Auditing

ESS services support the auditing of their activities.

Note

To ensure that audit messages are delivered, the Audit service may send the message more than once.

For instance, multiple instances of the Audit service may exist when applying configuration changes or when performing a rolling upgrade. During this time, duplicate messages (i.e., messages with the same unique id value) can occur if the old Audit instance processes a message but shuts down before acknowledging the message on Kafka, and then, the new Audit instance processes the unacknowledged message.

Audit Events

The following events are audited:

Services	Event Name	Notes
Most Services [4]	service-started service-shutdown	Service Startup/Shutdown.
Purgeable services (i.e. Access Grant, Authorization, Broker, Pod Provision, and WebID)	purge-init purge-completed	Purge start/complete.
Access Grant Service	access-grant-created access-grant-queried access-grant-read access-grant-revoked access-grant-verified access-request-created access-request-read access-request-revoked access-request-verified access-denial-created access-denial-read access-denial-revoked access-denial-verified request-authorized [1] revocation-status-read [2]	Access Request/Grant/Denial lifecycle events. Added in version 2.2.0: [3] [1] request-authorized events contain additional information (such as the access grant service endpoint, the WebID, client id, etc.) for authenticated access requests/grants/denial events. For these events, you can find the corresponding request-authorized event using the instrument field. However, other than the service endpoint, the same information may be found in the 2.2 access-* event messages themselves. [2] As part of the access grant to UMA access token exchange (which is an unauthenticated event), revocation-status-read events occur. These revocation-status-read events, which are also unauthenticated, do not have an associated request-authorized event. [3] Access Request/Grant/Denial lifecycle events are also available starting in 2.1.5 (i.e., the feature has been backported to 2.1.5). However, the 2.1.5 events do not include the various Pod/agent/client information enhancements that are available starting in 2.2.
Authorization Service	acr-created acr-updated acr-deleted provisioned-pod-access-control deprovisioned-pod-access-control	ACR Lifecycle events.
Pod Storage Service	resource-created resource-updated resource-deleted	Resource Lifecycle events.
	resource-read	Successful resource read events (GET and HEAD operations). To enable, see INRUPT_STORAGE_AUDIT_RESOURCE_READ_ENABLED configuration for Pod Storage Service. Added in version 2.1.
Pod Provision Service	pod-provisioned	Pod Provisioned event.
Purger application	purge-started purge-completed purge-failed	Purge start/end (successful or not).
Query Service	query-succeeded	Query events.
Query Indexer	ingest-succeeded ingest-failed	Query Indexer events.
Solid OIDC Broker Service	openid-backend-idp-login openid-token-requested openid-authorization-initialized	Authentication/Authorization flow.
UMA Service	uma-token-created	UMA Grant Flow.
WebID Service	webid-created webid-updated webid-deleted	WebID Profile events.

[4]

The following services do not produce audit events:

• Notification Gateway service

• WebSocket Notification service

• Start service

Note

ESS uses asynchronous messaging for auditing.

Audit Event Message Internal Format

Internally, ESS’ audit event messages are in RDF and serialized as ActivityStreams 2.0 JSON-LD documents:

Note

Although the following document shows all possible fields for an event message, the specific events determine which fields appear.

Starting in version 2.2, Pod-related event messages include:

• the actor information (WebID)

• the application information (Client ID).

• the Pod information (Pod root url and Pod data subject)

{
   "@context":[
      "https://www.w3.org/ns/activitystreams",
      "https://schema.inrupt.com/audit/v1.jsonld"
   ],
   "id":"<UUID of the event>",
   "type": [ "Activity", <type2>,... ],
   "name":"<event name>",
   "summary": "<event description>",
   "generator": <JSON document identifying the software producing the event>,
   "actor": [ <JSON document identifying the actor associated with the event>, ... ],
   "object": [ <JSON document identifying the object associated with the event>, ... ],
   "instrument": [ <JSON document identifying the client/application associated with the event>,
                   <JSON document with associated OpenTelemetry data>,
                   <JSON document with associated application-defined metadata>,
                   <JSON document identifying the Pod root and data subject>, ... ],
   "result": [ <JSON document containing associated results for the event, if any> ],
   "published": "<datetime>",
   "identifier":"<identifier to use for correlated events>"
}

@context	Specifies the JSON-LD contexts.
id	Universally Unique IDentifier (UUID) for the event.
type	An array of event types; e.g., [ "Activity", "Create" ].
name	Name that denotes the event; e.g., service-started, openid-backend-idp-login, resource-created, access-grant-created, etc. See Audit Events for a list of audited events names.
summary	Short description associated with the message name.
generator	JSON document identifying the software (e.g., service) producing the event. For example, the generator for a Pod provision event: "generator": { "id": "<service URL>" "type": ["SoftwareApplication"], "name": "<application name>", "qualifiedAssociation": "<processId>", "wasAssociatedWith": "<Kubernetes pod name>" }
actor	An array of JSON documents that identify the agents, if any, associated with the event. The actor’s identity can be denoted by various combination of fields, such as (list below is not exhaustive): id and type fields; name and type fields. The field can also be empty [] for events not initiated by a user (such as service start events, etc.). Note Starting in version 2.2, id and name fields are associated with a type field. In earlier versions, the id field and name field may appear without a clarifying type. For example: For an Access Request/Grant event, which include the id and the type: "actor" : [ { "id" : "https://id.example.com/someusername", "type" : [ "Agent" ] } ], For IdP login events: "actor": [ { "name": "someusername", "type" : [ "Agent" ] } ]
object	An array of JSON documents that identify the objects associated with the event; that is, the object(s) against which the action is performed. The object can be denoted by various combination of fields, such as (list below is not exhaustive): id and name fields; id and type fields; name and qualifiedAssociation fields; For example: A Pod provisioned event: "object": [ { "type": [ "Storage" ], "id": "<Pod Root URL>" } ] For access request/grant/denial creation events (available starting in version 2.1.5), the object field contains the created access request/grant/denial: Starting in 2.2, the object for the create access request/grant/denial events contains the created request/grant/denial in a document with: the content field that contains the access request/grant/denial as string and the mediaType field. Previously, the object contained the created access request/grant/denial directly as an element. Starting in 2.2, the object contains a new element that identifies the Pod resource. "object": [ { "content" : "{\"id\":\"https://vc.example.com/vc/79288a3 ... }", "mediaType": "application/ld+json" }, { "type" : [ "Resource" ], "id" : "https://storage.example.com/ad3b.../some/resource" } ] For Pod resource CRUD events ( resource-created, resource-read, resource-updated, and resource-deleted) events, the object field include an object that identifies the Pod resource. [5] For example: [6] "object": [ { "generated" : "6cba8240-9f79-40db-a129-c6b8edddb840", "invalidated" : "4e12ed06-c4e5-41bc-bcd3-0f6a821c08ca", "type" : [ "Resource" ], "id" : "https://storage.example.com/ad3b.../some/resource" } ] [5] Starting in 2.2.0, CRUD events no longer include an object that identifies the Pod data subject. See the instrument field instead for this information. [6] The generated and invalidated fields are internal references to the content that appear in some combination depending on whether it’s a create/read/update/delete event. The object may also include additional fields, such as accessControl for resource-created events.
instrument	An array of JSON documents that identify: The clients associated with the event, if any. For example: "instrument" : [ { "id" : "https://start.example.com/app/id", "summary" : "Client identifier", } ], Note Starting in version 2.2, instrument objects that identify the clients have an associated summary field with the value "Client identifier". In earlier versions, the id field may appear without a clarifying summary or may have included "name" : "Client Application" instead of summary. Some events such as service-started do not have associated clients, and thus may have empty instrument array. The associated OpenTelemetry instrument info (New in version 2.2) For example: "instrument" : [ // ... { "traceId" : "7decd3657a9efffc010a4b6a4b3da586", "spanId" : "91123fce3c668451", "isSampled" : true, "name" : "OpenTelemetry Span Context", "type" : [ "SpanContext" ] }, // ... ], OpenTelemetry traceId field can be used to correlate messages associated with a request. See Appendix: Audit Event Correlation for more information. The associated Pod information (New in version 2.2): hasDataSubject contains the Pod data subject information (Pod Data Subject refers to the agent who created the Pod. ) hasStorage contains the Pod root URL. For example: "instrument" : [ // ... { "hasDataSubject" : { "id" : "https://id.example.com/someusername", "type" : [ "https://w3id.org/dpv#DataSubject" ] }, "hasStorage" : "https://storage.example.com/root/", "type" : [ "http://www.w3.org/2004/02/skos/core#Concept" ] } // ... ], Application-defined metadata sent in client requests (New in version 2.2) For example: "instrument" : [ // ... { "name" : "Application-Defined Request Metadata", "items": [ { "mediaType" : "text/plain", "name" : "x-correlation-id", "content" : "2049875809728750827498245084" }, { "mediaType":"text/plain", "name":"my-client-version", "content":"1.0.3" } ], "type":[ "urn:uuid:1a05e301-4013-40c9-bae7-5d719b7151c8" ] } // ... ], See Application-Defined Metadata for the configuration needed to include in audit events.
result	An array of JSON documents that contains associated results. For example: an access-request-verified event includes the results of the verification, or an access-grant-revoked event includes the updated status. Added in version 2.1.5.
published	The timestamp of the event.
identifier	Identifier to use for correlated events from a single service that have occurred within the same request. To correlate events across services for a request, use the OpenTelemetry traceId in the instrument field.

For examples, see Appendix: Audit Events Examples.

Integration with External Logging Systems

The ESS Auditing service can log to:

• sysout (default) [7]

• Syslog

• Microsoft Sentinel.

[7]

When logging audit messages to sysout (the default), Auditing service uses the shared ESS logging components (available starting in 2.2.0), which use JSON as the default formatting for log messages. As such, starting in 2.2.0, audit messages logged to sysout are contained in the auditEvent field of the log message. The audit event message itself (i.e., the auditEvent value in v2.2+) is serialized as a JSON-LD documents (as in earlier releases).

Syslog

When integrating with Syslog, ESS audit events are converted to Syslog message format:

<priority>version timestamp hostname service processId messageId message

Where:

• hostname, service, processId and messageId values are taken extracted from the audit event agent field.

• message is the full ESS audit event in JSON.

For more information, see Integration with Syslog and

Microsoft Sentinel

When integrating with Microsoft Sentinel, the ESS audit events are converted into a Sentinel-specific format and POST’ed to the Sentinel service.

For more information, see Integration with Sentinel.