Microsoft 365 collaboration invite lure

Severity highCloud & SaaSThresholdstablecloud.identity.m365_collaboration_invite_lure · v1.0.0

A burst of collaboration invitations / mentions / comment notifications sent from one mailbox — the propagation phase of a business email compromise.

What it detects

A burst of collaboration invitations / mentions / comment notifications sent from one mailbox — the propagation phase of a business email compromise.

Once an account is taken over, the attacker stops reading and starts sending. A file-share or project invitation from a real, already-trusted colleague is a far better lure than a spoofed external email: it survives SPF/DKIM/DMARC because it genuinely is that account, it arrives with the sender's real display name, and the recipient has no reason to doubt it. The link in the invite leads to the credential-harvest or consent page, and the compromise spreads — often to other organisations, which turns an internal incident into a notifiable one.

This is the step KiroSec was blind to: on a real BEC case 317 such invitations went out from a compromised mailbox — 379 of the sending events came from the attacker's own IP — and not one rule fired, even while 41 separate recipients reported those exact messages as phishing.

MITRE ATT&CK mapping

  • T1534Internal Spearphishing· Lateral Movement
  • T1566.002Phishing: Spearphishing Link· Initial Access

False positives and triage

Collaboration invitations are ordinary work: onboarding a team, kicking off a project, or a chatty reviewer commenting through a document will all clear this threshold legitimately. What makes it a finding is the surrounding behaviour — the same account also showing mailbox access bursts, a hiding inbox rule, mass deletion, or send-as impersonation; invitations going to recipients outside the organisation; or the sending client/IP differing from that account's norm. Check whether the invitations name a project the account actually works on, and whether any recipient reported them. Treat a lone burst with no other signal as context, not a verdict.

The rule

Declarative YAML read by the detection engine — data, never executed. Confidence 60%. Source: rules/cloud/m365_collaboration_invite_lure.yaml, licensed under the Apache License 2.0.

id: cloud.identity.m365_collaboration_invite_lure
version: 1.0.0
status: stable
severity: high
confidence: 0.6
platforms: [application]
mitre_techniques: [T1534, T1566.002]
description: |
  A burst of collaboration invitations / mentions / comment notifications sent from
  one mailbox — the propagation phase of a business email compromise.

  Once an account is taken over, the attacker stops reading and starts sending. A
  file-share or project invitation from a real, already-trusted colleague is a far
  better lure than a spoofed external email: it survives SPF/DKIM/DMARC because it
  genuinely is that account, it arrives with the sender's real display name, and
  the recipient has no reason to doubt it. The link in the invite leads to the
  credential-harvest or consent page, and the compromise spreads — often to other
  organisations, which turns an internal incident into a notifiable one.

  This is the step KiroSec was blind to: on a real BEC case 317 such invitations
  went out from a compromised mailbox — 379 of the sending events came from the
  attacker's own IP — and not one rule fired, even while 41 separate recipients
  reported those exact messages as phishing.
match:
  provider: Office365
  event_action: Create
  raw_message_contains_any:
    - "invited you to a project"
    - "added you to a project"
    - "mentioned you in a Project"
    - "shared a file with you"
    - "shared with you"
    - "added a comment"
threshold:
  count: 5
  window_seconds: 3600
  group_by: [user_name]
correlation_keys: [user_name]
false_positive_guidance: |
  Collaboration invitations are ordinary work: onboarding a team, kicking off a
  project, or a chatty reviewer commenting through a document will all clear this
  threshold legitimately. What makes it a finding is the surrounding behaviour —
  the same account also showing mailbox access bursts, a hiding inbox rule, mass
  deletion, or send-as impersonation; invitations going to recipients outside the
  organisation; or the sending client/IP differing from that account's norm. Check
  whether the invitations name a project the account actually works on, and whether
  any recipient reported them. Treat a lone burst with no other signal as context,
  not a verdict (ADR-004).

Related detections

Run this rule on your own logs — offline

KIROSEC runs this rule and the rest of the library over Windows, Linux, macOS and cloud audit logs on the analyst's own machine. Every alert links to the events that raised it, and nothing leaves the device.

Browse all detections