Why alerts need evidence review: unified threat level decisions in netsa
The traditional approach alerts on every rule hit, so lots of background scanning is marked high risk while successful exploits drown in the noise. netsa adds a layer of evidence review after detection.
Evidence over appearances
- A firewall "allow" does not mean the attack reached a service: DNAT mappings and session evidence filter out background scanning that cannot reach anything.
- A 200 response does not mean the exploit worked: single-page app fallback pages, generic error pages and page assets all return 200, and netsa uses response baselines to rule out these "false successes".
- A successful sign-in after several failures needs a look at account clues to tell normal sign-ins from successful brute force.
Levels are decided in one place
The level a rule declares is only a suggestion. Every stage's conclusion goes to one decision logic, with a floor per category, escalation on success and caps by outcome, so stages do not add and subtract on their own and similar incidents get the same level. Blocked attacks are never marked "critical", and successful exploits are not missed.
Learning has limits and admits mistakes
Calibration that lowers levels only learns from people's decisions, one level at a time, and successful exploits stay critical; false-positive suppression that hides alerts must be accepted by a person. When an objective conclusion overturns something learned, it is reverted automatically and incidents already produced are corrected.
For everything netsa can do, visit the product page.