GHSA-QXMC-6F24-G86G
Vulnerability from github – Published: 2026-03-31 22:27 – Updated: 2026-03-31 22:27Summary
In the core update functionality of baserCMS, some parameters sent from the admin panel are passed to the exec() function without proper validation or escaping. This issue allows an authenticated CMS administrator to execute arbitrary OS commands on the server (Remote Code Execution, RCE).
This vulnerability is not a UI-level issue such as screen manipulation or lack of CSRF protection, but rather stems from a design that directly executes input values received on the server side as OS commands. Therefore, even if buttons are hidden in the UI, or even if CakePHP's CSRF/FormProtection (SecurityComponent) ensures that only legitimate POST requests are accepted, an attack is possible as long as a request containing a valid token is processed within an administrator session.
Vulnerability Information
| Item | Details |
|---|---|
| CWE | CWE-78: Improper Neutralization of Special Elements used in an OS Command |
| Impact | Remote Code Execution (RCE) |
| Severity | Critical |
| Attack Requirements | Administrator privileges required |
| Reproducibility | Reproducible (confirmed multiple times) |
| Test Environment | baserCMS 5.2.2 (Docker / development environment) |
Affected Areas
- Controller
PluginsController::get_core_update()- Service
PluginsService::getCoreUpdate()- Affected Endpoint
/baser/admin/baser-core/plugins/get_core_update
Technical Details
Vulnerable Code Flow
PluginsController::get_core_update()
↓ Retrieves php parameter from POST data
PluginsService::getCoreUpdate($targetVersion, $php, $force)
↓ Concatenates $php into command string without validation or escaping
exec($command)
Relevant Code (Excerpt)
PluginsController.php
$service->getCoreUpdate(
$request->getData('targetVersion') ?? '',
$request->getData('php') ?? 'php',
$request->getData('force'),
);
PluginsService.php
$command = $php . ' ' . ROOT . DS . 'bin' . DS . 'cake.php composer ' .
$targetVersion . ' --php ' . $php . ' --dir ' . TMP . 'update';
exec($command, $out, $code);
The $php parameter is user input, and none of the following countermeasures are in place:
- Restriction via allowlist
- Validation via regular expression
- Escaping via
escapeshellarg()or similar
Attack Scenario
- The attacker logs in as a CMS administrator
- Sends a POST request to the core update functionality in the admin panel
- Specifies a string containing OS commands in the
phpparameter exec()is executed on the server side, running the arbitrary OS command
Example Attack Input (Conceptual)
php=php;id>/tmp/rce_test;#
Verification Results (PoC)
Execution Result
$ docker exec bc-php cat /tmp/rce_test
uid=1000(www-data) gid=1000(www-data) groups=1000(www-data)
The above confirms that OS commands can be executed with www-data privileges.
Additional Notes
- Reproducible through the legitimate flow in the admin panel (browser)
- Succeeds even with CSRF/FormProtection tokens included in a legitimate request
- Failure cases (400/403) have also been investigated and differentiated
- Confirmed reproducible via resending HTTP requests with tools such as curl (resending the same request containing valid tokens)
Impact
If this vulnerability is exploited, the following becomes possible:
- Retrieval of server information
- Reading/writing arbitrary files
- Retrieval of application configuration information (DB credentials, etc.)
- OS-level operations beyond application permission boundaries
Although administrator privileges are required, this is a design issue where the impact extends from the application layer to the OS layer, and the impact is considered significant.
Recommended Fix
Primary Recommendation
- Do not accept the PHP executable path from user input
- Fix the PHP executable on the server side using the
PHP_BINARYconstant
$php = escapeshellarg(PHP_BINARY);
Supplementary Fix Recommendations
- Apply
escapeshellarg()escaping to other command-line arguments (version number, directory, etc.) as well - If possible, consider using execution methods that do not involve shell interpretation (array format, Process class, etc.)
Alternative (Not Recommended)
- Allowlist validation for the PHP executable path
- Combined use of regex validation and
escapeshellarg()
However, from the perspective of reducing the attack surface, a design that eliminates user input entirely is recommended.
Additional Notes
- This issue is independent of UI display controls (showing/hiding buttons)
- As long as the endpoint exists, an attack is possible if a request containing valid tokens is processed
- This is a problem stemming from the design-level handling of input, and cannot be prevented by CSRF or UI controls alone
Conclusion
Due to a design issue in baserCMS's core update functionality where user input is passed to exec() without validation, Remote Code Execution (RCE) is achievable with administrator privileges. This vulnerability can be fixed through input validation and design review, and prompt remediation is recommended.
This advisory was translated from Japanese to English using GitHub Copilot.
{
"affected": [
{
"database_specific": {
"last_known_affected_version_range": "\u003c= 5.2.2"
},
"package": {
"ecosystem": "Packagist",
"name": "baserproject/basercms"
},
"ranges": [
{
"events": [
{
"introduced": "0"
},
{
"fixed": "5.2.3"
}
],
"type": "ECOSYSTEM"
}
]
}
],
"aliases": [
"CVE-2026-21861"
],
"database_specific": {
"cwe_ids": [
"CWE-78"
],
"github_reviewed": true,
"github_reviewed_at": "2026-03-31T22:27:05Z",
"nvd_published_at": "2026-03-31T01:16:35Z",
"severity": "CRITICAL"
},
"details": "## Summary\n\nIn the core update functionality of baserCMS, some parameters sent from the admin panel are passed to the `exec()` function without proper validation or escaping. This issue allows **an authenticated CMS administrator to execute arbitrary OS commands on the server (Remote Code Execution, RCE)**.\n\nThis vulnerability is not a UI-level issue such as screen manipulation or lack of CSRF protection, but rather stems from **a design that directly executes input values received on the server side as OS commands**. Therefore, even if buttons are hidden in the UI, or even if CakePHP\u0027s CSRF/FormProtection (SecurityComponent) ensures that only legitimate POST requests are accepted, **an attack is possible as long as a request containing a valid token is processed within an administrator session**.\n\n---\n\n## Vulnerability Information\n\n| Item | Details |\n| ---- | ------- |\n| CWE | CWE-78: Improper Neutralization of Special Elements used in an OS Command |\n| Impact | Remote Code Execution (RCE) |\n| Severity | Critical |\n| Attack Requirements | Administrator privileges required |\n| Reproducibility | Reproducible (confirmed multiple times) |\n| Test Environment | baserCMS 5.2.2 (Docker / development environment) |\n\n---\n\n## Affected Areas\n\n- **Controller**\n - `PluginsController::get_core_update()`\n- **Service**\n - `PluginsService::getCoreUpdate()`\n- **Affected Endpoint**\n - `/baser/admin/baser-core/plugins/get_core_update`\n\n---\n\n## Technical Details\n\n### Vulnerable Code Flow\n\n```text\nPluginsController::get_core_update()\n \u2193 Retrieves php parameter from POST data\nPluginsService::getCoreUpdate($targetVersion, $php, $force)\n \u2193 Concatenates $php into command string without validation or escaping\nexec($command)\n```\n\n### Relevant Code (Excerpt)\n\n**PluginsController.php**\n\n```php\n$service-\u003egetCoreUpdate(\n $request-\u003egetData(\u0027targetVersion\u0027) ?? \u0027\u0027,\n $request-\u003egetData(\u0027php\u0027) ?? \u0027php\u0027,\n $request-\u003egetData(\u0027force\u0027),\n);\n```\n\n**PluginsService.php**\n\n```php\n$command = $php . \u0027 \u0027 . ROOT . DS . \u0027bin\u0027 . DS . \u0027cake.php composer \u0027 .\n $targetVersion . \u0027 --php \u0027 . $php . \u0027 --dir \u0027 . TMP . \u0027update\u0027;\n\nexec($command, $out, $code);\n```\n\nThe `$php` parameter is user input, and **none** of the following countermeasures are in place:\n\n- Restriction via allowlist\n- Validation via regular expression\n- Escaping via `escapeshellarg()` or similar\n\n---\n\n## Attack Scenario\n\n1. The attacker logs in as a CMS administrator\n2. Sends a POST request to the core update functionality in the admin panel\n3. Specifies a string containing OS commands in the `php` parameter\n4. `exec()` is executed on the server side, running the arbitrary OS command\n\n### Example Attack Input (Conceptual)\n\n```text\nphp=php;id\u003e/tmp/rce_test;#\n```\n\n---\n\n## Verification Results (PoC)\n\n### Execution Result\n\n```bash\n$ docker exec bc-php cat /tmp/rce_test\nuid=1000(www-data) gid=1000(www-data) groups=1000(www-data)\n```\n\nThe above confirms that OS commands can be executed with `www-data` privileges.\n\n### Additional Notes\n\n- Reproducible through the legitimate flow in the admin panel (browser)\n- Succeeds even with CSRF/FormProtection tokens included in a legitimate request\n- Failure cases (400/403) have also been investigated and differentiated\n- Confirmed reproducible via resending HTTP requests with tools such as curl (resending the same request containing valid tokens)\n\n---\n\n## Impact\n\nIf this vulnerability is exploited, the following becomes possible:\n\n- Retrieval of server information\n- Reading/writing arbitrary files\n- Retrieval of application configuration information (DB credentials, etc.)\n- OS-level operations beyond application permission boundaries\n\nAlthough administrator privileges are required, **this is a design issue where the impact extends from the application layer to the OS layer**, and the impact is considered significant.\n\n---\n\n## Recommended Fix\n\n### Primary Recommendation\n\n- Do not accept the PHP executable path from user input\n- Fix the PHP executable on the server side using the `PHP_BINARY` constant\n\n```php\n$php = escapeshellarg(PHP_BINARY);\n```\n\n### Supplementary Fix Recommendations\n\n- Apply `escapeshellarg()` escaping to other command-line arguments (version number, directory, etc.) as well\n- If possible, consider using execution methods that do not involve shell interpretation (array format, Process class, etc.)\n\n### Alternative (Not Recommended)\n\n- Allowlist validation for the PHP executable path\n- Combined use of regex validation and `escapeshellarg()`\n\nHowever, **from the perspective of reducing the attack surface, a design that eliminates user input entirely is recommended**.\n\n---\n\n## Additional Notes\n\n- This issue is independent of UI display controls (showing/hiding buttons)\n- As long as the endpoint exists, an attack is possible if a request containing valid tokens is processed\n- This is a problem stemming from the design-level handling of input, and cannot be prevented by CSRF or UI controls alone\n\n---\n\n## Conclusion\n\nDue to a design issue in baserCMS\u0027s core update functionality where user input is passed to `exec()` without validation, **Remote Code Execution (RCE) is achievable with administrator privileges**. This vulnerability can be fixed through input validation and design review, and prompt remediation is recommended.\n\nThis advisory was translated from Japanese to English using GitHub Copilot.",
"id": "GHSA-qxmc-6f24-g86g",
"modified": "2026-03-31T22:27:05Z",
"published": "2026-03-31T22:27:05Z",
"references": [
{
"type": "WEB",
"url": "https://github.com/baserproject/basercms/security/advisories/GHSA-qxmc-6f24-g86g"
},
{
"type": "ADVISORY",
"url": "https://nvd.nist.gov/vuln/detail/CVE-2026-21861"
},
{
"type": "WEB",
"url": "https://basercms.net/security/JVN_20837860"
},
{
"type": "PACKAGE",
"url": "https://github.com/baserproject/basercms"
},
{
"type": "WEB",
"url": "https://github.com/baserproject/basercms/releases/tag/5.2.3"
}
],
"schema_version": "1.4.0",
"severity": [
{
"score": "CVSS:3.1/AV:N/AC:L/PR:H/UI:N/S:C/C:H/I:H/A:H",
"type": "CVSS_V3"
}
],
"summary": "baserCMS has OS Command Injection Leading to Remote Code Execution (RCE)"
}
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.