Common Weakness Enumeration

CWE-436

Allowed-with-Review

Interpretation Conflict

Abstraction: Class · Status: Incomplete

Product A handles inputs or steps differently than Product B, which causes A to perform incorrect actions based on its perception of B's state.

378 vulnerabilities reference this CWE, most recent first.

GHSA-HGV6-X3P4-Q38V

Vulnerability from github – Published: 2023-10-10 18:31 – Updated: 2024-04-04 08:30
VLAI
Details

A interpretation conflict in Fortinet IPS Engine versions 7.321, 7.166 and 6.158 allows attacker to evade IPS features via crafted TCP packets.

Show details on source website

{
  "affected": [],
  "aliases": [
    "CVE-2023-40718"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-436"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2023-10-10T17:15:12Z",
    "severity": "HIGH"
  },
  "details": "A interpretation conflict in Fortinet IPS Engine versions 7.321, 7.166 and 6.158 allows attacker to evade IPS features via crafted TCP packets.",
  "id": "GHSA-hgv6-x3p4-q38v",
  "modified": "2024-04-04T08:30:20Z",
  "published": "2023-10-10T18:31:32Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2023-40718"
    },
    {
      "type": "WEB",
      "url": "https://fortiguard.com/psirt/FG-IR-23-090"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:H/A:N",
      "type": "CVSS_V3"
    }
  ]
}

GHSA-HH82-3PMQ-7FRP

Vulnerability from github – Published: 2022-12-12 21:25 – Updated: 2023-01-10 23:07
VLAI
Summary
Netty vulnerable to HTTP Response splitting from assigning header value iterator
Details

Impact

When calling DefaultHttpHeaders.set with an iterator of values (as opposed to a single given value), header value validation was not performed, allowing malicious header values in the iterator to perform HTTP Response Splitting.

Patches

The necessary validation was added in Netty 4.1.86.Final.

Workarounds

Integrators can work around the issue by changing the DefaultHttpHeaders.set(CharSequence, Iterator<?>) call, into a remove() call, and call add() in a loop over the iterator of values.

References

HTTP Response Splitting CWE-113: Improper Neutralization of CRLF Sequences in HTTP Headers

For more information

If you have any questions or comments about this advisory: * Open an issue in [example link to repo](https://github.com/netty/netty) * Email us at netty-security@googlegroups.com

Show details on source website

{
  "affected": [
    {
      "package": {
        "ecosystem": "Maven",
        "name": "io.netty:netty-codec-http"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "4.1.83.Final"
            },
            {
              "fixed": "4.1.86.Final"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2022-41915"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-113",
      "CWE-436"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2022-12-12T21:25:44Z",
    "nvd_published_at": "2022-12-13T07:15:00Z",
    "severity": "MODERATE"
  },
  "details": "### Impact\nWhen calling `DefaultHttpHeaders.set` with an _iterator_ of values (as opposed to a single given value), header value validation was not performed, allowing malicious header values in the iterator to perform [HTTP Response Splitting](https://owasp.org/www-community/attacks/HTTP_Response_Splitting).\n\n### Patches\nThe necessary validation was added in Netty 4.1.86.Final.\n\n### Workarounds\nIntegrators can work around the issue by changing the `DefaultHttpHeaders.set(CharSequence, Iterator\u003c?\u003e)` call, into a `remove()` call, and call `add()` in a loop over the iterator of values.\n\n### References\n[HTTP Response Splitting](https://owasp.org/www-community/attacks/HTTP_Response_Splitting)\n[CWE-113: Improper Neutralization of CRLF Sequences in HTTP Headers](https://cwe.mitre.org/data/definitions/113.html)\n\n### For more information\nIf you have any questions or comments about this advisory:\n* Open an issue in [[example link to repo](https://github.com/netty/netty)](https://github.com/netty/netty)\n* Email us at [netty-security@googlegroups.com](mailto:netty-security@googlegroups.com)\n",
  "id": "GHSA-hh82-3pmq-7frp",
  "modified": "2023-01-10T23:07:13Z",
  "published": "2022-12-12T21:25:44Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/netty/netty/security/advisories/GHSA-hh82-3pmq-7frp"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2022-41915"
    },
    {
      "type": "WEB",
      "url": "https://github.com/netty/netty/issues/13084"
    },
    {
      "type": "WEB",
      "url": "https://github.com/netty/netty/pull/12760"
    },
    {
      "type": "WEB",
      "url": "https://github.com/netty/netty/commit/c37c637f096e7be3dffd36edee3455c8e90cb1b0"
    },
    {
      "type": "WEB",
      "url": "https://github.com/netty/netty/commit/fe18adff1c2b333acb135ab779a3b9ba3295a1c4"
    },
    {
      "type": "WEB",
      "url": "https://github.com/netty/netty"
    },
    {
      "type": "WEB",
      "url": "https://lists.debian.org/debian-lts-announce/2023/01/msg00008.html"
    },
    {
      "type": "WEB",
      "url": "https://security.netapp.com/advisory/ntap-20230113-0004"
    },
    {
      "type": "WEB",
      "url": "https://www.debian.org/security/2023/dsa-5316"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:L/I:L/A:N",
      "type": "CVSS_V3"
    }
  ],
  "summary": "Netty vulnerable to HTTP Response splitting from assigning header value iterator"
}

GHSA-HRWM-HGMJ-7P9C

Vulnerability from github – Published: 2026-04-16 01:03 – Updated: 2026-04-16 01:03
VLAI
Summary
@fastify/express's middleware path doubling causes authentication bypass in child plugin scopes
Details

Summary

@fastify/express v4.0.4 contains a path handling bug in the onRegister function that causes middleware paths to be doubled when inherited by child plugins. This results in complete bypass of Express middleware security controls for all routes defined within child plugin scopes that share a prefix with parent-scoped middleware. No special configuration is required — this affects the default Fastify configuration.

Details

The vulnerability exists in the onRegister function at index.js lines 92-101. When a child plugin is registered with a prefix, the onRegister hook copies middleware from the parent scope and re-registers it using instance.use(...middleware). However, the middleware paths stored in kMiddlewares are already prefixed from their original registration.

The call flow demonstrates the problem: 1. Parent scope registers middleware: app.use('/admin', authFn) — use() calculates path as '' + '/admin' = '/admin' — stores ['/admin', authFn] in kMiddlewares 2. Child plugin registers with { prefix: '/admin' } — triggers onRegister(instance) 3. onRegister copies parent middleware and calls instance.use('/admin', authFn) on child 4. Child's use() function calculates path as '/admin' + '/admin' = '/admin/admin' — registers middleware with doubled path 5. Routes in child scope use the child's Express instance, where middleware is registered under the incorrect path /admin/admin 6. Requests to /admin/secret don't match /admin/admin — middleware is silently skipped

The root cause is in the use() function at lines 25-26, which always prepends this.prefix to string paths, combined with onRegister re-calling use() with already-prefixed paths.

PoC

const fastify = require('fastify');
const http = require('http');

function get(port, url) {
  return new Promise((resolve, reject) => {
    http.get('http://localhost:' + port + url, (res) => {
      let data = '';
      res.on('data', (chunk) => data += chunk);
      res.on('end', () => resolve({ status: res.statusCode, body: data }));
    }).on('error', reject);
  });
}

async function test() {
  const app = fastify({ logger: false });
  await app.register(require('@fastify/express'));

  // Middleware enforcing auth on /admin routes
  app.use('/admin', function(req, res, next) {
    if (!req.headers.authorization) {
      res.statusCode = 403;
      res.setHeader('content-type', 'application/json');
      res.end(JSON.stringify({ error: 'Forbidden' }));
      return;
    }
    next();
  });

  // Root scope route — middleware works correctly
  app.get('/admin/root-data', async () => ({ data: 'root-secret' }));

  // Child scope route — middleware BYPASSED
  await app.register(async function(child) {
    child.get('/secret', async () => ({ data: 'child-secret' }));
  }, { prefix: '/admin' });

  await app.listen({ port: 19876, host: '0.0.0.0' });

  // Root scope: correctly blocked
  let r = await get(19876, '/admin/root-data');
  console.log('/admin/root-data (no auth):', r.status, r.body);
  // Output: 403 {"error":"Forbidden"}

  // Child scope: BYPASSED — secret data returned without auth
  r = await get(19876, '/admin/secret');
  console.log('/admin/secret (no auth):', r.status, r.body);
  // Output: 200 {"data":"child-secret"}

  await app.close();
}
test();

Actual output:

/admin/root-data (no auth): 403 {"error":"Forbidden"}
/admin/secret (no auth): 200 {"data":"child-secret"}

Impact

Complete bypass of Express middleware security controls for all routes defined in child plugin scopes. Authentication, authorization, rate limiting, CSRF protection, audit logging, and any other middleware-based security mechanisms are silently skipped for affected routes.

  • No special request crafting is required — normal requests bypass the middleware
  • It affects the idiomatic Fastify plugin pattern commonly used in production
  • The bypass is silent with no errors or warnings
  • Developers' basic testing of root-scoped routes will pass, masking the vulnerability
  • Any child plugin scope that shares a prefix with middleware is affected

Applications using @fastify/express with path-scoped middleware and child plugins with matching prefixes are vulnerable in default configurations.

Affected Versions

  • @fastify/express v4.0.4 (latest at time of discovery)
  • Fastify 5.x in default configuration
  • No special router options required (ignoreDuplicateSlashes not needed)
  • Affects any child plugin registration where the prefix overlaps with middleware path scoping
  • Does NOT affect middleware registered without path scoping (global middleware)
  • Does NOT affect middleware registered on root path (/) due to special case handling

Variant Testing

Scenario Middleware Path Child Prefix Result
Root route /admin/root-data /admin N/A Middleware runs (403)
Child route /admin/secret /admin /admin BYPASS (200)
Child route /api/data /api /api BYPASS (200)
Nested child /admin/sub/data /admin /admin/sub BYPASS — path becomes /admin/sub/admin
Middleware on / with any child / /api No bypass — path === '/' && prefix.length > 0 special case

Suggested Fix

The onRegister function should store and re-use the original unprefixed middleware paths, or avoid re-calling the use() function entirely. Options include: 1. Store the original path and function separately in kMiddlewares before prefixing 2. Strip the parent prefix before re-registering in child scopes 3. Store already-constructed Express middleware objects rather than re-processing paths

Show details on source website

{
  "affected": [
    {
      "database_specific": {
        "last_known_affected_version_range": "\u003c= 4.0.4"
      },
      "package": {
        "ecosystem": "npm",
        "name": "@fastify/express"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "4.0.5"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-33807"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-436"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-04-16T01:03:25Z",
    "nvd_published_at": "2026-04-15T10:16:48Z",
    "severity": "CRITICAL"
  },
  "details": "### Summary\n\n`@fastify/express` v4.0.4 contains a path handling bug in the `onRegister` function that causes middleware paths to be doubled when inherited by child plugins. This results in complete bypass of Express middleware security controls for all routes defined within child plugin scopes that share a prefix with parent-scoped middleware. No special configuration is required \u2014 this affects the default Fastify configuration.\n\n### Details\n\nThe vulnerability exists in the `onRegister` function at `index.js` lines 92-101. When a child plugin is registered with a prefix, the `onRegister` hook copies middleware from the parent scope and re-registers it using `instance.use(...middleware)`. However, the middleware paths stored in `kMiddlewares` are already prefixed from their original registration.\n\nThe call flow demonstrates the problem:\n1. Parent scope registers middleware: `app.use(\u0027/admin\u0027, authFn)` \u2014 `use()` calculates path as `\u0027\u0027 + \u0027/admin\u0027 = \u0027/admin\u0027` \u2014 stores `[\u0027/admin\u0027, authFn]` in `kMiddlewares`\n2. Child plugin registers with `{ prefix: \u0027/admin\u0027 }` \u2014 triggers `onRegister(instance)`\n3. `onRegister` copies parent middleware and calls `instance.use(\u0027/admin\u0027, authFn)` on child\n4. Child\u0027s `use()` function calculates path as `\u0027/admin\u0027 + \u0027/admin\u0027 = \u0027/admin/admin\u0027` \u2014 registers middleware with doubled path\n5. Routes in child scope use the child\u0027s Express instance, where middleware is registered under the incorrect path `/admin/admin`\n6. Requests to `/admin/secret` don\u0027t match `/admin/admin` \u2014 middleware is silently skipped\n\nThe root cause is in the `use()` function at lines 25-26, which always prepends `this.prefix` to string paths, combined with `onRegister` re-calling `use()` with already-prefixed paths.\n\n### PoC\n\n```javascript\nconst fastify = require(\u0027fastify\u0027);\nconst http = require(\u0027http\u0027);\n\nfunction get(port, url) {\n  return new Promise((resolve, reject) =\u003e {\n    http.get(\u0027http://localhost:\u0027 + port + url, (res) =\u003e {\n      let data = \u0027\u0027;\n      res.on(\u0027data\u0027, (chunk) =\u003e data += chunk);\n      res.on(\u0027end\u0027, () =\u003e resolve({ status: res.statusCode, body: data }));\n    }).on(\u0027error\u0027, reject);\n  });\n}\n\nasync function test() {\n  const app = fastify({ logger: false });\n  await app.register(require(\u0027@fastify/express\u0027));\n  \n  // Middleware enforcing auth on /admin routes\n  app.use(\u0027/admin\u0027, function(req, res, next) {\n    if (!req.headers.authorization) {\n      res.statusCode = 403;\n      res.setHeader(\u0027content-type\u0027, \u0027application/json\u0027);\n      res.end(JSON.stringify({ error: \u0027Forbidden\u0027 }));\n      return;\n    }\n    next();\n  });\n  \n  // Root scope route \u2014 middleware works correctly\n  app.get(\u0027/admin/root-data\u0027, async () =\u003e ({ data: \u0027root-secret\u0027 }));\n  \n  // Child scope route \u2014 middleware BYPASSED\n  await app.register(async function(child) {\n    child.get(\u0027/secret\u0027, async () =\u003e ({ data: \u0027child-secret\u0027 }));\n  }, { prefix: \u0027/admin\u0027 });\n  \n  await app.listen({ port: 19876, host: \u00270.0.0.0\u0027 });\n  \n  // Root scope: correctly blocked\n  let r = await get(19876, \u0027/admin/root-data\u0027);\n  console.log(\u0027/admin/root-data (no auth):\u0027, r.status, r.body);\n  // Output: 403 {\"error\":\"Forbidden\"}\n  \n  // Child scope: BYPASSED \u2014 secret data returned without auth\n  r = await get(19876, \u0027/admin/secret\u0027);\n  console.log(\u0027/admin/secret (no auth):\u0027, r.status, r.body);\n  // Output: 200 {\"data\":\"child-secret\"}\n  \n  await app.close();\n}\ntest();\n```\n\nActual output:\n```\n/admin/root-data (no auth): 403 {\"error\":\"Forbidden\"}\n/admin/secret (no auth): 200 {\"data\":\"child-secret\"}\n```\n\n### Impact\n\nComplete bypass of Express middleware security controls for all routes defined in child plugin scopes. Authentication, authorization, rate limiting, CSRF protection, audit logging, and any other middleware-based security mechanisms are silently skipped for affected routes.\n\n- No special request crafting is required \u2014 normal requests bypass the middleware\n- It affects the idiomatic Fastify plugin pattern commonly used in production\n- The bypass is silent with no errors or warnings\n- Developers\u0027 basic testing of root-scoped routes will pass, masking the vulnerability\n- Any child plugin scope that shares a prefix with middleware is affected\n\nApplications using `@fastify/express` with path-scoped middleware and child plugins with matching prefixes are vulnerable in default configurations.\n\n### Affected Versions\n\n- `@fastify/express` v4.0.4 (latest at time of discovery)\n- Fastify 5.x in default configuration\n- No special router options required (`ignoreDuplicateSlashes` not needed)\n- Affects any child plugin registration where the prefix overlaps with middleware path scoping\n- Does NOT affect middleware registered without path scoping (global middleware)\n- Does NOT affect middleware registered on root path (`/`) due to special case handling\n\n### Variant Testing\n\n| Scenario | Middleware Path | Child Prefix | Result |\n|---|---|---|---|\n| Root route `/admin/root-data` | `/admin` | N/A | Middleware runs (403) |\n| Child route `/admin/secret` | `/admin` | `/admin` | **BYPASS** (200) |\n| Child route `/api/data` | `/api` | `/api` | **BYPASS** (200) |\n| Nested child `/admin/sub/data` | `/admin` | `/admin/sub` | **BYPASS** \u2014 path becomes `/admin/sub/admin` |\n| Middleware on `/` with any child | `/` | `/api` | No bypass \u2014 `path === \u0027/\u0027 \u0026\u0026 prefix.length \u003e 0` special case |\n\n### Suggested Fix\n\nThe `onRegister` function should store and re-use the original unprefixed middleware paths, or avoid re-calling the `use()` function entirely. Options include:\n1. Store the original path and function separately in `kMiddlewares` before prefixing\n2. Strip the parent prefix before re-registering in child scopes\n3. Store already-constructed Express middleware objects rather than re-processing paths",
  "id": "GHSA-hrwm-hgmj-7p9c",
  "modified": "2026-04-16T01:03:25Z",
  "published": "2026-04-16T01:03:25Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/fastify/fastify-express/security/advisories/GHSA-hrwm-hgmj-7p9c"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-33807"
    },
    {
      "type": "WEB",
      "url": "https://cna.openjsf.org/security-advisories.html"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/fastify/fastify-express"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:N",
      "type": "CVSS_V3"
    }
  ],
  "summary": "@fastify/express\u0027s middleware path doubling causes authentication bypass in child plugin scopes"
}

GHSA-HWPQ-RRPF-PGCQ

Vulnerability from github – Published: 2026-03-02 23:33 – Updated: 2026-03-30 13:44
VLAI
Summary
OpenClaw: system.run approval identity mismatch could execute a different binary than displayed
Details

Summary

system.run approvals in OpenClaw used rendered command text as the approval identity while trimming argv token whitespace. Runtime execution still used raw argv. A crafted trailing-space executable token could therefore execute a different binary than what the approver saw.

Affected Packages / Versions

  • Package: openclaw (npm)
  • Affected versions: <= 2026.2.24
  • Patched versions: >= 2026.2.25

Impact

This is an approval-integrity bypass that can lead to unexpected command execution under the OpenClaw runtime user when an attacker can influence command argv and reuse/obtain a matching approval context.

Trust Model Note

OpenClaw does not treat adversarial multi-user sharing of one gateway host/config as a supported security boundary. This finding is still valid in supported deployments because it breaks the operator approval boundary itself (approved display command vs executed argv).

Fix Commit(s)

  • 03e689fc89bbecbcd02876a95957ef1ad9caa176

Release Process Note

patched_versions is pre-set to the release (2026.2.25). Advisory published with npm release 2026.2.25.

OpenClaw thanks @tdjackey for reporting.

Show details on source website

{
  "affected": [
    {
      "database_specific": {
        "last_known_affected_version_range": "\u003c= 2026.2.24"
      },
      "package": {
        "ecosystem": "npm",
        "name": "openclaw"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "2026.2.25"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-32065"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-436",
      "CWE-863"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-03-02T23:33:08Z",
    "nvd_published_at": null,
    "severity": "MODERATE"
  },
  "details": "### Summary\n`system.run` approvals in OpenClaw used rendered command text as the approval identity while trimming argv token whitespace. Runtime execution still used raw argv. A crafted trailing-space executable token could therefore execute a different binary than what the approver saw.\n\n### Affected Packages / Versions\n- Package: `openclaw` (npm)\n- Affected versions: `\u003c= 2026.2.24`\n- Patched versions: `\u003e= 2026.2.25`\n\n### Impact\nThis is an approval-integrity bypass that can lead to unexpected command execution under the OpenClaw runtime user when an attacker can influence `command` argv and reuse/obtain a matching approval context.\n\n### Trust Model Note\nOpenClaw does not treat adversarial multi-user sharing of one gateway host/config as a supported security boundary. This finding is still valid in supported deployments because it breaks the operator approval boundary itself (approved display command vs executed argv).\n\n### Fix Commit(s)\n- `03e689fc89bbecbcd02876a95957ef1ad9caa176`\n\n### Release Process Note\n`patched_versions` is pre-set to the release (`2026.2.25`). Advisory published with npm release `2026.2.25`.\n\nOpenClaw thanks @tdjackey for reporting.",
  "id": "GHSA-hwpq-rrpf-pgcq",
  "modified": "2026-03-30T13:44:16Z",
  "published": "2026-03-02T23:33:08Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/openclaw/openclaw/security/advisories/GHSA-hwpq-rrpf-pgcq"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-32065"
    },
    {
      "type": "WEB",
      "url": "https://github.com/openclaw/openclaw/commit/03e689fc89bbecbcd02876a95957ef1ad9caa176"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/openclaw/openclaw"
    },
    {
      "type": "WEB",
      "url": "https://www.vulncheck.com/advisories/openclaw-approval-identity-mismatch-in-system-run-command-execution"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:H/PR:L/UI:R/S:U/C:N/I:H/A:N",
      "type": "CVSS_V3"
    },
    {
      "score": "CVSS:4.0/AV:N/AC:H/AT:P/PR:L/UI:A/VC:N/VI:H/VA:N/SC:N/SI:N/SA:N",
      "type": "CVSS_V4"
    }
  ],
  "summary": "OpenClaw: system.run approval identity mismatch could execute a different binary than displayed"
}

GHSA-HX8X-4MJR-GRC3

Vulnerability from github – Published: 2026-10-28 18:32 – Updated: 2026-11-05 21:34
VLAI
Details

MIME::Field::ParamVal versions before 5.521 for Perl do not flag a duplicate boundary parameter whose name contains a nested comment.

A parameter written boundary((c))="A" or boundary())="A" is stored under that literal name rather than as boundary. A second boundary parameter is then not reported as a duplicate, and MIME::Parser splits the message on it.

Readers that skip comments as RFC 5322 allows take the first boundary instead. A remote sender can hide an attachment from a filter that relies on ambiguous_content() while the recipient's mail client shows it.

Show details on source website

{
  "affected": [],
  "aliases": [
    "CVE-2026-116272"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-436"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2026-10-28T18:17:13Z",
    "severity": "MODERATE"
  },
  "details": "MIME::Field::ParamVal versions before 5.521 for Perl do not flag a duplicate boundary parameter whose name contains a nested comment.\n\nA parameter written boundary((c))=\"A\" or boundary(\\))=\"A\" is stored under that literal name rather than as boundary. A second boundary parameter is then not reported as a duplicate, and MIME::Parser splits the message on it.\n\nReaders that skip comments as RFC 5322 allows take the first boundary instead. A remote sender can hide an attachment from a filter that relies on ambiguous_content() while the recipient\u0027s mail client shows it.",
  "id": "GHSA-hx8x-4mjr-grc3",
  "modified": "2026-11-05T21:34:42Z",
  "published": "2026-10-28T18:32:13Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-116272"
    },
    {
      "type": "WEB",
      "url": "https://metacpan.org/release/DSKOLL/MIME-tools-5.521/changes"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:L/A:N",
      "type": "CVSS_V3"
    }
  ]
}

GHSA-J4H9-PM27-4RFW

Vulnerability from github – Published: 2026-06-23 18:03 – Updated: 2026-08-21 18:28
VLAI
Summary
OctoPrint has possible file exfiltration via query parameters on upload endpoints
Details

Impact

OctoPrint versions up until and including 1.11.7 as well as 2.0.0rc1 and 2.0.0rc2 contain a vulnerability that allows an attacker with the FILE_UPLOAD permission to exfiltrate files from the host that OctoPrint has read access to, by moving them into the upload folder where they then can be downloaded from. This vulnerability was already reported as GHSA-m9jh-jf9h-x3h2/CVE-2025-48067 but the fix provided in OctoPrint 1.11.2 turned out to be incomplete.

The primary risk lies in the potential exfiltration of secrets stored inside OctoPrint's config, or further system files. By removing important runtime files, this could also be used to impact the availability of the host after an attempted server restart. Given that the attacker requires a user account with file upload permissions, the actual impact of this should however hopefully be minimal in most cases.

Patches

The vulnerability has been patched in version 1.11.8 and 2.0.0rc3.

Details

OctoPrint's web application is implemented in Flask, but uploads are first intercepted by a custom upload handler built on Tornado that sits in front of it. The handler streams the upload to a temporary file on disk - so files larger than the available memory can be uploaded - and rewrites the request, adding internal form fields that tell Flask where to find that temporary file.

These fields are reserved and meant to be set only by the upload handler, never by the client. The previous fix from GHSA-m9jh-jf9h-x3h2/CVE-2025-48067 stripped them from the request received from the client when they were sent as multipart form fields, yet they could still reach Flask through other channels: as plain query parameters, or - since the Tornado handler and Flask did not parse requests identically - smuggled in via several "parser differentials" that looked harmless to the handler while Flask still saw the injected fields. Any of these let an attacker make OctoPrint treat an arbitrary file on the host as a freshly uploaded one and move it into the upload folder.

The following endpoints in OctoPrint are affected:

  • /api/files/{local|sdcard}
  • /api/languages
  • /plugin/backup/restore
  • /plugin/pluginmanager/upload_file

Further upload endpoints in third party plugins might be affected too.

The fix rejects requests carrying any of the reserved fields, aligns the Tornado handler's request parsing with Flask's (Werkzeug) to avoid any differential parsing, and re-validates the request rewritten by Tornado before forwarding it to Flask.

Credits

This vulnerability was discovered and responsibly disclosed to OctoPrint by Koh Jun Sheng and Jacopo Tediosi.

Timeline

2026-06-04: Report received 2026-06-04: Report acknowledged 2026-06-08: Report verified 2026-06-17: Fix ready for 1.11.x 2026-06-22: Fix ported to 2.0.0 2026-06-23: Fix released with 1.11.8 and 2.0.0rc3

Show details on source website

{
  "affected": [
    {
      "database_specific": {
        "last_known_affected_version_range": "\u003c= 1.11.7"
      },
      "package": {
        "ecosystem": "PyPI",
        "name": "OctoPrint"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "1.11.8"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    },
    {
      "database_specific": {
        "last_known_affected_version_range": "\u003c= 2.0.0rc2"
      },
      "package": {
        "ecosystem": "PyPI",
        "name": "OctoPrint"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "2.0.0rc1"
            },
            {
              "fixed": "2.0.0rc3"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-54134"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-436",
      "CWE-73"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-06-23T18:03:54Z",
    "nvd_published_at": null,
    "severity": "HIGH"
  },
  "details": "### Impact\n\nOctoPrint versions up until and including 1.11.7 as well as 2.0.0rc1 and 2.0.0rc2 contain a vulnerability that allows an attacker with the `FILE_UPLOAD` permission to exfiltrate files from the host that OctoPrint has read access to, by moving them into the upload folder where they then can be downloaded from. This vulnerability was already reported as [GHSA-m9jh-jf9h-x3h2/CVE-2025-48067](https://github.com/OctoPrint/OctoPrint/security/advisories/GHSA-m9jh-jf9h-x3h2) but the fix provided in OctoPrint 1.11.2 turned out to be incomplete.\n\nThe primary risk lies in the potential exfiltration of secrets stored inside OctoPrint\u0027s config, or further system files. By removing important runtime files, this could also be used to impact the availability of the host after an attempted server restart. Given that the attacker requires a user account with file upload permissions, the actual impact of this should however hopefully be minimal in most cases.\n\n### Patches\n\nThe vulnerability has been patched in version 1.11.8 and 2.0.0rc3.\n\n### Details\n\nOctoPrint\u0027s web application is implemented in Flask, but uploads are first intercepted by a custom upload handler built on Tornado that sits in front of it. The handler streams the upload to a temporary file on disk - so files larger than the available memory can be uploaded - and rewrites the request, adding internal form fields that tell Flask where to find that temporary file.\n\nThese fields are reserved and meant to be set only by the upload handler, never by the client. The previous fix from [GHSA-m9jh-jf9h-x3h2/CVE-2025-48067](https://github.com/OctoPrint/OctoPrint/security/advisories/GHSA-m9jh-jf9h-x3h2) stripped them from the request received from the client when they were sent as multipart form fields, yet they could still reach Flask through other channels: as plain query parameters, or - since the Tornado handler and Flask did not parse requests identically - smuggled in via several \"parser differentials\" that looked harmless to the handler while Flask still saw the injected fields. Any of these let an attacker make OctoPrint treat an arbitrary file on the host as a freshly uploaded one and move it into the upload folder.\n\nThe following endpoints in OctoPrint are affected:\n\n- `/api/files/{local|sdcard}`\n- `/api/languages`\n- `/plugin/backup/restore`\n- `/plugin/pluginmanager/upload_file`\n\nFurther upload endpoints in third party plugins might be affected too.\n\nThe fix rejects requests carrying any of the reserved fields, aligns the Tornado handler\u0027s request parsing with Flask\u0027s (Werkzeug) to avoid any differential parsing, and re-validates the request rewritten by Tornado before forwarding it to Flask.\n\n### Credits\n\nThis vulnerability was discovered and responsibly disclosed to OctoPrint by Koh Jun Sheng and Jacopo Tediosi.\n\n### Timeline\n\n2026-06-04: Report received\n2026-06-04: Report acknowledged\n2026-06-08: Report verified\n2026-06-17: Fix ready for 1.11.x\n2026-06-22: Fix ported to 2.0.0\n2026-06-23: Fix released with 1.11.8 and 2.0.0rc3",
  "id": "GHSA-j4h9-pm27-4rfw",
  "modified": "2026-08-21T18:28:35Z",
  "published": "2026-06-23T18:03:54Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/OctoPrint/OctoPrint/security/advisories/GHSA-j4h9-pm27-4rfw"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-54134"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/OctoPrint/OctoPrint"
    },
    {
      "type": "WEB",
      "url": "https://github.com/pypa/advisory-database/tree/main/vulns/octoprint/PYSEC-2026-2687.yaml"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:4.0/AV:A/AC:L/AT:N/PR:L/UI:N/VC:H/VI:L/VA:N/SC:N/SI:N/SA:N",
      "type": "CVSS_V4"
    }
  ],
  "summary": "OctoPrint has possible file exfiltration via query parameters on upload endpoints"
}

GHSA-J4JC-CQ9H-XRHR

Vulnerability from github – Published: 2026-08-12 18:31 – Updated: 2026-10-02 23:14
VLAI
Summary
Apache Airflow Backfill API parser discrepancy permits cross-DAG authorization bypass
Details

Apache Airflow's Backfill API authorized a request against a Dag id supplied by the caller whenever the backfill_id path segment failed to parse. The authorization dependency parsed it with int() while the route handler parsed it as pydantic's NonNegativeInt, which accepts values int() rejects (1.0 coerces to 1); FastAPI resolves dependencies before endpoint validation, so the two acted on different Dags. An authenticated user holding edit permission on any single Dag could therefore read, pause and cancel backfills belonging to any other Dag, including moving another Dag's queued runs to failed. No non-default configuration is required and backfill ids are sequential, so finding a target is trivial. Users are advised to upgrade to apache-airflow 3.3.1 or later, which parses the backfill id with the same type the routes declare.

Show details on source website

{
  "affected": [
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "apache-airflow"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "3.3.1"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-68968"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-436"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-10-02T23:14:02Z",
    "nvd_published_at": "2026-08-12T16:17:19Z",
    "severity": "HIGH"
  },
  "details": "Apache Airflow\u0027s Backfill API authorized a request against a Dag id supplied by the caller whenever the `backfill_id` path segment failed to parse. The authorization dependency parsed it with `int()` while the route handler parsed it as pydantic\u0027s `NonNegativeInt`, which accepts values `int()` rejects (`1.0` coerces to `1`); FastAPI resolves dependencies before endpoint validation, so the two acted on different Dags. An authenticated user holding edit permission on any single Dag could therefore read, pause and cancel backfills belonging to any other Dag, including moving another Dag\u0027s queued runs to `failed`. No non-default configuration is required and backfill ids are sequential, so finding a target is trivial. Users are advised to upgrade to apache-airflow 3.3.1 or later, which parses the backfill id with the same type the routes declare.",
  "id": "GHSA-j4jc-cq9h-xrhr",
  "modified": "2026-10-02T23:14:02Z",
  "published": "2026-08-12T18:31:18Z",
  "references": [
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-68968"
    },
    {
      "type": "WEB",
      "url": "https://github.com/apache/airflow/pull/70889"
    },
    {
      "type": "WEB",
      "url": "https://github.com/apache/airflow/commit/6fcae4e2b0aabd22b006659bc47139a124813865"
    },
    {
      "type": "WEB",
      "url": "https://github.com/apache/airflow/commit/bbb97f55e647dd09f0725c111b7a7081b9df7850"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/apache/airflow"
    },
    {
      "type": "WEB",
      "url": "https://github.com/apache/airflow/releases/tag/3.3.1"
    },
    {
      "type": "WEB",
      "url": "https://github.com/pypa/advisory-database/tree/main/vulns/apache-airflow/PYSEC-2026-3710.yaml"
    },
    {
      "type": "WEB",
      "url": "https://lists.apache.org/thread/f9zmw6xs5b4syhwzbl6fsxm4kf2632ol"
    },
    {
      "type": "WEB",
      "url": "http://www.openwall.com/lists/oss-security/2026/08/12/12"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:N",
      "type": "CVSS_V3"
    }
  ],
  "summary": "Apache Airflow Backfill API parser discrepancy permits cross-DAG authorization bypass"
}

GHSA-J55Q-P8QC-F6WG

Vulnerability from github – Published: 2026-10-30 18:32 – Updated: 2026-10-30 18:32
VLAI
Details

DBHub is a database MCP server for Postgres, MySQL, SQL Server, Oracle, MariaDB, SQLite. Prior to 0.22.6, read-only SQL validation can be bypassed on the MySQL and MariaDB connectors. In src/utils/sql-parser.ts, scanSingleLineComment treats every double dash as a comment start, whereas MySQL and MariaDB require following whitespace, a control character, or the end of the line. The read-only check in src/tools/execute-sql.ts validates split statements with isReadOnlySQL, but src/connectors/mysql/index.ts and src/connectors/mariadb/index.ts execute the original string with multipleStatements enabled. A double dash followed by a non-whitespace character can therefore hide a subsequent write or destructive DDL statement from the validator while the database interprets the dashes as minus operators and executes both statements. Anyone able to influence the assistant's SQL, including through prompt injection in content the assistant reads, can bypass read-only mode to modify or drop database data, subject to the database account's privileges. PostgreSQL is not affected by this path because its double-dash parsing agrees with the checker and its connector executes individual split statements. This issue is fixed in version 0.22.6.

Show details on source website

{
  "affected": [],
  "aliases": [
    "CVE-2026-61789"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-436"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2026-10-30T17:17:26Z",
    "severity": "MODERATE"
  },
  "details": "DBHub is a database MCP server for Postgres, MySQL, SQL Server, Oracle, MariaDB, SQLite. Prior to 0.22.6, read-only SQL validation can be bypassed on the MySQL and MariaDB connectors. In `src/utils/sql-parser.ts`, `scanSingleLineComment` treats every double dash as a comment start, whereas MySQL and MariaDB require following whitespace, a control character, or the end of the line. The read-only check in `src/tools/execute-sql.ts` validates split statements with `isReadOnlySQL`, but `src/connectors/mysql/index.ts` and `src/connectors/mariadb/index.ts` execute the original string with `multipleStatements` enabled. A double dash followed by a non-whitespace character can therefore hide a subsequent write or destructive DDL statement from the validator while the database interprets the dashes as minus operators and executes both statements. Anyone able to influence the assistant\u0027s SQL, including through prompt injection in content the assistant reads, can bypass read-only mode to modify or drop database data, subject to the database account\u0027s privileges. PostgreSQL is not affected by this path because its double-dash parsing agrees with the checker and its connector executes individual split statements. This issue is fixed in version 0.22.6.",
  "id": "GHSA-j55q-p8qc-f6wg",
  "modified": "2026-10-30T18:32:42Z",
  "published": "2026-10-30T18:32:42Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/bytebase/dbhub/security/advisories/GHSA-j656-3hf2-fvjc"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-61789"
    },
    {
      "type": "WEB",
      "url": "https://github.com/bytebase/dbhub/pull/342"
    },
    {
      "type": "WEB",
      "url": "https://github.com/bytebase/dbhub/commit/872bb338f7d31f6afe517a076ac3e3edafaaaf08"
    },
    {
      "type": "WEB",
      "url": "https://github.com/bytebase/dbhub/releases/tag/v0.22.6"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:P/VC:N/VI:H/VA:H/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X",
      "type": "CVSS_V4"
    }
  ]
}

GHSA-J5RX-43MX-FV6C

Vulnerability from github – Published: 2026-11-10 18:31 – Updated: 2026-11-10 18:31
VLAI
Details

Nodemailer 3.0.0 before 10.4.1 contains an address parsing flaw in addressparser that truncates domains at folding whitespace or comments before a dot, allowing attackers to bypass recipient domain restrictions. Attackers can submit crafted to, cc, bcc or replyTo addresses that pass RFC 5322 domain checks but deliver mail to an attacker-chosen address under the application's sender identity.

Show details on source website

{
  "affected": [],
  "aliases": [
    "CVE-2026-125755"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-436"
    ],
    "github_reviewed": false,
    "github_reviewed_at": null,
    "nvd_published_at": "2026-11-10T16:17:22Z",
    "severity": "HIGH"
  },
  "details": "Nodemailer 3.0.0 before 10.4.1 contains an address parsing flaw in addressparser that truncates domains at folding whitespace or comments before a dot, allowing attackers to bypass recipient domain restrictions. Attackers can submit crafted to, cc, bcc or replyTo addresses that pass RFC 5322 domain checks but deliver mail to an attacker-chosen address under the application\u0027s sender identity.",
  "id": "GHSA-j5rx-43mx-fv6c",
  "modified": "2026-11-10T18:31:41Z",
  "published": "2026-11-10T18:31:41Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/nodemailer/nodemailer/security/advisories/GHSA-3q53-mvfw-q9v8"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-125755"
    },
    {
      "type": "WEB",
      "url": "https://github.com/nodemailer/nodemailer/commit/c2b9c41bb81b2294ebadd294ad15ff035a17d0a5"
    },
    {
      "type": "WEB",
      "url": "https://github.com/nodemailer/nodemailer/blob/3b5cf1eec32c81f287870c8a1c9c09640062cfd5/src/addressparser/index.ts#L330-L345"
    },
    {
      "type": "WEB",
      "url": "https://github.com/nodemailer/nodemailer/releases/tag/v10.4.1"
    },
    {
      "type": "WEB",
      "url": "https://www.vulncheck.com/advisories/nodemailer-3.0.0-before-10.4.1-recipient-allow-list-bypass-via-addressparser"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:L/A:N",
      "type": "CVSS_V3"
    },
    {
      "score": "CVSS:4.0/AV:N/AC:H/AT:P/PR:N/UI:N/VC:H/VI:L/VA:N/SC:N/SI:N/SA:N/E:X/CR:X/IR:X/AR:X/MAV:X/MAC:X/MAT:X/MPR:X/MUI:X/MVC:X/MVI:X/MVA:X/MSC:X/MSI:X/MSA:X/S:X/AU:X/R:X/V:X/RE:X/U:X",
      "type": "CVSS_V4"
    }
  ]
}

GHSA-J656-3HF2-FVJC

Vulnerability from github – Published: 2026-11-02 17:57 – Updated: 2026-11-02 17:57
VLAI
Summary
DBHub: MySQL and MariaDB comment interpretation bypasses read-only SQL validation
Details

The read-only mode can be bypassed on the MySQL and MariaDB connectors. In read-only mode a caller can still execute arbitrary write and DDL statements (for example DROP TABLE), which defeats the protection that read-only mode is meant to provide when a database is exposed to an AI agent.

Root cause

The problem is a mismatch between how DBHub's SQL parser handles single-line comments and how MySQL and MariaDB actually parse them.

In src/utils/sql-parser.ts, scanSingleLineComment treats "--" as the start of a comment no matter what character follows it. MySQL and MariaDB only begin a "--" comment when the two dashes are followed by whitespace, a control character, or the end of the line. Otherwise the dashes are read as two minus signs and the rest of the line is ordinary SQL.

The read-only check in src/tools/execute-sql.ts splits the input and validates each statement with isReadOnlySQL, but the connector then executes the original, unmodified string. The MySQL and MariaDB connectors open their connection pool with multipleStatements: true (src/connectors/mysql/index.ts:54 and src/connectors/mariadb/index.ts:53) and run the whole string in a single call.

Because of this, the following payload:

SELECT 1--1;DROP TABLE victim

is seen by DBHub as a single SELECT statement, since everything after "--" is treated as a comment and the read-only check passes. MySQL and MariaDB instead parse it as "SELECT 1 - -1" followed by "DROP TABLE victim" and execute both statements.

Proof of concept

1) DBHub's own parser. Running splitSQLStatements and isReadOnlySQL from this repository on the payload returns a single statement and readOnly = true, so the query is allowed in read-only mode. The same code correctly rejects "SELECT 1; DROP TABLE victim", which confirms that the "--1" is what gets through.

2) Execution. Connecting to MariaDB 11.4 with mysql2 and multipleStatements: true (the same option the connector uses) and sending the payload returns 2 for the first statement, which confirms "--" was not treated as a comment, and then drops the table. The table exists before the call and is gone afterwards.

Impact

Read-only mode is the control operators rely on to let an assistant read from a database without being able to change it. With this bypass, anyone who can influence the SQL the assistant sends, most realistically through prompt injection in content the assistant reads, can perform writes and destructive DDL against MySQL and MariaDB. PostgreSQL is not affected by this path, because it treats "--" as a comment regardless and executes each split statement individually.

Affected code (current main, commit e99c3aa9, package version 0.22.3)

  • src/utils/sql-parser.ts — scanSingleLineComment does not require whitespace after "--"
  • src/tools/execute-sql.ts — validates the split statements but executes the original string
  • src/connectors/mysql/index.ts and src/connectors/mariadb/index.ts — multipleStatements: true

Suggested fix

The most reliable fix is to stop executing the original string: run only the statements that were validated, and disable multipleStatements on the MySQL and MariaDB connections unless it is genuinely required. As defense in depth, make the comment scanner dialect-aware so that for MySQL and MariaDB "--" only starts a comment when it is followed by whitespace, a control character, or the end of the line, similar to the handling already in place for MySQL conditional comments.

Show details on source website

{
  "affected": [
    {
      "package": {
        "ecosystem": "npm",
        "name": "@bytebase/dbhub"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "0.22.6"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [],
  "database_specific": {
    "cwe_ids": [
      "CWE-436",
      "CWE-863"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-11-02T17:57:51Z",
    "nvd_published_at": null,
    "severity": "MODERATE"
  },
  "details": "The read-only mode can be bypassed on the MySQL and MariaDB connectors. In read-only mode a caller can still execute arbitrary write and DDL statements (for example DROP TABLE), which defeats the protection that read-only mode is meant to provide when a database is exposed to an AI agent.\n\nRoot cause\n\nThe problem is a mismatch between how DBHub\u0027s SQL parser handles single-line comments and how MySQL and MariaDB actually parse them.\n\nIn src/utils/sql-parser.ts, scanSingleLineComment treats \"--\" as the start of a comment no matter what character follows it. MySQL and MariaDB only begin a \"--\" comment when the two dashes are followed by whitespace, a control character, or the end of the line. Otherwise the dashes are read as two minus signs and the rest of the line is ordinary SQL.\n\nThe read-only check in src/tools/execute-sql.ts splits the input and validates each statement with isReadOnlySQL, but the connector then executes the original, unmodified string. The MySQL and MariaDB connectors open their connection pool with multipleStatements: true (src/connectors/mysql/index.ts:54 and src/connectors/mariadb/index.ts:53) and run the whole string in a single call.\n\nBecause of this, the following payload:\n\n    SELECT 1--1;DROP TABLE victim\n\nis seen by DBHub as a single SELECT statement, since everything after \"--\" is treated as a comment and the read-only check passes. MySQL and MariaDB instead parse it as \"SELECT 1 - -1\" followed by \"DROP TABLE victim\" and execute both statements.\n\nProof of concept\n\n1) DBHub\u0027s own parser. Running splitSQLStatements and isReadOnlySQL from this repository on the payload returns a single statement and readOnly = true, so the query is allowed in read-only mode. The same code correctly rejects \"SELECT 1; DROP TABLE victim\", which confirms that the \"--1\" is what gets through.\n\n2) Execution. Connecting to MariaDB 11.4 with mysql2 and multipleStatements: true (the same option the connector uses) and sending the payload returns 2 for the first statement, which confirms \"--\" was not treated as a comment, and then drops the table. The table exists before the call and is gone afterwards.\n\nImpact\n\nRead-only mode is the control operators rely on to let an assistant read from a database without being able to change it. With this bypass, anyone who can influence the SQL the assistant sends, most realistically through prompt injection in content the assistant reads, can perform writes and destructive DDL against MySQL and MariaDB. PostgreSQL is not affected by this path, because it treats \"--\" as a comment regardless and executes each split statement individually.\n\nAffected code (current main, commit e99c3aa9, package version 0.22.3)\n\n- src/utils/sql-parser.ts \u2014 scanSingleLineComment does not require whitespace after \"--\"\n- src/tools/execute-sql.ts \u2014 validates the split statements but executes the original string\n- src/connectors/mysql/index.ts and src/connectors/mariadb/index.ts \u2014 multipleStatements: true\n\nSuggested fix\n\nThe most reliable fix is to stop executing the original string: run only the statements that were validated, and disable multipleStatements on the MySQL and MariaDB connections unless it is genuinely required. As defense in depth, make the comment scanner dialect-aware so that for MySQL and MariaDB \"--\" only starts a comment when it is followed by whitespace, a control character, or the end of the line, similar to the handling already in place for MySQL conditional comments.",
  "id": "GHSA-j656-3hf2-fvjc",
  "modified": "2026-11-02T17:57:51Z",
  "published": "2026-11-02T17:57:51Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/bytebase/dbhub/security/advisories/GHSA-j656-3hf2-fvjc"
    },
    {
      "type": "WEB",
      "url": "https://github.com/bytebase/dbhub/pull/342"
    },
    {
      "type": "WEB",
      "url": "https://github.com/bytebase/dbhub/commit/872bb338f7d31f6afe517a076ac3e3edafaaaf08"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/bytebase/dbhub"
    },
    {
      "type": "WEB",
      "url": "https://github.com/bytebase/dbhub/releases/tag/v0.22.6"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:P/VC:N/VI:H/VA:H/SC:N/SI:N/SA:N",
      "type": "CVSS_V4"
    }
  ],
  "summary": "DBHub: MySQL and MariaDB comment interpretation bypasses read-only SQL validation"
}

No mitigation information available for this CWE.

CAPEC-105: HTTP Request Splitting

An adversary abuses the flexibility and discrepancies in the parsing and interpretation of HTTP Request messages by different intermediary HTTP agents (e.g., load balancer, reverse proxy, web caching proxies, application firewalls, etc.) to split a single HTTP request into multiple unauthorized and malicious HTTP requests to a back-end HTTP agent (e.g., web server).

See CanPrecede relationships for possible consequences.

CAPEC-273: HTTP Response Smuggling

An adversary manipulates and injects malicious content in the form of secret unauthorized HTTP responses, into a single HTTP response from a vulnerable or compromised back-end HTTP agent (e.g., server).

See CanPrecede relationships for possible consequences.

CAPEC-34: HTTP Response Splitting

An adversary manipulates and injects malicious content, in the form of secret unauthorized HTTP responses, into a single HTTP response from a vulnerable or compromised back-end HTTP agent (e.g., web server) or into an already spoofed HTTP response from an adversary controlled domain/site.

See CanPrecede relationships for possible consequences.