Learn › SharePoint & OneDrive › 6 · Security & external access

DLP for SharePoint and OneDrive

DLP scans SharePoint and OneDrive content for sensitive information and acts on it — blocking sharing, warning users, restricting access. The catch is timing and coverage: DLP sees what the index sees, so it protects indexed content, not the moment of upload.

What DLP does here

Data Loss Prevention (authored in Microsoft Purview) applies policies to SharePoint and OneDrive locations that detect sensitive content and take action. Detection is by:

  • Sensitive information types (SITs) — patterns like credit-card, national ID, bank details, with confidence and count thresholds.
  • Sensitivity labels — a DLP rule can key off the label on a file (the labels-on-sites concept), so "if labelled Confidential, block external sharing" becomes enforceable.
  • Trainable classifiers / keywords / document fingerprints for less structured content.

Actions: block or restrict external (or all) sharing, show policy tips to the user, notify/alert admins, and restrict access to the content.

The scan timing (the crucial caveat)

DLP for SharePoint/OneDrive evaluates content that has been crawled and indexed (search concept). The implication that trips people up:

  • It's not a synchronous gate at upload. A freshly uploaded file with sensitive content may not be caught for a short window until it's indexed and the policy evaluates it.
  • Sharing actions are evaluated closer to real time (a block-external-sharing rule can prevent the share), but at-rest detection of already-stored content depends on the crawl.

So DLP is strong at governing already-indexed content and sharing events, weaker as an instantaneous upload firewall — plan it as detect-and-remediate, not an airtight door.

The failure modes

  • "It didn't catch my test file immediately." Indexing latency — expected; re-test after the content is indexed.
  • Encrypted/unreadable content is invisible. DLP inspects text the service can read; an encrypted file (or an image-only PDF without OCR) may not match a SIT — the same partially-indexed gap eDiscovery has.
  • False positives from loose SITs — low thresholds flag benign content; tune confidence and instance counts.
  • Label vs content precedence — a label-based rule and a content-based rule can both apply; understand which action wins (typically the most restrictive).
  • Policy scope gaps — a policy scoped to specific sites misses content elsewhere; OneDrive is a separate location that must be explicitly included.
  • Policy tips need supported clients — tips surface in the web/Office experience; not every access path shows them.

What to watch (proofs)

  • Policy coverage: Purview → Data loss prevention → policies — which locations (SharePoint, OneDrive) each policy covers and its status (test vs enforce).
  • Matches and actions: the DLP activity/alerts and Activity Explorer — what was detected, where, and whether it was blocked or tipped.
  • The timing reality: upload a test file with a known SIT, then observe the match appear after indexing — proof of the crawl-dependent behaviour.
  • Label interplay: a DLP rule conditioned on a sensitivity label firing on labelled content — confirms labels and DLP are wired together.

PowerShell for this concept

Discussion

No messages yet — start the thread.

Sign in with your email to join the discussion — we send a one-time link, no password.