
Bilingual analysis and safe lab PoC for CVE-2026-90931, a stored XSS in LaraDashboard's SVG media upload, with source verifier, marker payload, and remediation guidance.
English | Português do Brasil
Public research evidence and safe local-lab reproduction material for stored cross-site scripting in LaraDashboard's media library.
| Field | Value |
|---|---|
| Product | LaraDashboard |
| Official affected range | >= 0.9.0, < 1.3.0 |
| Fixed boundary | 1.3.0 |
| Tested revision | 9dc1335835701651aef3adcb3e885acecb32402c |
| Weakness | CWE-79, CWE-434 |
| Official severity | CVSS 4.0 5.1 / CVSS 3.1 5.4 |
| Required permission | media.create |
| Advisory | GHSA-9gxw-qpx8-x9c7 |
| CNA / reporter | VulnCheck / capivara-research |
The CVE title retains the originally reported 0.9.0 through 1.2.2 wording.
The canonical record updated on 2026-10-29 expresses the affected range as
>= 0.9.0, < 1.3.0 and marks 1.3.0 unaffected. This repository follows the
current structured range and separately preserves the tested revision.
The production upload pipeline accepts SVG files without sanitizing active content. The files are stored under the application's origin and served inline as image/svg+xml. A content-level user holding only media.create can therefore store JavaScript that executes when another user opens the file.
Outside demo mode, MediaUploadRequest applies only generic file and size validation. SVG is explicitly recognized as image/svg+xml, while the subsequent checks validate extension/MIME consistency but never sanitize the SVG DOM or reject scripts and event handlers.
The original coordinated disclosure validated the issue end to end using a custom role with only media.create and media.view.
The poc/ directory contains:
--execute.The marker changes visible text in the SVG document and sets the fragment to XSS_MARKER; it does not read cookies, tokens, or other application data.
For a source-only reproducible check pinned to the audited revision:
docker build -t cve-2026-90931 .
docker run --rm cve-2026-90931
Exit status 1 means vulnerable markers were found; 0 means none were found.
JavaScript executes in the dashboard origin after user interaction. When opened by an authenticated administrator, same-origin script can act with that user's browser privileges. The published advisory conservatively scores the issue as CVSS 3.1 5.4.
Content-Security-Policy and X-Content-Type-Options: nosniff as defense in depth.Reported through GitHub Private Vulnerability Reporting and published by the project on 2026-08-30. VulnCheck assigned CVE-2026-90931 on 2026-09-14 and credits capivara-research as reporter.
.
├── README.md / README.pt-BR.md
├── poc/
│ ├── static_verify.py # non-invasive source verifier
│ ├── marker.svg # harmless visible browser marker
│ └── upload_marker.py # explicit, loopback-only upload helper
├── evidence/official-record-summary.md
├── Dockerfile
├── SECURITY.md
└── LICENSE
Use this material only on systems you own or are explicitly authorized to test. The PoCs are intended for isolated, disposable lab environments.