCH, and DELETE in many cases to effect a state change. Subsequently, when multiple SCIM protocol client entities update SCIM resources, an individual client may need to be informed of changes that occur over time. This may be achieved through the use of event messages or signals in the form of Security Event Tokens (SET). SET tokens convey information about changes that have occured in a publishing domain that may be of interest to a receiving domain. Unlike SCIM Protocol requests, Security Events do not describe actions that a receiver must take, rather they are simple statements of fact about changes that have already occurred. The intent, is to allow the event receiver to determine the best follow-uip action to take within the context of the receiving domain. This gives each domain independent operation and control.

Requirements Language 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.

Notational Conventions For purposes of readability examples are not URL encoded. Implementers MUST percent encode URLs as described in Section 2.1 of. Throughout this documents all figures MAY contain spaces and extra line-wrapping for readability and space limitations. Similarly, some URI's contained within examples, have been shortened for space and readability reasons.

Definitions This specification uses definitions from the following specifications:

• Security Event Tokens , and
• System for Cross-Domain Identity Management Protocol .

Additionally, the following terms are defined:

Attributes and Claims

The JWT specification upon which SET is based uses the term "claims" to refer to attributes in a JSON token. SCIM in contrast uses the term "attributes" to refer to JSON attributes. For the purposes of this draft, the terms "attributes" and "claims" are equivalent.

CP

Abbreviation for "Co-ordinated Provisioning" as defined in . In these relationships an Event Publisher and Receiver typically exchange resource change events without exchanging change data. For a receiver to know the value of the data, the Event Receiver usually has to call back to the Event Publisher domain to receive a new copy of data (e.g. Using a SCIM GET request).

DBR

Abbreviation for "Domain Based Replication" as defined in . In this mode because there is an administrative relationship spanning multiple operational domains, data shared in Events typically includes all of the change data. This eliminates the need for a call back (e.g. a SCIM GET request), to retrieve raw data.

Event Feed

This describes the notion that a feed MAY be individualized per client. A service provider MAY offer to allow Event Receiver's to "subscribe" to specific event types or events about specific resources. If no option is offered, it is assumed the client will receive all events about all resources.

Event Receiver

A service that receives events for the purpose of subsequent action (e.g. such as replication), co-ordination of workflow, or signalling. In the case of SET Push Transfer , the Event Receiver is an HTTP Service Endpoint that receives requests. In the case of SET Poll-Based Transfer , the receiver is an HTTP client that initiates HTTP request to an Event Publisher endpoint.

Event Publisher

A system that issues SETs based on a change that has occurred at a SCIM Service Provider. For example, events MAY originate from a SCIM Create, Modify, or Delete per request. A SCIM Service Provider MAY be an Event Publisher or an independent service that aggregates events into Event Receiver feeds. As described above, when using , the Event Publisher is an HTTP Client that initiates HTTP POST requests to a defined Event Receiver endpoint. When using , the Event Publisher provides an HTTP endpoint which a receiver MAY use to "poll" for Security Events.

RS

Abbreviation for "Risk Signals" as defined in .

SCIM Client

An HTTP client that initiates SCIM Protocol requests and receives responses.

SCIM Service Provider

An HTTP server that implements SCIM Protocol and SCIM Schema .

SET

Abbreviation for "Security Event Token" as defined in

SCIM Events A SCIM event is a message, in the form of a Security Event Token . A SET event consists of a set of standard JWT "top-level" claims, plus one or more Event Payload claims that contains additional information relevant to a defined event.

Identifying the Subject of an Event SCIM Events SHALL use the "sub_id" claim defined by Subject Identifiers for Security Event Tokens specification to identify the subject of events. Since it is possible to issue a SCIM SET Event with multiple events, the sub_id claim MUST NOT be contained within an Event payload. A SET with multiple event URI's SHALL indicate that the events originate from the same transaction or resource state change. Finally, to distinguish SET Events from JWT tokens, the JWT "sub" claim MUST NOT be used.

SCIM Subject Id Example

The top-level claim "sub_id" contains a sub-claim "format" whose value is set to scim to indicate the other attributes present are SCIM attributes. The following sub_id attributes are defined:

uri

The SCIM relative path for the resource which usually consists of the resource type endpoint plus the resource id. For example /Users/2b2f880af6674ac284bae9381673d462. This attributes MUST be provided in a SCIM Event sub_id claim. Note the relative path is the path component after the SCIM Service Provider Base URI as defined in Section 1.3 . In cases where the Event Receiver is unable to match a URI, the Event Receiver MAY issue a call-back to a previously agreedn SCIM Service Provider Base URI plus the relative uri value and perform a SCIM GET request per Section 3.4.1 .

externalId

If known, the externalId value of the SCIM Resource that MAY be used by a receiver to identify the corresponding resource in the Event Receiver's domain.

id

The SCIM Id attribute MAY be used for backwards compatibility reasons in addition to the uri claim.

emails,username, ...

A SCIM attribute which is uniquely identifiable (e.g. username, emails). Used in situtations where the normal SCIM identifiers ( id and externalId) are insufficient to identfy a common resource between an Event Publisher and Event Receiver.

Common Event Attributes The following attributes are available for all events defined. Some attributes are defined as SET/JWT claims, while others are "Event Payload" claims as defined in Section 1.2 .

txn

For the purposes of SCIM, this SET defined "top-level" claim identifies a unique transaction originating at a SCIM Service Provider and/or its underlying data repository or database. The claim is used to detect duplicate transactions that may have been received (e.g. in the case of a re-transmitted or recovered event). If not provided, the SET jti claim MAY be used. Where txn identifies uniqueness within a SCIM Service Provider, multiple SETs may be issued each with distinct JTI's stemming from a common originating transaction with identical txn values.

data

Defined in SCIM Bulk Operations, Section 3.7 , thie payload attribute contains the equivalent SCIM command processed by the SCIM Service provider. For example, after processing a SCIM Create operation, the data contained includes the final representation of the created entity by the SCIM Service Provider including the assigned id value.

attributes

This payload attribute contains an array of attributes that were added, revised, or removed. For example:
"attributes": ["username","emails"]

Only one of data or attributes claims SHALL be provided depending on the event definition. This specifications defines a new URI prefix urn:ietf:params:event:SCIM which is used as the prefix for the following defined SCIM Events.

SCIM Feed Events This section defines events related to notices about which resources are being added or removed from an event feed. These events are used in Co-operative Provisioning scenarios where only a sub-set of entities are shared across an Event Feed. The URI prefix for these events is: urn:ietf:params:event:SCIM:feed

urn:ietf:params:event:SCIM:feed:add The specified resource was added to the Event Feed. A feed:add does not indicate a resource is new or has been recently created. For example, an existing user has had a new role (e.g. CRM_User) added to their profile which has caused their resource to join a feed.

Example SCIM Feed Add Event

urn:ietf:params:event:SCIM:feed:remove The specified resource has been removed from the feed. Removal does not indicate that the resource was deleted or otherwise deactivated. This event has minimal disclosure.

Example SCIM Feed Remove Event

SCIM Provisioning Events This section defines provisioning events that have occurred within a SCIM Service Provider. These events are used in both Domain Based Replication (DBR) and Co-operative Provisioning (CP) mode. The URI prefix for these events is: urn:ietf:params:event:SCIM:prov

urn:ietf:params:event:SCIM:prov:create Indicates a new SCIM resource has been created by the SCIM Service Provider and has been added to the Event Feed. Note that when a create event is sent, a corresponding urn:ietf:params:event:SCIM:feed:add event SHOULD NOT be issued in the same feed. In DBR mode mode, all claims of the new resource are included. In CP mode, the attributes returned discloses what attributes were created at the publisher. In DBR mode, the set of values reflecting the final state of the resource at the service provider are provided using the "data" attribute. Note that because this is a replication request, the uri attribute that was assigned by the SCIM Service Provider is shared so that all replicas in the domain use the same resource identifier.

Example SCIM Create (Domain Replication Mode)

Example SCIM Create Event (CP Mode)

The event above notifies the Event Receiver which attributes have changed but does not convey the actual information. The Event Receiver MAY retrieve that information by performing a SCIM GET to the sub value specified.

urn:ietf:params:event:SCIM:prov:patch The specified resource has been updated using SCIM PATCH. When in DBR mode, the data attribute contains the PATCH Request body. In CP mode, only the modified attribute name is included.

Example SCIM Patch Event (DBR Mode)

Example SCIM Patch Event (CP Mode)

urn:ietf:params:event:SCIM:prov:put The specified resource has been updated (e.g. one or more attributes has changed). In DBR mode, the SCIM PUT request body is included in the data attribute; or, In CP mode the modified attributes are listed using attributes.

Example SCIM Put Event (DBR Mode)

Example SCIM Put Event (CP Mode)

urn:ietf:params:event:SCIM:prov:delete The specified resource has been deleted from the SCIM publisher. The resource is also removed from the feed. When a DELETE is sent, a corresponding feedRemove is not issued. A delete event has minimal disclosure profile only.

Example SCIM Delete Event

urn:ietf:params:event:SCIM:prov:activate The specified resource (e.g. User) has been activated. This event indicates a high-level change in state as agreed between the Event Publisher and Event Receiver. For example, an activated resource is one that can now have an active session (may log in) from a security perspective.

Example SCIM Activate Event

urn:ietf:params:event:SCIM:prov:deactivate The specified resource (e.g. User) has been deactivated and disabled. The exact meaning must be agreed to by the Event Publisher and its corresponding Event Receiver. Typically this means the sub may no longer have an active security session. As with the activate event, this event has minimal disclosure requirements.

SCIM Signals Events This section defines security signal events that have occured within a SCIM Service Provider.The URI prefix for these events is: urn:ietf:params:event:SCIM:signal

urn:ietf:params:event:SCIM:sig:authMethod A new authentication method has been added to the User profile. As attackers often use new authentication methods to lock-out Users from their account, this signal can be used by the receiver that the chance of account them may be temporarily elevated. The receiver MAY also wish to take action such as resetting current authorizations or sessions.

Example SCIM Authentication Factor Change Event

urn:ietf:params:event:SCIM:sig:pwdReset The specified resource (e.g. User) has changed its password or the password has been reset. When the password has changed, the attributes attribute is supplied with the value "password".

Example SCIM Password Change Event

Miscellaneous Events This section defines events related miscellaneous events such as Asynchronous Request complettion that have occured within a SCIM Service Provider. The URI prefix for these events is: urn:ietf:params:event:SCIM:misc

urn:ietf:params:event:SCIM:misc:asyncResp This event signals the completion of a SCIM request. The payload contains the attributes defined in SCIM Bulk Section 3.7 and is the same a single SCIM Bulk Response Operation as per Section 3.7.3.

Example SCIM Async Response Event

Event Delivery For the purposes of SCIM, this draft uses the following methods exchange of events using the Security Event Token exchange methods:

Push-Based Security Event Token (SET) Delivery delivers tokens to an Event Receiver with an accessible HTTP Entpoint.

Poll-Based Security Event Token (SET) Delivery enables an event receiver to initiate calls to SET Event publisher endpoint to receive 1 or more events. This is particularly advantageous when an Event Receiver is behind a firewall or other network restriction that would prevent Push Delivery. An Event Receiver may also use HTTP "Long-polling" to achieve real time event delivery.

In the figure below, a possible distribution architecture is shown. This specification is only normatively concerned with the actual Security Event Transfer mechanism. SCIM Service providers MAY choose to implement SET transfer directly or they may use a method of allowing a single Event Publisher to assemble streams of events for transfer to a receiver. Likewise, on the receiving side, the only normative requirement is to be able to receive events and implement storage of Events to local recovery needs.

As Security Event Tokens are based on JWT tokens, it is possible to exchange events by a number of transfer mechanisms such as: XMPP , HTTP , and Message Buses (e.g. , Apache Kafka ). For example, on the publishing side, a cluster or network of SCIM servers may publish events to a common SET publisher service for distribution to 1 or more receivers. The Event Publisher MAY be incorporated directly into each SCIM Server, or a Local Event Delivery System might be used to collect events for an Event Publisher service for forwarding. How this is done is up to the implementer. This specification is only concerned with the interoperability of SET transfer between domains using and . The SET Transfer specifications provide a short-term method of recovery to ensure SET Events are successfully transferred. Once a receiving domain has successfully stored events to its own recovery needs, the receiving domain acknowledges the transfer of SET Events to the publisher using the method defined in and . SET does not specify local server event recovery mechanisms, this is up to the service implementation within each domain. This is done to enable cross-domain independence between domains. As an example, a hosted SCIM service provider with a series of SCIM servers replicates through a proprietary system. An enterprise customer who is the common client across both providers wants to establish a single administrative domain and wants to share SCIM change events between providers. Each domain has its own SCIM implementation and its own local replication strategy. The publishing domain issues events that the receiving domain picks up. Once the receiving domain has processed the event, the receiving domains own internal replication and recovery takes over. Because there is no need for the publishing domain to retain the event, it has the option to purge the event once the receiving domain has acknowledged it. This may be particularly critical if a publishing domain has dozens or even thousands of event receiving domains each with their own sub-set of data. Retaining all events for all receivers would become impracticle.

Security Event Token Signing and Encryption This specification uses Security Event Tokens as the message format for SCIM Events. As SETs are based on JWT tokens , they can be transmitted unsecurity, signed, or encrypted. For more information see the JWT Cookbook specification for examples. The decision on whether to use JWS and JWE depends on operational considerations. For each SCIM Feed relationship, it is up to deployers to decide on signing, encryption and algorithm requirements. Deployers SHOULD be aware that too much emphasis on turning on every possible encryption feature may cause operational performance to suffer. Deployers MUST weigh the security trade-offs of up-to-date SCIM services, vs. the potential information loss of an event.

Unsecured

Per Section 6 , tokens MAY be generated with {"alg":"none"}. This mode speeds up processing and is best used in DBR scenarios. Unencrypted tokens MUST be transferred over authenticated TLS layer encryption and SHOULD only be used in a restricted network environment.

Signed

JWS () signed SETs are useful when it is important to be able to verify the issuer of a SET as valid. In addition, some systems MAY wish to validate the authenticity of the event in a review process which may occur at a later date. While the content can be validated as originating from the correct issuer and is unmodified, the message contents remain unsecure. Signed SETs MUST be transferred over encrypted transport.

Encrypted

JWE () are encrypted SETs and are useful when the transport mechanism is not fully securable (e.g. messages carried by a third party). The use of JWEs ensures only the designated receiver can read the event and provides mutual authentication within the SET mesage itself.

Point-to-Point Delivery Over HTTP Security Event Tokens MAY be delivered using push-based HTTP delivery , or pull-based HTTP Polling . Both of these protocols define a method of transfer and acknowledgement to prevent loss-of-information and to provide re-transmission and recover. The method of transfer is best decided by considering the following advantages and disadvantages in a production scenario: Push-based delivery has the following advantages:

• Message transfer is instant (when compared to using a common Event Publisher acting as a relay), and in high event frequency scenarios, HTTP connections can be kept open.
• Scales well when an SCIM Event Publisher has thousands of event receivers and TCP resources may be limited.
• Does not require events to be routed to a single publisher node. SCIM Events may be issued by SCIM Service Provider nodes where the transaction occurred.
• SCIM Events only need to be retained until they have been delivered to designated receivers.

Push-delivery has the following disadvantages:

• A SCIM Event Publisher system needs authorization credentials enabling it to access the HTTP SCIM Event delivery endpoint.
• When synchronizing business data that is behind protected firewalls, a virtual network or other firewall policy may be required to allow external network based SCIM providers to deliver SCIM Events to internally hosted systems.

Delivery by HTTP Polling has the following advantages:

• It is possible for a SCIM Event Receiver to use the same SCIM credentials it uses when access the normal SCIM Service Provider service defined by .
• Systems behind protected network boundaries can reach externally hosted systems without requiring special firewall or network configuration.
• Instantaneous transfer can be used using HTTP Long-polling as described Section 2.1 of .

Polling-based delivery has the following disadvantages:

• Long-polling requires the use of persistent connections for which TCP resources may be limited. HTTP Long-polling is best used in scenarios when there are relatively few Event Receivers.
• The SCIM Event Publisher MUST retain events for the Event Receiver until delivered.

Security Considerations This specification depends on the Security Considerations for . The use of Json Web Encryption (JWE) can impose performance limitations when used in high event frequency scenarios. JWE is primarily useful only when the transfer of SETs involves an unsecured transfer method (e.g. URL) that would not otherwise be protected by the transfer protocol (e.g. SET Transfer over TLS ). For SCIM Provisioning events, the long-term series of changes may be critical to both sides. As such Event Publishers SHOULD consider storing events for receivers for longer periods of time in the case of an extended SET Transfer service failure. Similarly Event Receivers MUST ensure events are persisted directly or indirectly sufficient to meet local recovery needs before acknowledging received SET Events. When SET Events are stored for future delivery or retained local recovery MUST be limited only to the parties needed to support recovery or SET forwarding. JWS signed SET Events SHOULD be used to verify authenticity of the origin of a SET Event. Validating event signatures is both useful on the initial transfer of SET Events, and may also be useful for auditing purposes. In operation, some SCIM resources such as SCIM Groups may have a high rate of change. Implementors and operators SHOULD consider use of throttling techniques to balance immediacy and frequency. For example, a large group whose members change dozens of times per second may not need discrete SET Patch Events per change. Instead, issuing a single consolidated change per second or even minute may be beneficial to keeping Event Receivers up-to-date. Likewise, a Co-ordinated Provisioning Event Receiver (Section 2.2), does not necessarily need to retrieve the full Group on every change request. It MAY choose to do lookups on a less frequent scale for reconciliation.

Privacy Considerations This specification enables the sharing of information between domains. The specification assumes that implementers and deployers are operating under one of the following scenarios:

• A common administrative domain where there is one administrative owner of the data. In these cases the objective is to protect privacy and security of the owner and user data by keeping information systems co-ordinated and up-to-date. For example, the domains decide to use Domain Based Replication mode in order to keep employee information synchronized.
• In a co-operative or co-ordinated relationship, parties have decided to share a limited amount of data and or signals for the benefits of their users. Depending on end-user consent, information is shared on an as authorized and/or as needed basis. For example, the domains agree to use Co-ordinated Provision mode that exchanges things like account status, or specific minimal attribute information needed that must be fetched on request after receiving notice of a change. This enables authorization to be verified each time data is transferred.

In general the sharing of SCIM Event information falls within a pre-existing SCIM Client and Service Provider relationship. In the case of SCIM Risk Signals Events, the existing relationship may need to be reviewed. By their nature, however, SCIM Signals carry no personal information and aid parties in ensuring the protection of privacy information and account security. Privacy considerations of MUST also be observed.

IANA Considerations This section registers the schema extensions found in in the "Event" registry as per Section 4.2.

Schema URI:

See.

Schema Name:

See corresponding names under.

Intented ResourceType:

N/A. Events are not intended to be persisted in SCIM.

Purpose:

See each description in.

Single-valued Attributes:

None.

Multi-valued Attributes:

All schemas in this specification share the same attributes. See.

Summary of schema URI registrations:

Schema URI	Name	Reference
urn:ietf:params:event:SCIM:feed:add	Resource added to Feed Event
urn:ietf:params:event:SCIM:feed:remove	Remove resouce From Feed Event
urn:ietf:params:event:SCIM:prov:create	New Resource Event
urn:ietf:params:event:SCIM:prov:patch	Resource Patch Event
urn:ietf:params:event:SCIM:prov:put	Resource Put Event
urn:ietf:params:event:SCIM:prov:delete	Resource Deleted Event
urn:ietf:params:event:SCIM:prov:activate	Resource Activated Event
urn:ietf:params:event:SCIM:prov:deactivate	Resource Deactivated Event
urn:ietf:params:event:SCIM:sig:authMethod	New authentication method added
urn:ietf:params:event:SCIM:sig:pwdReset	Password Reset Event

References Normative References Amazon Coinbase Fastly Google Microsoft Coinbase Amazon Yubico This Shared Signals and Events (SSE) Framework enables sharing of signals and events between cooperating peers. It enables multiple applications such as Risk Incident Sharing and Coordination (RISC) and the Continuous Access Evaluation Profile, Microsoft Google This document defines the Continuous Access Evaluation Profile (CAEP) of the Shared Signals and Events Framework. It specifies a set of event types conforming to the SSE Framework. These event types are intended to be used between cooperating Transmitters and Receivers such that Transmitters may send continuous updates using which Receivers can attenuate access to shared human or robotic users, devices, sessions and applications. Informative References Apache Software Foundation

Use Cases SCIM Events may be used in a number of ways. The following non-normative sections describe some of the expected uses.

Domain Based Replication The objective of "Domain Based Replication" events (DBR) is to synchronize resource changes between SCIM service providers in a common administrative domain. In this mode, complete information about changes for resources are shared between replicas for immediate processing.

Domain Based Replication Sequence │ │ │ │ │ │ │ │ "SCIM Response" │ │ ┌┴┐ │ <────────────────────│ │ │ │ │ └┬┘ │ │ │ │ "Event SCIM:prov:│ │ │ │ id:xyz" │ │ │ │ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─>│ │ │ │ │ │ │ │ │ │ │ │ │ │────┐ "Update local node"│ │ │ │ │<───┘ └┬┘]]>

From a security perspective, it is assumed that servers sharing DBR events are secured by a common access policy and all servers are required to be up-to-date. From a Privacy Perspective, because all servers are in the same administrative domain, the primary objective is to keep individual service provider nodes or cluster synchronized.

Co-ordinated Provisioning In "Co-ordinated Provisioning" (CP), SCIM resource change events perform the function of change notification without the need to provide raw data. In any Event Publisher and Receiver relationship, the set of SCIM resources (e.g. Users) that are linked or co-ordinated is managed within the context of a an event feed and which MAY be a subset of the total set of resources on either side. For example, an event feed could be limited to users who have consented to the sharing of information between domains. To support capability, "feed" specific events are defined to indicate the addition and removal of SCIM resources from a feed. For example, when a user consents to the sharing of information between domains, events about the User MAY be added to the feed between the Event Publisher and Receiver.

Co-Ordinated Provisioning Sequence │ │ │ │ │ │ │ │ │ │ "SCIM Resp" │ │ ┌┴┐ │ │<──────────────│ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ │ ╔═══════╤╪═╪════════════════╪═╪═════════════════╪════╗ │ ║ LOOP ││ │ │ │ │ ║ │ ╟───────┘└┬┘ Event: │ │ │ ║ │ ║ │ SCIM:prov:│ │ │ ║ │ ║ │ id:xyz │ │ │ ║ │ ║ │ ─ ─ ─ ─ ─ ─ ─ >│ │ │ ║ │ ║ │ │ │ │ ║ │ ║ │ ╔════════════╧═╧══════════════╗ │ ║ │ ║ │ ║Receiver may accumulate ║ │ ║ │ ║ │ ║events for periodic action. ║ │ ║ │ ║ │ ╚════════════╤═╤══════════════╝ │ ║ │ ║ │ SCIM GET │ │ │ ║ │ ║ │ <───────────────│ │ │ ║ │ ║ │ │ │ │ ║ │ ║ │ Filtered │ │ │ ║ │ ║ │ Resource Resp │ │ │ ║ │ ║ │ ───────────────>│ │ │ ║ │ ║ │ │ │ │ ║ │ ║ │ │ │ "Co-ord Action" │ ║ │ ║ │ │ │ ───────────────>│ ║ │ ╚═════════╪═════════════════╪═╪═════════════════╪════╝]]>

In CP mode, the receiver of an event must call back to the originating SCIM Service Provider (e.g. using a SCIM GET request) to reconcile the newly changed resource in order to obtain the changes. Co-ordinated provisioning has the following benefits:

• Differences in schema (e.g. attributes) between domains. For example, a receiving domain may only be interested or only be allowed access to a few attributes (e.g. role based access data) to enable access to an application.
• Different Event Receivers MAY have differing needs access to information and thus be assigned varying access rights. Minimal information events combined with call-backs for data allows data filtering to be applied.
• Receivers can take independent action. For example deciding which attributes or resource lifecycle changes to accept. For example, in the case of a conflict, a receiver can prioritize one domain source over another.
• A receiver MAY throttle or buffer changes rather than act immediately on a notification. For example, for a frequently changing resource, the receiver MAY choose to make scheduled SCIM GET for resources that have been marked "dirty" by events received in the last scheduled cycle.

A disadvantage of the CP approach is that it may be considered costly in the sense that each event received might trigger a call back to the event issuer. This cost should be weighed against the cost producing filtered information in each event for each receiver. Further a receiver is not required to make a call-back on every provisioning event. It is assumed that an underlying relationship between domains exists that permits the exchange of personal information and credentials. For example the decision to perform SCIM provisioning operations at the SCIM Service Provider issuing change events, was previously authorized and appropriate confidentiality and privacy agreements have been met in cross-domain scenarios. Examples of this might be services for hire by an employer or a specific consent from an end-user as part of a online authorization where individual consent was obtained. When sharing information between parties, CP Events minimize the information shared in each message requiring the Security Event Receiver to call back to the event publisher to retrieve more information if required. In this way, the Event Publisher is able to regular access to information through normal SCIM protocol access restrictions.

Risk Signals The sharing of risk signals (RS) is used for the purpose of co-ordinating account change events between a SCIM Service Provider and another related security service. For example, when a password or other authentication factor has changed, a receiving security system can choose to terminate current User sessions to force a re-authentication against the modified User resource. These signals MAY also include those described in the OpenID Shared Signals Working Group Specifications. These events are intended for receivers where there is a prior relationship on behalf of the users described in the SCIM Service Provider. The intent of sharing information about security events is for the purpose of securing a user account and ensuring privacy.

Async Requests A SCIM provisioning client MAY wish to request "asynchronous" processing using the "Prefer Header for HTTP", Section 4.1 . In this mode, a normal SCIM protocol POST, PUT, PATCH, or DELETE request is made, and the HTTP Header Prefer is included with the value respond-async. When a SCIM Client signals respond-async, the SCIM server response changes to HTTP Status 202 Accepted as defined in . The Location header returned is the final resource location and no payload is present. Following acceptance of an asynchronous request, a notification of completion can be issued using the Async Event Notification per . The location returned SHALL correspond to the sub claim in the future Async Event SET message.

Asychronous Request Sequence │ │ │ │ │ │ │ "Accepted Status 202" │ │ │ "Location: /Users/xyz" │ │ │ <──────────────────────────────────────────────────│ │ │ │ │ └┬┘ │ │ │ "Event SCIM:misc:asyncResp │ │ │ │ sub: /Users/xyz │ │ │ │ id:xyz, method: PUT" │ │ │ │ <─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─│ │ └┬┘ │ │ "ID: xyz created" │ │ │ <─ ─ ─ ─ ─ ─ ─ ─ ─ ─ ─ │ ]]>

Acknowledgments Thanks to Morteza Ansari who contributed significantly to draft-hunt-idevent-scim-00, upon which this draft is based. The editor would like to thank the participants in the SCIM working group and the id-event list for their support of this specification.

Change Log Draft 00 - PH - First Draft