//nefariousplan

Scope Correction By Reissue

When a CVE's affected-versions field is wrong, the protocol is a new CVE, not an edit. The two records together describe the full scope. Neither one describes it alone, and the gap between them is a strict inequality that automated scanners honour.

CVE records are immutable by design. Once a CNA assigns an identifier and publishes the record, the affected-products and affected-versions fields are statements of record. The provenance guarantee is useful. The side effect is that a record whose scope was wrong on the day of publication remains wrong afterward.

Scope-correction-by-reissue is the shape where the protocol for fixing that error is a new CVE, not an edit. A second identifier is assigned, a second advisory is written, a second affected-versions field is populated. The second record takes the range the first record missed. Because the two records cannot overlap, overlap would mean one host matches two CVE identifiers for one bug, the second record stops short of the first record's version by strict inequality. NVD renders that as a CPE match with versionEndExcluding set to the first record's version. The CPE grammar is explicit that the bound is open. Automated scanners respect the openness and exclude the first record's version from the second record's hits.

The two records together describe the full population at risk. Neither one describes it alone. An administrator who reads only the second CVE sees a version range they are not in. An administrator who reads only the first CVE sees nineteen years of unmentioned releases. Downstream processors that score and triage at CVE-ID granularity, CISA KEV, scanner feeds, vulnerability dashboards, attach deadlines and exploitation notes to one of the two records and not the bug.

Mechanism

The pattern names the shape where correcting a wrong affected-versions field on an existing CVE is, in the catalog's own grammar, impossible. Once a CNA publishes a CVE record, the affected-products and affected-versions fields are frozen. The immutability is deliberate. Downstream consumers (NVD's CPE match rules, CISA's KEV catalog, vulnerability scanners, SBOM tools) read each record as authoritative and merge it into their own data at CVE-ID granularity; a mutable scope field would make every prior scanner result an unprovenanced claim. The side effect is that a scope that was wrong on the day of publication remains wrong afterward, and the only legitimate remedy inside the protocol is another number.

The correction protocol issues a second CVE. A second identifier is assigned, a second advisory is written, a second affected-versions field is populated. The second record takes the range the first record missed. Because the two records cannot overlap, overlapping would mean one host matches two CVE identifiers for one bug, which breaks CVE-ID-granular counting in every downstream pipeline, the second record's range stops short of the first record's version by strict inequality. NVD translates the advisory phrasing ("4.0.0 before 6.1.5") into a CPE match rule with versionEndExcluding set to the first record's version. The CPE grammar is explicit that the bound is open; the version named in the field is not in the range the record covers. Automated scanners honour the openness.

Three downstream effects compose. First, administrators reading the second record alone see a false negative for the version the first record covered. The second record's text names the excluded version only to exclude it; nothing else inside the record points at the first. Second, downstream processors that operate at CVE-ID granularity attach exploitation notes, patch deadlines, and severity scores to one CVE at a time. When only one of the two records carries the "actively exploited" language, only that record lands in KEV, and the version only the other record covers is invisible to patch-deadline enforcement. Third, research write-ups, PoC READMEs, and mitigation guides written against the second record propagate its range without re-reading the first; the gap between the records reinforces itself in downstream text.

The Apache Roller exhibit grounds the shape. CVE-2026-82384 named Roller 6.1.5 as affected, following the CNA shorthand of "the version the researcher reported against is the version that got patched." CVE-2026-129448, issued seven weeks later, named "Apache Roller 4.0.0 before 6.1.5" and included the explicit note "This is the same vulnerability as CVE-2026-82384. This issue is being actively exploited." NVD resolved the November advisory into a CPE match with versionEndExcluding: 6.1.5. CISA added the November record to the Known Exploited Vulnerabilities catalog on 20 November with a three-day patch deadline. CISA did not add the September record. The web.xml servlet declaration that enables the vulnerable code path was written in 2007 by the same author who shipped Roller 4.0 and is byte-identical across every release tag from roller-4.0 through roller-6.1.5. The code says "every version since 2007." The catalog says "one version in record A, every other version in record B, and the exclusion between them is a strict inequality that an automated scanner will respect."

The mechanism distinguishes itself from near classes on the axis of why the record understates. Scope Is Support Status names the shape where the affected-versions line encodes a vendor's end-of-support matrix rather than the code's reach; the understatement is a business decision expressed in catalog grammar. CVE Scope Is The Patcher names the shape where the record names the fork that triaged the report and omits the upstream shipping the byte-identical code; the understatement is a maintainer-coordination artefact. Scope-correction-by-reissue is the shape where the catalog protocol itself, once a wrong number has been recorded, forecloses an edit and requires a new number whose range is non-overlapping by construction. The understatement persists because the correction protocol is working as designed. The protocol's job is to preserve the audit trail. The side effect is a strict inequality in the middle of the population at risk.

Exhibits

CVE-2026-129448: Apache's Second Roller CVE Excludes 6.1.5 By Name. CISA's KEV Catalog Lists Only The Second CVE.. CVE-2026-82384 named Roller 6.1.5 as affected. CVE-2026-129448, issued seven weeks later with the explicit note "same vulnerability," names 4.0.0 before 6.1.5. NVD resolves the second record into a CPE match with versionEndExcluding: 6.1.5, excluding 6.1.5 by strict inequality. CISA added only the second record to KEV. The config line is nineteen years old and identical across every tag.

Boundaries

Not every CVE record whose affected-versions line understates the code is scope-correction-by-reissue. Scope Is Support Status names the shape where the field encodes a vendor's end-of-support matrix rather than the code's reach; the understatement is a business decision. CVE Scope Is The Patcher names the shape where the record covers the fork that triaged the report and omits the upstream shipping byte-identical code; the understatement is a maintainer-coordination artefact. This pattern is specifically the shape where two CVE records exist for one bug, and the strict inequality between them is an artefact of CVE-catalog immutability rather than vendor choice or coordination failure. If there is only one record, the pattern is one of the other two.

Not every pair of related CVE identifiers is a scope correction. Two independent bugs in the same library get two CVE IDs because they are two bugs; two chained primitives that require each other get two CVE IDs because the chain has two steps. Scope-correction-by-reissue requires one bug class and two records describing non-overlapping portions of its reach, usually announced with explicit "same vulnerability as" language in the reissue's advisory and resolved in NVD as two CPE match rules that share criteria and abut at a single version. Where the two records describe different code paths or different primitives, the pattern does not apply.

Not every CVE-record-does-not-match-code gap is a record-vs-record shape. Backport Gap names the shape where one CVE record honestly covers every affected branch and the patch reaches only some of them; the mismatch is between record and patch. Release Lag Disclosure names the shape where the fix is a public commit before the release carrying it; the mismatch is between source and binary. Scope-correction-by-reissue requires two CVE records whose affected-versions fields abut at a strict inequality. Where the record is one and the fix reality differs, the pattern is elsewhere.

Defender playbook

When you encounter an advisory bearing "same vulnerability as CVE-X" or "this issue was previously disclosed as," retrieve both records and union their affected-versions fields before acting. The CPE match for either record alone will understate coverage. If a KEV deadline or scanner feed is attached to only one of the two records, treat the deadline as covering the full union for your hosts; tooling that gates on CVE-ID-granular matching will not do the union on your behalf. For the Apache Roller exhibit, the union is "4.0.0 through 6.1.5 inclusive," which is not what either CVE record says in isolation.

In your vulnerability-management pipeline, parse the advisory text (not just the structured affected-versions field) for cross-references between CVE identifiers that assert "same vulnerability" or equivalent language. Merge the affected ranges of the cross-referenced records into a single bug-level track; retain the per-CVE records for provenance, but gate patch urgency on the union. Scanners that key on versionEndExcluding will honour the strict inequality and leave the excluded host invisible; KEV ingests at CVE-ID granularity and will surface only the record that carried the exploitation note. Neither tool will reconcile the records for you.

When you are the CNA correcting a scope error, assume automated consumers will not merge the two records. Publish the reissue's advisory with explicit prose that the first record is incomplete, and with text an automated pipeline can index ("same vulnerability as CVE-X," "supersedes," "previously disclosed as"). Where the pipeline permits it, attach a human-readable mitigation document alongside the record that unions the ranges and names the excluded version by name. The CVE catalog's immutability protects the audit trail; it does not communicate scope. The second task is yours.

To test whether a scanner or dashboard honours scope-correction-by-reissue in practice, stage a host at the excluded version (6.1.5 for the Roller exhibit, or the equivalent for the two records you are investigating) and verify whether both CVE identifiers fire against it. A scanner that only matches the second record's versionEndExcluding range will not; a scanner that resolves by product plus version against the union of records will. The test is cheap and the result is the ground truth on whether your tooling reads the catalog at bug granularity or at record granularity.

Kinship

Backport Gap. Adjacent in the CVE-record-contradicts-fix-reality family, on a different axis. Backport-gap names the shape where one CVE record honestly covers every affected branch and the patch reaches only some of them; the gap is between record and patch. Scope-correction-by-reissue names the shape where the fix is in one release and the record scope is corrected across two CVE identifiers; the gap is between record and record. Both patterns leave administrators making a patch decision from a CVE document that does not match what their hosts are running. The direction of the mismatch differs: backport-gap over-states remediation, scope-correction-by-reissue under-states exposure.

Disclosure After Exploitation. Co-occurring in the Apache Roller exhibit. The November reissue carried the first in-wild-exploitation note the bug had received, and that note is what drove CISA to add the November record to KEV. The two patterns describe different parts of the same event. Disclosure-after-exploitation describes the vendor's decision to upgrade the advisory after active exploitation was recognised; scope-correction-by-reissue describes the catalog mechanics that made the upgrade a second CVE rather than an edit to the first. Where the two co-occur, the exploitation signal attaches to only one of the two records, and KEV honours the CVE-ID granularity rather than the bug. The record that names the exploited version is not always the record KEV lists.

The first record describes some of the bug. The second record describes the rest. The strict inequality between them is a design feature of the catalog, not a feature of the code.