SaaS ✓ fields verified against sample logs

Microsoft 365 Unified Audit Log Fields: Reference With Examples

Published

One JSON record per user or admin activity in Exchange, SharePoint, OneDrive, Entra ID and other Microsoft 365 services: who (UserId), what (Operation, with the cmdlet's Parameters for Exchange admin actions), on which object, from which address, and when (UTC).

Official documentation: Microsoft: Office 365 Management Activity API schema · Microsoft: Office 365 Management Activity API reference

Sample record

{
"CreationTime": "2024-03-15T09:00:00",
"Id": "m365-b001",
"Operation": "FileAccessed",
"OrganizationId": "corp-tenant",
"RecordType": 6,
"ResultStatus": "Succeeded",
"UserKey": "[email protected]",
"UserType": 0,
"Version": 1,
"Workload": "SharePoint",
"ClientIP": "203.0.113.42",
"UserId": "[email protected]",
"ObjectId": "https://corp.sharepoint.com/sites/Marketing/Shared Documents/Campaign-Brief.docx"
}

Numbered lines are the fields in the table.

From the LogTriage parser test corpus — synthetic data, no real hosts or users. The first record of the sample, pretty-printed.

Explain this record field by field →

Show the sample lines as they appear in the file
{"CreationTime":"2024-03-15T09:00:00","Id":"m365-b001","Operation":"FileAccessed","OrganizationId":"corp-tenant","RecordType":6,"ResultStatus":"Succeeded","UserKey":"[email protected]","UserType":0,"Version":1,"Workload":"SharePoint","ClientIP":"203.0.113.42","UserId":"[email protected]","ObjectId":"https://corp.sharepoint.com/sites/Marketing/Shared Documents/Campaign-Brief.docx"},

Field reference

Field Type Meaning Why it matters for security
1CreationTimee.g. 2024-03-15T06:00:00 date-time Microsoft: "The date and time in Coordinated Universal Time (UTC) when the audit log record was generated." UTC even though the value has no Z; add one before parsing, or a library may read it as local time.
2Ide.g. m365-001 GUID Microsoft: "Unique identifier of an audit record." Use it to remove duplicates when the same content blob is downloaded twice.
3Operatione.g. New-InboxRule string Microsoft: "The name of the user or admin activity." For Exchange admin activity, "the name of the cmdlet that was run." The field most hunts start with — New-InboxRule, Set-Mailbox, MailItemsAccessed, FileDownloaded, SharingInvitationCreated, Add member to role.
4RecordTypee.g. 1 number Microsoft: "The type of operation indicated by the record." 1 is ExchangeAdmin, 2 ExchangeItem, 6 SharePointFileOperation, 8 AzureActiveDirectory, 14 SharePointSharingOperation, 15 AzureActiveDirectoryStsLogon, 50 ExchangeItemAggregated (MailItemsAccessed). Decides which service-specific fields the record has; filter on it before reading Parameters or ObjectId.
5Workloade.g. Exchange string Microsoft: "The Office 365 service where the activity occurred." Grouping one user's records by workload shows a session moving from mail to files to directory changes.
6UserTypee.g. 2 number Microsoft: "The type of user that performed the operation." 0 is Regular ("A regular user without admin permissions"), 2 is Admin. Admin-type records against a normal user's mailbox, or a regular user running admin cmdlets, need explaining.
7UserIde.g. [email protected] string Microsoft: "The UPN (User Principal Name) of the user who performed the action." System accounts and app@sharepoint also appear. The actor. A compromised account shows as itself, so look at ClientIP and the sequence of operations rather than the name.
8UserKeye.g. [email protected] string Microsoft: "An alternative ID for the user identified in the UserId property" — the passport unique ID (PUID) for SharePoint, OneDrive and Exchange events. Stable when a user is renamed; a UserKey that differs from UserId can mean an admin acting on someone else's mailbox, as in the malicious sample's Set-Mailbox.
9ClientIPe.g. 185.220.101.45 IP address Microsoft: "The IPv4 or IPv6 address of the device that was used when the activity was logged." For some services it is the address of a Microsoft application acting for the user. The main pivot — the same unknown address behind a sign-in, an inbox rule and file downloads within minutes is the business email compromise pattern.
10ResultStatuse.g. Succeeded string Microsoft: "Possible values are Succeeded, PartiallySucceeded, or Failed. For Exchange admin activity, the value is either True or False." For Entra ID STS logon records Microsoft warns that Succeeded "indicates only that the HTTP operation was successful; it doesn't mean the logon was successful" — read LogonError.
11ObjectIde.g. https://corp.sharepoint.com/sites/Finance/Shared Documents/2024-Budget.xlsx string Microsoft: "For SharePoint and OneDrive activity, the full path name of the file or folder accessed by the user. For Exchange admin audit logging, the name of the object that was modified by the cmdlet." Many FileDownloaded records across finance and HR sites from one session is bulk collection.
Parameters.Namee.g. ForwardingSmtpAddress string The name of a parameter passed to the Exchange cmdlet named in Operation; each Parameters entry is a Name/Value pair. ForwardTo, RedirectTo, ForwardingSmtpAddress and DeleteMessage are the parameters that turn an inbox rule or a mailbox setting into mail theft.
Parameters.Valuee.g. [email protected] string The value given for that parameter. An external address here is the evidence; check whether the domain belongs to your organization.
OperationProperties.Namee.g. MailAccessType string Name of an additional property of the operation, as Name/Value pairs — for MailItemsAccessed, MailAccessType (Bind or Sync) among others. Bind records single messages being read; Sync records a client downloading a whole mail folder, so a Sync from an unknown address can mean the folder's contents left the tenant.
ModifiedProperties.Namee.g. Role.DisplayName string Name of a property the operation changed, on Entra ID directory records; NewValue and OldValue give the change. On Add member to role it names the role that was granted — a privileged role given to a new account right after a suspicious sign-in is persistence.

Have one of your own records? Paste it into the log line explainer and each field is explained from this table, with times, codes and addresses decoded — in your browser.

What it records

The unified audit log collects user and administrator activity from Microsoft 365 services into one stream of JSON records. Every record carries the common schema Microsoft documents in the Management Activity API schema: Id, CreationTime, Operation, RecordType, Workload, UserId, UserKey, UserType, ClientIP, OrganizationId and, usually, ResultStatus and ObjectId. On top of that each record type adds its own fields: Exchange admin records carry the cmdlet’s Parameters, Entra ID records Target and ModifiedProperties, MailItemsAccessed records OperationProperties and Folders.

The malicious sample shows business email compromise end to end from one address: New-InboxRule forwarding mail out, Set-Mailbox with an external ForwardingSmtpAddress, three FileDownloaded records across finance and HR sites, a SharingInvitationCreated to a guest, Add member to role for a new account and MailItemsAccessed reading mailbox content.

Where to get it

  • Microsoft Purview audit search. Search by user, activity and date range, and export the result as CSV. The CSV puts the whole JSON record in an AuditData column, which is what most investigations end up parsing.
  • PowerShell. Search-UnifiedAuditLog in Exchange Online PowerShell returns the same records, with AuditData holding the JSON.
  • Office 365 Management Activity API. Subscriptions per content type (Audit.Exchange, Audit.SharePoint, Audit.AzureActiveDirectory, Audit.General) deliver blobs of these records; SIEM connectors use this route.
  • Microsoft Sentinel. The Office 365 connector writes them to the OfficeActivity table under different column names.

Parsing pitfalls

  • Times without a zone. CreationTime is UTC but written without Z.
  • The CSV wraps the JSON. In a Purview export the useful fields are inside the AuditData column as a JSON string, not separate columns.
  • Name/Value lists. Parameters, OperationProperties and ModifiedProperties are arrays of pairs; turn them into a map before searching, or a rule on ForwardTo will miss it.
  • ResultStatus differs by workload. Exchange admin records use True/False, others Succeeded/PartiallySucceeded/Failed, and for STS logons it is only the HTTP result.
  • ClientIP is not always the user. Microsoft notes it can be the address of a Microsoft application calling the service on the user’s behalf, and that some Entra ID records leave it empty.
  • Delay. Records appear in search after a delay that varies by service, so a search right after an incident can be incomplete.

Detection ideas

  • Mail forwarding: New-InboxRule or Set-InboxRule with ForwardTo, RedirectTo or DeleteMessage, and Set-Mailbox with an external ForwardingSmtpAddress — the inbox rule rule.
  • One address, many steps: the same ClientIP behind a sign-in, a mailbox change and file downloads within an hour, from an address the user has not used before — the pattern in the business email compromise use case.
  • Bulk collection: many FileDownloaded or FileSyncDownloadedFull records for one user across sites, or MailItemsAccessed with MailAccessType Sync.
  • External sharing: SharingInvitationCreated or anonymous links to addresses outside the tenant shortly after unusual activity.
  • Privilege changes: Add member to role granting an admin role, especially to a new or guest account.

The free browser analyzer reads unified audit log exports, including the Purview CSV, and runs these checks locally.

Sending Microsoft 365 Unified Audit Log to an LLM

One megabyte of this format is about 497,530 tokens on current Claude models and 301,091 on GPT models, measured with the models' own tokenizers — so a gigabyte sent raw to Claude Sonnet 5.5 costs about $995 in input tokens alone (prices of 2026-10-24). Estimate your own volume →

Detections and guides for Microsoft 365 Unified Audit Log

Frequently Asked Questions

Is CreationTime in UTC?
Yes. Microsoft defines it as the time in Coordinated Universal Time when the record was generated, even though the value has no trailing Z. Parsers that treat a time without a zone as local will shift every record by the analyst's offset, so append the Z or set UTC explicitly.
Why does a failed sign-in show ResultStatus Succeeded?
For Entra ID secure token service (STS) logon records, Microsoft says Succeeded only means the HTTP operation worked; it does not say whether the logon did. Whether the user actually signed in is in the LogonError property of the same record. Do not count sign-in failures from ResultStatus alone.
What is the difference between UserId and UserKey?
UserId is the UPN of the person or account that performed the action. UserKey is an alternative identifier — Microsoft documents it as the passport unique ID for SharePoint, OneDrive and Exchange events — and it does not change when a user is renamed. When an administrator acts on another user's mailbox, the two can point at different people.
Where are the details of an Exchange admin action?
In Parameters: a list of Name/Value pairs holding the arguments of the cmdlet named in Operation. A New-InboxRule record's Parameters carry the rule name and fields such as ForwardTo or DeleteMessage; a Set-Mailbox record's carry ForwardingSmtpAddress. The Operation alone does not tell you whether the change was harmful.

Analyze a Microsoft 365 Unified Audit Log file in your browser

The free analyzer detects this format, parses every field above and flags attack patterns — the file never leaves your machine. For threat-intel enrichment and an AI-written incident report, run it through LogTriage.

Investigating something right now? LogTriage checks every address in the file against threat intelligence and writes the incident report: 10 analyses a month free. The incident pack adds the in-depth AI analysis and files up to 100 MB: 25 analyses for $19, one payment, no subscription. Pricing

How this page was made. The official definitions, field names and examples on it are copied by script from the sources it links and re-checked against them on every build. The investigation advice was drafted with AI from those sources and passed a separate review before publishing. What is checked, and how · found a mistake? [email protected]