| name | purview-advanced-dlp |
| description | Guidance for advanced Microsoft Purview DLP capabilities - Adaptive Protection (risk-based DLP driven by Insider Risk), Exact Data Match, contextual conditions, and advanced Endpoint DLP egress controls. Covers dynamic, risk-adaptive enforcement and high-precision matching for mature programs. WHEN: advanced DLP, Adaptive Protection, risk-based DLP, Exact Data Match DLP, dynamic DLP enforcement, Insider Risk driven DLP, advanced Endpoint DLP, contextual DLP conditions, dynamic block actions. |
| license | MIT |
| metadata | {"author":"Microsoft","version":"0.1.0"} |
Microsoft Purview Advanced DLP
Advanced DLP extends baseline policies with risk-adaptive and high-precision controls,
notably Adaptive Protection, which dynamically adjusts DLP enforcement based on a user's
Insider Risk level - so low-risk users get light controls and elevated-risk users get
stricter actions automatically.
When to use
Moving from static DLP to dynamic, context-aware enforcement and high-precision matching for
mature data security programs that already have classification and baseline DLP working.
Do not use this skill to design a first DLP policy from scratch (use purview-dlp-policy) or to
stand up Insider Risk Management (use insider-risk-baseline).
Pick the right advanced capability
| Goal | Use |
|---|
| Dynamic enforcement keyed to user risk | Adaptive Protection (IRM-driven) |
| High-precision match against your own records | Exact Data Match (EDM) SITs |
| Context-aware rules (label + recipient + volume) | Contextual conditions in DLP rules |
| Block USB / unsanctioned cloud / clipboard | Advanced Endpoint DLP egress controls |
| Stop paste into ChatGPT / unsanctioned AI | Endpoint DLP + DSPM for AI |
Rule of thumb: pick one advanced capability per quarter; layering Adaptive Protection, EDM, and
new Endpoint controls simultaneously makes failure attribution impossible.
Approach
- Confirm prerequisites - Baseline DLP in enforce mode for at least one workload; Insider
Risk Management running with tuned indicators (for Adaptive Protection).
Verify: IRM is producing meaningful risk levels for a pilot population.
- Adaptive Protection - Integrate Insider Risk Management risk levels (minor/moderate/
elevated) with DLP so low-risk users get light controls and elevated-risk users get stricter
actions (e.g., block) automatically - minimising friction while focusing protection.
Verify: a test user moved to elevated triggers the stricter DLP rule branch.
- Exact Data Match (EDM) - Use EDM SITs to match against a hashed copy of your structured
data (e.g., customer records) for high-precision, low-false-positive detection.
Verify: EDM schema definition uploaded, hash refresh succeeded, and test sample matches.
- Contextual conditions - Combine sensitivity labels, classifiers, content volume,
recipient, and access scope for precise rules (e.g., "Confidential + external recipient +
≥5 SIT matches").
Verify: test events show the rule fires only on the intended combinations.
- Endpoint egress controls - Restrict copy to USB, upload to unsanctioned cloud/AI apps,
network share, clipboard, and printing on managed endpoints.
Verify: a test paste of labelled content into an unsanctioned AI app is blocked or flagged.
- Tune continuously - Use Activity Explorer and alerts to refine; review override
justifications weekly.
Verify: false-positive rate trends down month over month.
Guardrails
- Adaptive Protection requires Insider Risk Management configured and consented - plan privacy
and pseudonymisation with legal/works councils before enabling.
- EDM requires schema definition, a hashed data upload, and a dedicated upload role - this is
not a quick win and needs data-engineering ownership.
- Validate dynamic enforcement in audit/simulation before enabling block actions - dynamic block
in production with no audit phase is a guaranteed incident.
- Endpoint DLP advanced controls (clipboard, USB, network share) need onboarded devices and
E5 Compliance; mac and Windows behaviour can differ.
- Keep an exception path - business processes break; an override workflow with justification is
required.
Common anti-patterns
- Turning Adaptive Protection straight to "block on elevated" before IRM thresholds are tuned -
blocks innocent users on noisy indicators.
- EDM with a 50-column schema and no refresh job - matches go stale silently.
- Stacking contextual conditions until nothing matches, then declaring DLP "doesn't work".
- Blocking clipboard tenant-wide without piloting on knowledge workers.
- No metrics on override frequency - tuning becomes guesswork.
Example prompts
Configure risk-based DLP driven by Insider Risk Adaptive Protection.
Set up Exact Data Match (EDM) for precise DLP detection.
How do I add contextual conditions to advanced Endpoint DLP?
Validate dynamic DLP enforcement in simulation before blocking.
Block clipboard paste of Highly Confidential into unsanctioned AI on managed endpoints.
Microsoft Learn