DLP

Outside supplier

1. Compose. M103 holds the labeled roster attachment for the outside supplier. No rule evaluation or delivery has occurred. Maya’s ability to read R101 for staffing is assumed and does not authorize every destination. X101 is an unchanged independent copy with a trusted Confidential label. The primary recipient is print@example.test; the comparison recipient is staffing@example.test. The model covers a corporate-mail send, not a clipboard, personal device or every possible application.
2. Inspect. P103 matches the Confidential attachment and outside recipient. The matching condition alone has not yet blocked or delivered the message. This supplied label and exact destination are the configured inputs, not the output of a universal content-recognition algorithm. The policy matches Confidential plus a destination other than the approved staffing mailbox. The same attachment to the approved mailbox does not meet that blocking condition. Evaluation produces a result while the send is still held; enforcement is a separate action.
3. Enforce. The configured blocking action stops M103 before delivery. V103 records the decision; no attachment reaches the outside mailbox. The actual send gate rechecks the policy before creating any mailbox copy, so a fabricated allow decision cannot override it. The blocked path remains unsent and records message, policy, recipient and decision in V103 without copying the roster payload. The permitted path queues delivery but still has no received message. The sender’s file stays unchanged, and recording a decision does not complete an analyst review.
4. Check inbox. An independent destination read returns 404 for M103. The blocked send has created no roster copy there; the sender still has X101. In the allowed path, the queued send first creates an independent destination copy, then the mailbox read returns HTTP 200 with the same roster row. In the blocked path, no send runs and the read returns a bodyless 404 with no message or attachment. Queuing was not treated as receipt. These simulated mailbox outcomes demonstrate P103’s selected behavior, without a live SMTP or cloud-service test. V103, R101 and sender-held X101 remain, while analyst review and retention decisions stay open.

Approved HR desk

1. Compose. M103 holds the same labeled roster attachment for the approved HR mailbox. No rule evaluation or delivery has occurred. Maya’s ability to read R101 for staffing is assumed and does not authorize every destination. X101 is an unchanged independent copy with a trusted Confidential label. The primary recipient is print@example.test; the comparison recipient is staffing@example.test. The model covers a corporate-mail send, not a clipboard, personal device or every possible application.
2. Inspect. The Confidential attachment is going to the approved HR mailbox. P103’s outside-recipient condition does not match; nothing has been delivered yet. This supplied label and exact destination are the configured inputs, not the output of a universal content-recognition algorithm. The policy matches Confidential plus a destination other than the approved staffing mailbox. The same attachment to the approved mailbox does not meet that blocking condition. Evaluation produces a result while the send is still held; enforcement is a separate action.
3. Enforce. The gate permits M103 and queues delivery to the approved HR mailbox. V103 records the decision; queuing is not receipt. The actual send gate rechecks the policy before creating any mailbox copy, so a fabricated allow decision cannot override it. The blocked path remains unsent and records message, policy, recipient and decision in V103 without copying the roster payload. The permitted path queues delivery but still has no received message. The sender’s file stays unchanged, and recording a decision does not complete an analyst review.
4. Check inbox. Delivery creates a mailbox copy. An independent read then returns M103 with the unchanged roster row. The sender’s original attachment remains. In the allowed path, the queued send first creates an independent destination copy, then the mailbox read returns HTTP 200 with the same roster row. In the blocked path, no send runs and the read returns a bodyless 404 with no message or attachment. Queuing was not treated as receipt. These simulated mailbox outcomes demonstrate P103’s selected behavior, without a live SMTP or cloud-service test. V103, R101 and sender-held X101 remain, while analyst review and retention decisions stay open.

This is a configured label-and-recipient rule in one simulated corporate-mail channel, not a sensitive-content detector or a universal DLP flow. It has no override. Other channels, missing or incorrect labels, transformations and legitimate exceptions need separate tests. Decision V103 is recorded without the payload; analyst review and retention remain unfinished.

Flow references: Microsoft Purview: DLP conditions, actions and channel scope · CIS Control 3: Data Protection · NIST SP 800-53 Rev. 5: processing purposes and security attributes