Sep 14, 2026·5 min read·14 visits
A flaw in the October CMS image resizer allows processing of non-HTTP protocols, potentially leading to arbitrary code execution via PHAR deserialization.
This report provides a technical analysis of GHSA-2XMM-M4WV-3FJH, an incomplete scheme validation vulnerability in the image resizing utility of October CMS. By exploiting this flaw, authenticated or privileged users can pass dangerous URI schemes to trigger deserialization of untrusted metadata.
October CMS incorporates an image resizing component that processes local and remote image resources. The core of this system handles URLs and local file paths to retrieve, resize, and cache images.
The vulnerability, designated as GHSA-2XMM-M4WV-3FJH, resides within the scheme validation routine of the image resizer. The component classifies any input string containing the substring :// as an external URL. This over-inclusive classification allows arbitrary URI schemes, including dangerous PHP stream wrappers, to bypass designed validation checks.
This security weakness affects October CMS installations from version 4.3.0 up to, but not including, version 4.3.5. Exploitation of the flaw can lead to untrusted metadata deserialization, which may result in remote code execution under specific conditions.
The root cause of this vulnerability lies in the implementation of the ResizeImageItem::fromObject() method within October CMS. This method is responsible for instantiating image representation objects based on user-supplied sources.
To determine if a source string represents a remote resource, the logic checks for the presence of the scheme indicator ://. The implementation assumes that any string containing :// is an external HTTP or HTTPS URL, subsequently routing the path to the self::fromUrl() initialization path.
Because the verification is limited to the existence of :// anywhere in the string, attackers can supply non-HTTP schemes like phar://, file://, or ftp://. When subsequent image operations are executed on the parsed path, PHP’s stream wrapper subsystem intercepts the request. For example, processing a phar:// stream triggers the parsing of the PHP Archive (PHAR) manifest, which automatically deserializes any serialized metadata embedded within the archive.
The following code illustrates the vulnerable routing mechanism. The method incorrectly uses a general substring search to identify external URLs instead of verifying explicit, permitted protocols.
// Vulnerable routing mechanism in ResizeImageItem::fromObject()
public static function fromObject($source) {
if (is_string($source)) {
// Loose validation: allows phar:// or file://
if (strpos($source, '://') !== false) {
return self::fromUrl($source);
}
// ... local file handling ...
}
}The remediation implemented in version 4.3.5 restricts the accepted schemes specifically to HTTP and HTTPS protocols. The updated logic parses the scheme and matches it against an explicit whitelist of safe communication protocols.
// Remediated validation pattern
public static function fromObject($source) {
if (is_string($source)) {
$scheme = parse_url($source, PHP_URL_SCHEME);
// Hardened check: only allow safe http/https schemes
if ($scheme && in_array(strtolower($scheme), ['http', 'https'])) {
return self::fromUrl($source);
}
// ... local file handling ...
}
}Exploitation of GHSA-2XMM-M4WV-3FJH requires a multi-stage attack path. The attacker must possess credentials or exploit a separate vector to upload a malicious PHP Archive disguised as an innocuous file type onto the server.
The uploaded file contains a serialized PHP object payload designed to trigger a gadget chain present within the application's dependencies. The attacker then submits the payload path using the phar:// stream wrapper to a feature that passes the value directly into the image resizer, such as a vulnerable backend template parameter or an unvalidated frontend input.
When the image resizer attempts to determine the dimensions or process the cached image, it invokes standard PHP filesystem functions. PHP parses the PHAR file structure, deserializes the malicious object from the metadata, and triggers the gadget chain's magic methods, executing code on the underlying operating system.
The primary impact of this vulnerability is untrusted deserialization (CWE-502). If an attacker successfully triggers the deserialization of a crafted PHAR archive, they can potentially achieve remote code execution (RCE) in the context of the web server process.
The realization of RCE depends on the presence of a valid deserialization gadget chain within the application's runtime class path. Since October CMS relies on various Laravel components and external Composer libraries, numerous well-documented gadget chains may exist.
The CVSS score is evaluated as 3.9 (Low) due to high requirements for privileges and user interaction. In standard configurations, only authenticated users with administrative or developer-level template authorization can supply strings directly to the image resizer component.
The recommended course of action is upgrading the October CMS package (october/october) to version 4.3.5 or higher. This version implements strict validation of schemes, restricting accepted remote protocols to HTTP and HTTPS.
If immediate upgrading is not possible, organizations should apply defense-in-depth mitigations. Developers must audit Twig templates and PHP controllers to ensure no user-controlled strings are passed directly into the |resize filter or the ResizeImages::resize() API.
Additionally, administrators can disable the PHAR stream wrapper by adding phar to the stream_resolve_include_path configuration or using third-party PHP extensions to restrict the protocols available to PHP's file handling functions.
CVSS:3.1/AV:N/AC:H/PR:H/UI:R/S:U/C:L/I:L/A:L| Attribute | Detail |
|---|---|
| CWE ID | CWE-20, CWE-502 |
| Attack Vector | Network (AV:N) |
| CVSS Score | 3.9 (Low) |
| Exploit Status | PoC (Proof of Concept) |
| KEV Status | Not Listed |
| Impact | Untrusted Deserialization / Code Execution |
The product receives input that is expected to be validated, but the validation is missing, incorrect, or incomplete, allowing unauthorized inputs.
CVE-2026-9799 is an authorization bypass vulnerability in the User-Managed Access (UMA) subsystem of Keycloak (org.keycloak.authorization). When a resource server is configured with the non-default PERMISSIVE policy enforcement mode, an authenticated user holding a valid UMA permission ticket for a single typed resource can obtain an authorization Requesting Party Token (RPT) granting access to all resources of the same type across the resource server. By submitting an authorization request with the type selector prefix 'resource-type-any:<type>', the attacker bypasses instance-level owner approval checks for unshared peer resources that lack explicit static administrative policies.
CVE-2026-9083 is a path traversal and information disclosure vulnerability (CWE-22) affecting Keycloak and the Red Hat build of Keycloak. The vulnerability resides in the Java Keystore key provider component (JavaKeystoreKeyProviderFactory), which fails to restrict the user-supplied keystore file path to a designated directory. An authenticated realm administrator holding the 'manage-realm' role can supply arbitrary absolute or relative filesystem paths when configuring a key provider. Because the application returns distinct error messages depending on whether the target path does not exist, lacks read permissions, or fails keystore parsing, an attacker can use this component as an oracle to probe the underlying host filesystem.
CVE-2026-9086 is a stored Cross-Site Scripting (XSS) vulnerability in Keycloak and the Red Hat build of Keycloak. The vulnerability stems from a case-sensitive URI scheme validation check in DefaultClientValidationProvider, allowing attackers with client registration or management permissions to inject javascript: or data: URIs using mixed-case characters.
CVE-2026-62290 is an incorrect authorization vulnerability (CWE-863) in cert-manager that allows low-privileged Kubernetes namespace tenants to bypass ClusterIssuer DNS01 solver policies and exfiltrate cluster-scoped DNS credentials. By directly creating standalone ACME Challenge custom resources rather than going through the standard Certificate and Order workflow, an attacker can specify arbitrary solver provider configurations while referencing a privileged ClusterIssuer. The cert-manager ACME challenge controller reconciles these directly instantiated Challenge resources without verifying their parentage or validating the user-supplied solver parameters against the referenced ClusterIssuer's selector policies.
CVE-2026-118662 (also tracked as GHSA-8g5g-mxqh-grgj) is an improper access control and path canonicalization vulnerability in the Vite frontend build and development tool. In version 6.4.4, the Vite development server fails to strip transform and loader query parameters before evaluating file paths against the `server.fs.deny` configuration directive. When the development server is exposed to a network and sensitive files located within `server.fs.allow` match patterns in `server.fs.deny`, an unauthenticated remote attacker can bypass file access restrictions by appending specific query strings such as `?raw` or `?inline` to HTTP requests. This allows arbitrary retrieval of restricted files, including environment configuration files and private cryptographic keys.
A vulnerability in the Rust `zip` crate (versions 1.3.0 prior to 9.0.0) allows an attacker to trigger an uncatchable process abort via a crafted ZIP archive. When extracting symbolic link entries on Unix or Windows targets, `ZipArchive::extract` passes the untrusted, attacker-controlled `uncompressed_size` header value directly to `Vec::with_capacity` before reading or validating entry payload data. By supplying a 112-byte archive with a ZIP64 extended information field declaring an excessive uncompressed size, an attacker forces the memory allocator to fail and invoke `handle_alloc_error()`, immediately terminating the entire process with `SIGABRT`.