One decision per finding
Triage is deciding, and the fixing comes later. For each finding choose one of three outcomes and write down why.
Confirm. The finding is real and applies to this system. It gets an owner and a due date and goes into the normal work queue.
False positive. The scanner is wrong: the code path cannot be reached, the input is already validated upstream, or the flagged string is a test fixture and not a credential. Record the reason in a full sentence. "Not an issue" is not a reason. A reviewer in six months should be able to read the note and agree or disagree with it.
Accept the risk. The finding is real and you are choosing not to fix it now, because the fix costs more than the exposure warrants or because another control covers it. Acceptance needs three things: a named person with the authority to accept it, the reasoning, and an expiry date. On that date the finding comes back for another look. An acceptance with no expiry is a decision nobody will revisit, and the conditions that justified it will change without anyone noticing.
If a finding needs more than a few minutes of investigation, confirm it and let the owner investigate. A triage session that turns into debugging does not finish.
Severity is not exploitability
Severity, usually expressed as a CVSS score, describes how bad a vulnerability would be if exploited: how it is reached, what privileges it needs, and the effect on confidentiality, integrity and availability. It says nothing about whether anyone is exploiting it.
Two public data sources fill that gap for vulnerabilities that have a CVE identifier.
EPSS, the Exploit Prediction Scoring System published by FIRST.org, gives each CVE a probability between 0 and 1 that exploitation activity will be observed in the next 30 days. It is recalculated daily. Most CVEs score very low, and that is the useful part: it lets you separate the few that attackers are likely to use from the many they are not.
The CISA Known Exploited Vulnerabilities (KEV) catalogue is a list of CVEs for which there is evidence of exploitation in the wild. It is maintained by the US Cybersecurity and Infrastructure Security Agency. Being on the list is a fact about the past and not a forecast, which makes it the strongest single signal you can get.
Neither source knows your environment. A KEV-listed flaw in a library you ship but never call is less urgent than the list suggests, and a low-EPSS flaw on your public login page may deserve more attention than its score. Findings in your own code, such as an injection flaw found by static or dynamic testing, have no CVE and so no EPSS score at all. For those, severity and exposure are what you have.
A workable ordering: anything on KEV first, then high or critical severity with a high EPSS score, then the remaining high and critical findings on internet-facing systems, then everything else by severity.
Due dates by severity
A service level for remediation is a promise about how long a confirmed finding may stay open. Set it by severity, and keep it short enough to mean something and long enough to be kept. A reasonable starting point is critical within a day or two, high within a week, medium within a month and low within a quarter. The exact numbers matter less than three rules.
1. The clock starts at detection. If it starts at triage, a slow triage hides a slow fix.
2. A breach is reported and explained. A missed date that nobody notices teaches the team that the dates are decorative.
3. KEV-listed findings take the shortest deadline whatever their severity score.
If you miss most of your deadlines in the first month, the deadlines are wrong for your capacity or the queue is too large. Lengthen them or raise the bar for what gets confirmed. Do not leave a policy in place that everyone knows is fiction.
What to automate, and what not to
Automate the parts that need no judgement.
De-duplication. The same flaw reported by two scans, or by two tools, should be one finding with two sightings.
Enrichment. Attaching EPSS, KEV status and the affected asset to each finding saves the lookup that makes triage slow.
Closing what has gone. If a later complete scan of the same target no longer detects a finding, it can be resolved without a meeting.
Routing. Send findings to the team that owns the repository or the service.
Reminders for due dates and for acceptances that are about to expire.
Keep the judgement calls with people. Marking something a false positive or accepting a risk are decisions with consequences, and a person should make them and put their name to them. Automated suggestions are useful as a first draft, provided someone reviews them and the record shows who decided. Suppression rules deserve particular care. A broad rule written to silence one noisy check will hide the next real finding in the same category.
A weekly routine
Thirty minutes, the same day each week, two people. One of them should know the system and one should own the security queue.
1. Look at new findings in priority order and decide each one. Stop when the time is up. Whatever is left waits a week unless it is on KEV.
2. Review breached due dates and agree what happens to each.
3. Review acceptances that expire in the coming month.
4. Glance at the totals: findings awaiting triage, open by severity, and median time to fix. If awaiting triage grows three weeks running, the session is too short or the scanner needs tuning.
The measure of a triage process is whether every finding has a decision and a date, and not whether the open count is zero.
How Spoorder handles it
Spoorder's triage queue offers Confirm, False positive, Accept and Escalate for each finding. A false positive needs a written reason of at least 20 characters, and an accepted finding goes to the risk register. Findings with a known CVE are enriched with EPSS and CISA KEV status and grouped as Fix now, Fix soon or Watch. Default due dates are 24 hours for critical, 3 days for high, 7 days for medium and 30 days for low.
Ready to scan your application?
Free health check: HTTPS, TLS and security headers for one URL, in about a minute. Free account, no credit card.
Scan your app free →