GHSA-W937-FG2H-XHQ2

Vulnerability from github – Published: 2026-04-22 20:32 – Updated: 2026-04-22 20:32
VLAI?
Summary
locize Client SDK: Cross-origin DOM XSS & Handler Hijack Through Missing e.origin Validation in InContext Editor
Details

Summary

Versions of the locize client SDK (the browser module that wires up the locize InContext translation editor) prior to 4.0.21 register a window.addEventListener("message", …) handler that dispatches to registered internal handlers (editKey, commitKey, commitKeys, isLocizeEnabled, requestInitialize, …) without validating event.origin.

The pre-patch listener in src/api/postMessage.js gates dispatch on event.data.sender === "i18next-editor-frame" — that value sits inside the attacker-controlled message payload, not the browser-enforced origin. Any web page that could embed or be embedded by a locize-enabled host — an iframe on a third-party page, a window.open-ed victim, a parent frame reaching down — could send a crafted postMessage and trigger the internal handlers.

Impact

Depending on which handler the attacker invokes, distinct consequences follow. All of them share the same root cause: the handlers implicitly assumed the payload came from the real editor iframe.

  • Cross-origin DOM XSS via editKey / commitKeys: the pre-patch handleEditKey assigned attacker-controlled payload values to item.node.innerHTML and to item.node.setAttribute(attr, value). That allowed planting <script>, <img onerror>, or onclick/onload/onfocus event handlers; and on attribute writes, href="javascript:…" / src="data:text/html,<script>…" / style="…" / etc.

  • api.source / api.origin hijack via isLocizeEnabled: the handler set api.source = e.source; api.origin = e.origin — attacker-controlled values. All subsequent sendMessage calls (which post translations, callbacks, etc., back toward api.source) would go to the attacker window rather than the real editor, leaking translation content and any metadata the SDK forwards.

  • CSS-injection / layout-escape via requestPopupChanges: containerStyle.height / .width were interpolated into calc() expressions and popup.style.setProperty() without validation, allowing attackers to inject additional CSS declarations (semicolons, behavior:url() on legacy IE, CSS-exfil patterns) into the popup inline style.

Exploitation requires the attacker-owned page to share a window reference with the locize-enabled host: typical vectors are an iframe on an attacker-controlled page, a window.opener/window.open relationship, or a parent frame that can postMessage into an embedded locize host. The SDK intended model is that only the editor iframe at https://incontext.locize.app (or the configured staging/development origin) can reach these handlers.

Affected versions

All versions of locize prior to 4.0.21.

Patch

Fixed in 4.0.21. Two layers:

  1. Primary — validate event.origin at the top of window.addEventListener("message", …) in src/api/postMessage.js. The expected origin is the configured iframe origin (getIframeUrl()), so custom environments continue to work. Messages from any other origin are silently dropped before any handler runs.

  2. Defence-in-depthhandleEditKey now rejects dangerous attribute-name writes (on*, style) and javascript: / data: / vbscript: / file: URLs on href / src / action / formaction / xlink:href; innerHTML assignments are sanitised through a throwaway DOMParser document (stripping <script>, <iframe>, <object>, <embed>, <link>, <meta>, <base>, <style> plus event handlers and dangerous URL schemes). Legitimate translation formatting (<b>, <em>, <strong>, <a href="https://…">, etc.) passes through.

  3. CSS-injectionhandleRequestPopupChanges now requires containerStyle.height / .width to match a strict CSS length pattern; malformed values are dropped silently.

Workarounds

No workaround short of upgrading.

Credits

Discovered via an internal security audit of the locize ecosystem.

Show details on source website

{
  "affected": [
    {
      "package": {
        "ecosystem": "npm",
        "name": "locize"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "4.0.21"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [],
  "database_specific": {
    "cwe_ids": [
      "CWE-79",
      "CWE-346"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-04-22T20:32:11Z",
    "nvd_published_at": null,
    "severity": "HIGH"
  },
  "details": "### Summary\n\nVersions of the `locize` client SDK (the browser module that wires up the locize InContext translation editor) prior to 4.0.21 register a `window.addEventListener(\"message\", \u2026)` handler that dispatches to registered internal handlers (`editKey`, `commitKey`, `commitKeys`, `isLocizeEnabled`, `requestInitialize`, \u2026) **without validating `event.origin`**.\n\nThe pre-patch listener in `src/api/postMessage.js` gates dispatch on `event.data.sender === \"i18next-editor-frame\"` \u2014 that value sits inside the attacker-controlled message payload, not the browser-enforced origin. Any web page that could embed or be embedded by a locize-enabled host \u2014 an iframe on a third-party page, a `window.open`-ed victim, a parent frame reaching down \u2014 could send a crafted `postMessage` and trigger the internal handlers.\n\n### Impact\n\nDepending on which handler the attacker invokes, distinct consequences follow. All of them share the same root cause: the handlers implicitly assumed the payload came from the real editor iframe.\n\n- **Cross-origin DOM XSS** via `editKey` / `commitKeys`: the pre-patch `handleEditKey` assigned attacker-controlled payload values to `item.node.innerHTML` and to `item.node.setAttribute(attr, value)`. That allowed planting `\u003cscript\u003e`, `\u003cimg onerror\u003e`, or `onclick`/`onload`/`onfocus` event handlers; and on attribute writes, `href=\"javascript:\u2026\"` / `src=\"data:text/html,\u003cscript\u003e\u2026\"` / `style=\"\u2026\"` / etc.\n\n- **`api.source` / `api.origin` hijack** via `isLocizeEnabled`: the handler set `api.source = e.source; api.origin = e.origin` \u2014 attacker-controlled values. All subsequent `sendMessage` calls (which post translations, callbacks, etc., back toward `api.source`) would go to the attacker window rather than the real editor, leaking translation content and any metadata the SDK forwards.\n\n- **CSS-injection / layout-escape** via `requestPopupChanges`: `containerStyle.height` / `.width` were interpolated into `calc()` expressions and `popup.style.setProperty()` without validation, allowing attackers to inject additional CSS declarations (semicolons, `behavior:url()` on legacy IE, CSS-exfil patterns) into the popup inline style.\n\nExploitation requires the attacker-owned page to share a window reference with the locize-enabled host: typical vectors are an `iframe` on an attacker-controlled page, a `window.opener`/`window.open` relationship, or a parent frame that can `postMessage` into an embedded locize host. The SDK intended model is that only the editor iframe at `https://incontext.locize.app` (or the configured staging/development origin) can reach these handlers.\n\n### Affected versions\n\nAll versions of `locize` prior to **4.0.21**.\n\n### Patch\n\nFixed in **4.0.21**. Two layers:\n\n1. **Primary** \u2014 validate `event.origin` at the top of `window.addEventListener(\"message\", \u2026)` in `src/api/postMessage.js`. The expected origin is the configured iframe origin (`getIframeUrl()`), so custom environments continue to work. Messages from any other origin are silently dropped before any handler runs.\n\n2. **Defence-in-depth** \u2014 `handleEditKey` now rejects dangerous attribute-name writes (`on*`, `style`) and `javascript:` / `data:` / `vbscript:` / `file:` URLs on `href` / `src` / `action` / `formaction` / `xlink:href`; `innerHTML` assignments are sanitised through a throwaway DOMParser document (stripping `\u003cscript\u003e`, `\u003ciframe\u003e`, `\u003cobject\u003e`, `\u003cembed\u003e`, `\u003clink\u003e`, `\u003cmeta\u003e`, `\u003cbase\u003e`, `\u003cstyle\u003e` plus event handlers and dangerous URL schemes). Legitimate translation formatting (`\u003cb\u003e`, `\u003cem\u003e`, `\u003cstrong\u003e`, `\u003ca href=\"https://\u2026\"\u003e`, etc.) passes through.\n\n3. **CSS-injection** \u2014 `handleRequestPopupChanges` now requires `containerStyle.height` / `.width` to match a strict CSS length pattern; malformed values are dropped silently.\n\n### Workarounds\n\nNo workaround short of upgrading.\n\n### Credits\n\nDiscovered via an internal security audit of the locize ecosystem.",
  "id": "GHSA-w937-fg2h-xhq2",
  "modified": "2026-04-22T20:32:11Z",
  "published": "2026-04-22T20:32:11Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/locize/locize/security/advisories/GHSA-w937-fg2h-xhq2"
    },
    {
      "type": "WEB",
      "url": "https://github.com/locize/locize/commit/d006b75fadb8e8ab77b023e462850fc6e9170735"
    },
    {
      "type": "WEB",
      "url": "https://developer.mozilla.org/en-US/docs/Web/API/Window/postMessage#security_concerns"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/locize/locize"
    },
    {
      "type": "WEB",
      "url": "https://github.com/locize/locize/releases/tag/v4.0.21"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:C/C:L/I:H/A:L",
      "type": "CVSS_V3"
    }
  ],
  "summary": "locize Client SDK: Cross-origin DOM XSS \u0026 Handler Hijack Through Missing e.origin Validation in InContext Editor "
}


Log in or create an account to share your comment.




Tags
Taxonomy of the tags.


Loading…

Loading…

Loading…

Sightings

Author Source Type Date

Nomenclature

  • Seen: The vulnerability was mentioned, discussed, or observed by the user.
  • Confirmed: The vulnerability has been validated from an analyst's perspective.
  • Published Proof of Concept: A public proof of concept is available for this vulnerability.
  • Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
  • Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
  • Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
  • Not confirmed: The user expressed doubt about the validity of the vulnerability.
  • Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.


Loading…

Detection rules are retrieved from Rulezet.

Loading…

Loading…