Vulnerability false positive
A vulnerability false positive is a detected vulnerability that is not present in an exploitable state. FedRAMP's 2026 definition excludes any flaw that was ever present, so a mitigated or fixed one does not qualify. Close it with evidence from the system itself, not a second scan. Read 19 Nov 2026.
On this page
Three definitions
| Source | Wording | Difference |
|---|---|---|
| NIST SP 800-115 | "An alert that incorrectly indicates that a vulnerability is present." | Alert-based |
| NIST SP 800-83 Rev. 1 | "An instance in which a security tool incorrectly classifies benign content as malicious." | Malware context, not vulnerabilities |
| FedRAMP 2026 | "A detected vulnerability that is not actually present in an exploitable state in the information resource" | Includes code that is present but not in an operating state |
FedRAMP's definition has two notes. The first covers vulnerable software "not loaded, running, or otherwise in an operating state required for exploitation". The second limits it: "This only applies if the vulnerability is not and was not present; a remediated vulnerability or a fully mitigated vulnerability cannot also be a false positive vulnerability." A flaw blocked by a control is mitigated; it is not a false positive, and it stays on the exception register.
What looks like a false positive and is not
| Looks like a false positive when | Actually | Where it goes |
|---|---|---|
| The flaw exists but exploitation needs conditions that never occur | A "not likely exploitable" evaluation | Not reachable, not on the KEV |
| The flaw exists and a control blocks the path | A mitigated flaw | Control already in front |
| You disagree with the severity the scanner assigned | A change of risk assessment, not an absent flaw | PCI SSC FAQ 1597: under Requirement 6.3.1 the entity may rate a vulnerability for its own environment instead of accepting an outside rating |
| The code is on disk but the vulnerable feature is not loaded or running | False positive under FedRAMP's first note, if it cannot be loaded either | This page; if it can be loaded, it is present and needs an exposure check |
| The condition is truly absent on this system | False positive | This page |
FedRAMP's rule VER-EVA-ELX gives the neighboring warning: "The simple reality is that most traditional vulnerabilities discovered by scanners or during assessment are not likely to be exploitable", but providers "who regularly evaluate vulnerabilities as not likely exploitable without careful consideration are more likely to suffer from an adverse impact". Declaring a false positive deserves the same care. A flaw that is present but unlikely to be exploited is evaluated and reported; a flaw that is absent is closed. SP 800-53 RA-5 keeps the order: "Analyze vulnerability scan reports and results from vulnerability monitoring" comes before "Remediate legitimate vulnerabilities".
Why a second tool is weak evidence
NIST SP 800-115 says "Automated tools can produce a significant number of findings, but these findings often need to be validated to isolate false positives. Assessors may validate vulnerabilities by manually examining the vulnerable system or by using a second automated tool and comparing the results." It warns that although a second tool is quick, "these comparison tools can often produce similar results", which include the same false positives. It adds that manual examination "typically provides more accurate results than comparing results from multiple tools, but it also has the potential to be time-consuming."
The same publication says "vulnerability scanners can have a high false positive error rate (i.e., reporting vulnerabilities when none exist)".
Worked example: a version-only scan on a patched Red Hat kernel
CVE-2026-46300 is a Linux kernel flaw, named Fragnesia, in the code that merges network receive buffers. The function skb_try_coalesce() used to drop the marker that flags page fragments as shared. IPsec ESP input code trusts that marker: without it, ESP decrypts in place over fragments that belong to the page cache, so a local low-privilege user can alter the cached copy of a read-only file such as a setuid binary (Red Hat Security Data API). The flaw is in socket buffer coalescing, and it is reachable only where the ESP code is loaded or can be loaded. NVD lists several affected mainline ranges; the one containing 5.14 runs from 5.11 up to but excluding 5.15.208. Red Hat fixed it in RHEL 9 by backport: errata RHSA-2026:19568 of 20 May 2026 ships kernel 5.14.0-687.10.1.el9_8. CISA added the CVE to the KEV catalog on 28 Oct 2026 with a due date of 31 Oct 2026 (KEV feed).
Suppose a scanner that compares version numbers reports the invented host rhel9-app-04 on 29 Oct 2026, because the kernel number 5.14 falls inside the mainline range. The scanner never saw the backport. Two commands settle it: uname -r shows the running kernel, and rpm -q --changelog kernel | grep CVE-2026-46300 shows whether the installed package carries the fix. Distribution kernels carry backports, so the version number alone can mislead.
| What the host shows | Finding is | Next step |
|---|---|---|
uname -r shows 5.14.0-687.10.1.el9_8 or a later build, and the changelog lists CVE-2026-46300 | False positive | Close with the two outputs attached, 29 Oct 2026, two days before the 31 Oct due date |
The package is installed but uname -r shows an older kernel | Present: the host runs the vulnerable code until it reboots | Reboot, or apply one of Red Hat's two mitigations (below) and record it |
| Older kernel, no changelog line | Present | Remediate by the 31 Oct due date |
The second row is the common mistake: an installed fix that is not running is not a false positive. FedRAMP's note on code that is not loaded does not rescue it either, because the ESP code can be loaded and the running kernel is the vulnerable one.
- Block the IPsec modules (Red Hat's broader method, bulletin RHSB-2026-003). Run
lsmod | grep -E 'esp4|esp6'first. Red Hat treats IPsec as not in active use when neither module is loaded, and warns that blocking the modules breaks IPsec where it is configured. Writeinstall esp4 /bin/falseandinstall esp6 /bin/falseas two lines into /etc/modprobe.d/dirtyfrag.conf, then runrmmod esp4 esp6. The rmmod unloads the modules at once (it refuses if they are in use) and the file stops them loading again, including after a reboot. A host that starts IPsec later loses it then, so confirm with the network team that no tunnel is planned. Deleting the file reverts it. - Disable unprivileged user namespaces (Red Hat's targeted method, for hosts where IPsec must stay): write
user.max_user_namespaces=0into /etc/sysctl.d/dirtyfrag.conf and runsysctl --system. Rootless containers, sandboxed browsers and Flatpak can stop working.
With either in place the host is mitigated, not a false positive: FedRAMP's second note says a fully mitigated vulnerability "cannot also be a false positive". It stays on the exception register until the reboot.
Steps that leave a record
- Log the detection. Owner: whoever raised the finding. Record: finding ID, tool, signature, and the time and source of the detection.
- Examine the system, not the report. Owner: the system's administrator. Record: installed and running version, whether the component is loaded or running, the setting involved, with date and command output (SP 800-115, manual examination).
- Have someone other than the person who raised or will close the finding validate it. Owner: a second engineer or the assessor. Record: validator name and date. FedRAMP's legacy playbook marked a false positive Yes and moved it to the Closed tab when the third-party assessor validated it, because "validated FPs are not considered open risks"; one not validated stayed Pending until the agency authorizing official approved it.
- Tune or suppress the signature, with the reason. Owner: the scanner owner. Record: the rule change, so the same alert does not return next scan.
- Close, and write the trigger that reopens it. Owner: the finding owner. Record: closing date, and the reopen triggers: a change of build or configuration, or a new scanner signature. Evidence is true for one build and one configuration; see verifying a mitigation for drift checks.
If verification shows the flaw is present after all, the finding moves to the row of the Decision Table that fits, and the clock does not restart at verification. BOD 26-04 says the timelines begin when CISA adds the vulnerability to the KEV catalog or the agency identifies it on an asset, and "Whichever event occurs first starts the remediation timeline"; see patch available and exposed. A real finding is recorded on the risk acceptance form if it will be accepted. For the legacy deviation request types that preceded the 2026 rules, which also covered false positives, see FedRAMP deviation requests.
Sources
- NIST CSRC glossary, false positive, read 19 Nov 2026
- NIST SP 800-115, Section 7.3 and glossary, read 19 Nov 2026
- FedRAMP 2026 definitions, False Positive Vulnerability, read 19 Nov 2026
- FedRAMP 2026, VER-EVA-EFP, VER-EVA-ELX, VER-RPT-VDT and VER-RPT-AVI, read 19 Nov 2026
- FedRAMP legacy playbook, POA&M (reference only after 24 Jun 2026), read 19 Nov 2026
- PCI SSC FAQ 1597, read 19 Nov 2026
- CISA BOD 26-04, Appendix A, read 19 Nov 2026
- NIST SP 800-53 Rev. 5, RA-5, read 19 Nov 2026
- Red Hat Security Data API, CVE-2026-46300, read 19 Nov 2026
- CISA KEV feed, catalog version 2026.11.18, read 19 Nov 2026
- Red Hat security bulletin RHSB-2026-003, Dirty Frag and Fragnesia, read 19 Nov 2026
Last reviewed