Skip to content

NVD configurations with target_sw=- never match Fleet CPEs that carry a concrete target_sw (Google brotli CVE-2026-111497 missed on Python and Homebrew) #58295

Open

Description

@marcusallen97

Fleet versions

Web browser and operating system: N/A (vulnerability processing)


💥 Actual behavior

NVD lists CVE-2026-111497 (Google brotli decoder overflow; affects 1.0.3–1.0.9, 1.1.0 and 1.2.0) with a single configuration whose target_sw is - (NA):

cpe:2.3:a:google:brotli:*:*:*:*:*:-:*:*

Fleet generates CPEs for the Python package Brotli and the Homebrew formula brotli with the software's own target_sw (python and macos, and google:brotli after the #57209 vendor pin):

cpe:2.3:a:google:brotli:1.0.9:*:*:*:*:python:*:*
cpe:2.3:a:google:brotli:1.2.0:*:*:*:*:macos:*:*

Neither matches the CVE, so Fleet reports no vulnerability for Python Brotli 1.0.9 or Homebrew brotli 1.2.0 even though both are inside the affected range. CVE-2020-8927 on the same product is found only because its configuration uses target_sw=*.

The cause is in the WFN matcher: matchAttr in server/vulnerabilities/nvd/tools/wfn/matching.go returns true for * or an exact match and false whenever either side is -, so an NVD configuration with target_sw=- can never match a Fleet CPE that carries a concrete target_sw. On 2026-10-29 NVD re-issued the google:brotli CPEs with target_sw=- (noted in #57201), so this is not brotli-specific: any NVD configuration written with target_sw=- is invisible to every software source for which Fleet sets a concrete target_sw (python_packages, homebrew_packages, npm_packages, extensions, macOS/Windows apps, …).

🛠️ Expected behavior

CVE-2026-111497 is reported for Python Brotli 1.0.9 and Homebrew brotli 1.2.0 (and any other in-range Google brotli row), the same way CVE-2020-8927 is for Brotli 1.0.7. For application CPEs, an NVD target_sw of - should be treated as "applies regardless of platform" when compared with Fleet's generated target_sw, for example by expanding the Fleet CPE with a target_sw=- / * alias in expandCPEAliases (server/vulnerabilities/nvd/cve.go) the way the visual_studio_code / visual_studio aliases already are, or by relaxing the NA case in matchAttr for that attribute only.

🧑‍💻 Steps to reproduce

These steps:

  • Have been confirmed to consistently lead to reproduction in multiple Fleet instances.
  • Describe the workflow that led to the error, but have not yet been reproduced in multiple Fleet instances.
  1. Harness, from the repo root with a synced vulnerability DB:
    go run ./tools/nvd/nvdvuln -sync -db_dir /tmp/vulndbs -software_name Brotli -software_version 1.0.9 -software_source python_packages
    go run ./tools/nvd/nvdvuln -db_dir /tmp/vulndbs -software_name brotli -software_version 1.2.0 -software_source homebrew_packages
    
    Both resolve to google:brotli with target_sw=python / macos and report no CVEs (the tool logs "NVD scan produced no matches with non-empty input", its no-vulnerabilities path). Control: -software_name Brotli -software_version 1.0.7 -software_source python_packages reports CVE-2020-8927 (config target_sw=*) but still not CVE-2026-111497.
  2. Live: on a macOS host run pip install Brotli==1.0.9 and brew install brotli (1.2.0 or 1.2.1 — only 1.2.0 is in range), refetch, run the vulnerabilities cron. software_cpe holds the two google:brotli CPEs above; software_cve has no row for CVE-2026-111497 on either software id, and GET /api/v1/fleet/software/versions?query=brotli shows vulnerabilities: null for them.
  3. Compare with the NVD feed entry for CVE-2026-111497: the only cpe23Uri is cpe:2.3:a:google:brotli:*:*:*:*:*:-:*:* with versionEndIncluding ranges covering 1.0.9 and 1.2.0.

🕯️ More info (optional)

Activity

  1. moved this from 📨 Inbox to 🥚 Ready in 🛡️ #g-supply-chainon Nov 11, 2026
  2. marcusallen97 commented on Nov 13, 2026

    @marcusallen97
    ContributorAuthor

    Related, opposite failure in the same matcher: #58685 — since #55058 gave Android apps target_sw=android, the OS shortcut in match_cpe.go (matchTargetSW) lets a vulnerable unversioned o:google:android entry flag an Android app at any version (false positive), while this issue is matchAttr refusing target_sw=- (false negative). Different code paths and fixes, but both are the matcher's platform handling, so linking them for whoever picks either up.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    #g-supply-chainSupply Chain product groupbugSomething isn't working as documented

    Type

    No type

    Projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions