Explore our new Cybersecurity Terms Glossary, explained by our experts.
Cybersecurity Terms Glossary
A false positive is an alert for something harmless; a false negative is silence for something harmful. Every detection trades one against the other, and the base rate makes that trade harsher than intuition suggests.

What Are False Positives and False Negatives?
Detection systems make two kinds of mistake. A false positive raises an alert for activity that is not malicious. A false negative fails to raise one for activity that is. Every adjustment to a rule moves error from one column to the other, and no amount of engineering removes both, which is why the honest question is never “how do we eliminate false positives” but “which error can we afford where”.
| Activity malicious | Activity benign | |
|---|---|---|
| Alert raised | True positive | False positive |
| No alert | False negative | True negative |
From these come the two metrics worth tracking. Precision is true positives divided by all alerts raised: of what we were told, how much mattered. Recall is true positives divided by all malicious activity: of what happened, how much were we told about. Precision is measurable from your own triage records. Recall is not, because you cannot count what you never saw, and anyone quoting a recall figure is estimating from tests rather than measuring.
Much of what a SOC records as a false positive is nothing of the kind. When a rule designed to detect remote execution fires on an administrator using PsExec, the rule worked exactly as specified. That is a benign true positive: correct detection of authorised activity.
The distinction drives different remedies. A false positive means the logic is wrong and needs correcting. A benign true positive means the environment contains legitimate instances of the behaviour, and the answer is a scoped exception, an asset-based restriction, or a change to the practice that generates it. Teams that record both as false positives end up weakening rules that are working, and their reported false positive rate tells them nothing useful.
Consider a detection with genuinely excellent characteristics: 99% recall and a 1% false positive rate, running over one million authentication events a day of which ten are malicious. It catches roughly ten of the ten, and it also raises about 10,000 alerts on the benign 999,990. Precision is around 0.1%. An analyst would check a thousand alerts to find one real finding.
That arithmetic, not laziness, is why single-indicator detections over high-volume data do not work, and why practitioners combine signals, scope rules to where the behaviour is abnormal, and accumulate risk across several weak indicators rather than alerting on each. It is also why canary accounts are so effective: they engineer the base rate so that the benign population is nearly empty.
The decision is contextual. For the last line of defence against ransomware, a false negative is catastrophic and a false positive costs an hour, so bias towards sensitivity. For a detection feeding automated containment on production servers, a false positive causes an outage, so bias towards precision and accept that quieter variants will be missed. Write the choice down per rule, because an undocumented preference becomes someone else’s accident.
False negatives are the harder error to manage precisely because they are silent. The only ways to find them are deliberate: executing techniques through atomic tests or simulation, hunting for behaviour that no rule covers, purple team exercises, and the painful retrospective review after an incident that asks which existing rule should have fired and did not.
A board asking about false positive rates generally wants to know whether the security spend is producing usable output. Give them precision per rule family, the proportion of the queue that is actually reviewed, and the validated coverage figure from testing, rather than a single aggregate percentage that mixes a daily noise generator with a rule that fires twice a year.
And resist the target of zero false positives. A detection estate with no false positives is one that has been tuned until it only catches what it has already seen, and the next incident will not resemble the last one. Reasonable precision with known gaps, reviewed regularly through tuning and the rule lifecycle, is the achievable goal.
There is no universal figure, because it depends on volume and consequence. A rule firing twice a month can be valuable at 20% precision, since checking costs minutes. A rule firing 200 times a week needs much higher precision or it will be ignored no matter how accurate it is. Set a precision target per rule based on expected volume and review it against measured outcomes.
Only by looking for them deliberately. Execute attacker techniques with atomic tests or a simulation platform and check whether alerts appear. Hunt for behaviours no current rule covers. Run purple team exercises where the offensive side shares what it did. And after every incident, ask which rule should have detected each stage and why it did not, then fix the rule rather than noting the gap.
Because malicious events are extremely rare compared with benign ones. Even a rule with a 1% false positive rate produces thousands of alerts when applied to a million daily events containing a handful of real ones, so precision collapses. The practical consequences are to scope rules narrowly, combine multiple signals before alerting, and prefer detections over populations where benign activity is close to zero.
No, and conflating them causes real harm. A false positive means the rule logic matched something it was not meant to match, so the logic needs fixing. A benign true positive means the rule correctly identified the behaviour but the behaviour was authorised, so the answer is a scoped exception or a change to the practice. Track them separately or your tuning decisions will be based on the wrong signal.
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.