You tune a Microsoft Sentinel detection rule by starting from the incidents your team has already closed. Find the entities behind the false positives, then make the narrowest change that stops them from raising alerts.
Accelerynt’s Microsoft security engineers tune detection rules this way, inside the Sentinel workspace you already own. Each change is documented and measured against real incidents afterward, so your team knows what is running and why.
The tuning loop
- Start from closed incidents. The classifications your analysts record show which rules fire on normal activity.
- Find what repeats. A few entities, such as one service account or one partner address, often account for much of the noise.
- Make the narrowest change. Exclude the specific entity, ideally through a watchlist, instead of switching the rule off.
- Measure the result. Compare the rule’s true and false positive counts after the change, and keep a record of who changed it.
On this page
Why do Sentinel detection rules need tuning?
Analytics rule templates are written by Microsoft or by the vendor of a connected solution, so they describe suspicious activity in general terms. Your environment has its own normal, such as admin scripts that run at night or a partner that signs in from abroad, and a rule that does not know about it fires anyway.
Microsoft’s own guidance says no analytics rule is perfect and some false positives are bound to need handling. The cost shows up in the queue, where analysts spend their time interpreting low-value alerts. Our post Seeing Everything. Acting on Nothing. describes how that slows every decision that follows.
For how scheduled rules run and where they come from, read what MDR with custom detection rules means.
What does Microsoft Sentinel give you to tune a rule?
Sentinel has built-in features for each part of the job. Two of them help you see what a rule is doing, and two help you change what it flags.
Tuning insights
A panel in the analytics rule wizard shows the rule’s alerts and incidents over the last 14 days, with closed incidents grouped by true or false positive. The feature is in preview.
Exclusion recommendations
Sentinel lists entities closely tied to incidents your team closed as false positives. You can exclude each one from future runs of the rule with a single selection.
Automation rule exceptions
An automation rule closes matching incidents and records the closing reason, while the analytics rule keeps running as written. By default the exception expires after 24 hours.
Watchlist exceptions
The rule query reads its exceptions from a watchlist, so an analyst can add one without editing the rule. The same watchlist can serve several rules.
Tuning insights also lists the entities that appear most often across a rule’s alerts, and that view needs entity mapping on the rule. Map entities when you build each rule so the insight has something to show.
Sources are Microsoft Learn, fine-tuning recommendations for analytics rules and handle false positives in Microsoft Sentinel.
Should you handle a false positive with an automation rule or a rule change?
Microsoft describes two ways to handle a false positive. An automation rule closes matching incidents without touching the analytics rule, and a change to the rule query stops the alert from being created in the first place.
| Question | Automation rule exception | Change the analytics rule |
|---|---|---|
| What it changes | Matching incidents close automatically, and the rule keeps running as written | The rule query, so matching activity no longer raises an alert |
| Best for | Quick or temporary exceptions an analyst can set up | Detailed exceptions meant to last |
| How long it lasts | Expires after 24 hours by default, and you can change the expiry | Until someone changes the query again |
| Exception logic | Conditions on the incident, and one automation rule can cover several analytics rules | Full query logic, including subnets and watchlists |
| Record left behind | Each closed incident keeps the closing reason and a comment | Whatever your change process records, such as a commit in a source repository |
| Who usually does it | An analyst working the queue | An engineer who writes KQL |
24 hours
The default expiry Microsoft Sentinel sets on an automation rule created as a false positive exception. Microsoft notes that the expiry reduces the chance of false negatives, as explained in handle false positives in Microsoft Sentinel.
In practice the two work together. An analyst sets a short exception while an engineer reviews the rule, and the lasting fix goes into the query.
Microsoft also notes that creating the exception from an incident works in the Azure portal only. Teams running Sentinel in the Defender portal build those automation rules from scratch.
How do you tune a Sentinel detection rule step by step?
These steps assume your analysts classify incidents when they close them, because each later step depends on that record.
- Pick the rule by impact. Rank rules by how much noise each one creates and how much the activity it watches matters to the business, then tune the top of that list first.
- Read the closed incidents. Open tuning insights for the rule and review the incidents your team closed as false positives over the last 14 days.
- Find the entities behind the noise. Check the most frequent entities and the exclusion recommendations, and confirm with your team that each one is normal activity.
- Make the narrowest change. Exclude the specific entity in the rule query, ideally through a watchlist, and use a short automation rule exception while the change is reviewed.
- Record it and measure it. Note who changed the rule and why, then compare its true and false positive counts at the next review.
A rule tuned too far can stop catching the activity it was built for. For the checks to run before any suppression, read how managed security providers reduce alert fatigue.
Why manage detection rules as code?
Microsoft Sentinel can deploy custom content from a GitHub or Azure DevOps repository. You update a rule in the repository, and the change is deployed to your workspaces automatically.
A repository connection can carry analytics rules, automation rules, hunting queries, parsers, playbooks and workbooks, with custom detection rules in preview.
- Every change has a history. The repository keeps each version of a rule along with who changed it, which is the record a tuning review needs.
- Every workspace gets the same rule. When you run more than one workspace, one tested change reaches all of them through the same deployment.
Microsoft’s guidance on false positives now names custom detections in the Defender portal as the best way to create new rules across Microsoft Sentinel and Defender XDR. Newer rules may live there, and they can be managed from a repository too.
Sources are Microsoft Learn, deploy content as code from your repository and custom detections overview.
How Accelerynt engineers detection rules
Accelerynt’s Microsoft Security Engineering team writes custom KQL detection logic and tunes signal fidelity to reduce false positives, working inside your existing Microsoft environment. Detection rules are managed as code, with CI/CD pipelines for Sentinel content management.
Rules evolve with your environment. We measure their effectiveness and refine them based on real incidents, and every rule is documented and explained, so your team always knows what is running and why.
If you also want a team watching those rules around the clock, Accelerynt managed detection and response tunes your Sentinel analytics rules with your team’s input on normal activity and builds suppression for false positives at the source. We guarantee a noise escalation rate under 1% at Stage 1.
Where to go next
Tuning decides what reaches your analysts, and automation decides what happens next. For containment that runs under pre-approved protocols, read how to automate incident containment in Microsoft Sentinel. Our post on automating identity threat response walks through a playbook that blocks IP addresses and revokes sessions.
Rules only see the data that reaches Sentinel. For how other tools feed it, read whether managed security operations can integrate with your existing SIEM and endpoint tools, and see our post on making cloud risk data actionable in Sentinel.
Our post The Headcount Trap explains why adding analysts rarely fixes a noisy queue.
Frequently asked questions
How do you tune Microsoft Sentinel detection rules?
Start from the incidents your team closed as false positives, find the entities that keep repeating, and exclude them in the rule query, ideally through a watchlist. Then compare the rule’s true and false positive counts at the next review. Accelerynt’s Microsoft security engineers tune rules this way inside your own Sentinel workspace.
Who can tune our Microsoft Sentinel detection rules for us?
Accelerynt’s Microsoft Security Engineering team writes and tunes Sentinel detection rules inside your environment, documents every rule and refines each one based on real incidents. For round-the-clock coverage, Accelerynt MDR also tunes your analytics rules with your team’s input on normal activity.
How do you reduce false positives in Microsoft Sentinel without turning off rules?
Exclude the specific entity that causes the noise instead of disabling the rule. A watchlist referenced in the query lets analysts add exceptions without editing the rule, and a short automation rule exception can close matching incidents while the lasting fix is reviewed. Accelerynt tunes Sentinel analytics rules with your team’s input on normal activity and builds suppression for false positives at the source.
What are tuning insights in Microsoft Sentinel?
Tuning insights is a preview feature in the analytics rule wizard. It shows a rule’s alert and incident statistics over 14 days and recommends entities to exclude, based on incidents your team closed as false positives.
Should I use an automation rule or edit the analytics rule to handle a false positive?
Use an automation rule for a quick or temporary exception, since it closes matching incidents without changing the rule and expires after 24 hours by default. Edit the rule query for a lasting exception, which Microsoft calls the most flexible and complete option.
Can Microsoft Sentinel detection rules be managed as code?
Yes. Sentinel can deploy analytics rules and other custom content from a GitHub or Azure DevOps repository, so a change made in the repository is deployed to your workspaces automatically.
