Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
Tuning is the deliberate adjustment of a detection so that it keeps catching what it was built for while producing fewer worthless alerts. Done carelessly, it is how blind spots are created.

What Is Detection Rule Tuning?
A new detection rule rarely arrives at the right sensitivity. It either misses variants of the technique or buries the analyst under administrative activity that looks superficially identical. Tuning is the work of closing that gap, and it is one of the few security activities where the wrong approach makes the organisation measurably less safe while appearing to improve things.
Before changing anything, pull 30 days of the rule’s alerts and classify them. Three buckets matter: genuine defects in the logic, correct detections of authorised activity, and correct detections of activity that should not be happening but is tolerated. The third is the interesting one, because the fix is a conversation with an engineering team rather than a change to the rule.
Then calculate precision: true positives divided by total alerts. Without that number the tuning conversation is anecdotal, and rules are usually tuned or deleted on the basis of whoever complained most recently.
Reach for the narrowest fix that solves the problem. Ordered from safest to most dangerous:
Exclusion lists are where coverage quietly disappears. Four disciplines keep them honest. Be specific: exclude a full path plus a signature or hash, never a bare filename, because C:\Windows\Temp\backup.exe is a file an attacker can create. Record the reason and the approver on every entry. Give each exclusion an expiry date and review it then, since the project that justified it in March may have finished in April. And never exclude by attacker-controllable value, which rules out hostname patterns, command line fragments, and user agent strings; a rule that ignores anything mentioning monitoring_agent is a rule an attacker can switch off by naming their process accordingly.
The exclusions themselves deserve a periodic audit. In most mature estates a quarter of them are obsolete, and a handful cover more activity than anyone intended.
Some noise is a sign of a bad practice rather than a bad detection. Administrators using domain admin credentials for routine work, service accounts with interactive logon rights, deployment tools that run encoded PowerShell, or shared accounts that make attribution impossible all generate alerts that are technically correct. Fixing the practice removes the noise permanently and improves security at the same time, which tuning the rule does neither.
Raise these as findings rather than absorbing them. A rule producing 200 alerts a month because of one badly designed deployment process is an argument for changing the process, and it is usually a more persuasive one than the security team could otherwise make.
Treat every tuning change as a change: peer reviewed, recorded in version control with a reason, and deployed through the same path as the original rule. Re-validate after tuning by executing the technique again, because the most common outcome of over-tightening is a rule that no longer fires for the thing it was built to catch. Then watch the precision figure for the next month.
Both the before and after belong in the rule lifecycle record, and the whole exercise is the most direct remedy available for alert fatigue. At Privia Security, tuning engagements normally start with 30 days of alert classification, because the fix is obvious once the composition of the queue is visible. Our SOC maturation services include that analysis.
Prefer fixes that preserve detection of the technique: correct the logic, add required context, or group duplicate alerts into one case. Use value exclusions last, make them specific enough that an attacker cannot reproduce them, give every exclusion an owner and an expiry, and re-run a test of the technique after tuning to confirm the rule still fires for the behaviour it exists to catch.
It depends on consequence and volume. A rule firing twice a month can be useful at 20% precision because the cost of checking is trivial. A rule firing 200 times a week needs to be far better or it will be ignored regardless of its accuracy. Set a target per rule based on its expected volume, record it, and review against measured outcomes quarterly.
Move them to audit mode rather than disabling them outright. Results are still recorded for later analysis and for retrospective searching, but analysts are not paged. That preserves the data needed to tune the rule properly, which disabling destroys, and it avoids the situation where a rule is switched off for a fortnight and quietly never switched back on.
Someone other than the person making the change, through the same review process used for new rules. Tuning reduces detection capability by design, so it warrants at least as much scrutiny as adding a rule. Record the reason, the scope of what is now excluded, and the expiry date, and make the change through version control so it can be reverted and audited.
You May Be Interested In These
We use a few cookies that are needed to make this website work, and they are always on. We would also like to set analytics cookies, to see which pages are useful; functional cookies, so that our live chat works; and advertising cookies, to measure our campaigns. We will not set any of those until you say yes, and you can change your mind at any time from the Cookie Settings link in the footer.
Choose the cookies you are happy for us to use. Strictly necessary cookies are always on; everything else is off until you turn it on.
These cookies keep the website working: they hold your security checks, your form submissions and the choice you make here. The site cannot run without them, so they are always on.
These let our live chat open and remember your conversation, and let the newsletter sign-up work. Switch them off and the chat window will not appear.
These tell us which pages are read and how people find us, so that we can improve the site. The reports are about groups of visitors, never about you personally.
These let us measure our advertising and show you our adverts on other websites, such as LinkedIn, Google and Meta. Switch them off and you will still see adverts, but they will be less relevant.