PatchFirst · Latest vulnerabilities
Vulnerabilities published for PHP in the last 30 days: 179. Listed in the KEV catalogue: 2.
Fresh vulnerability reports from vendor announcements and security news. These warnings do not change the security rating.
According to OpenNet, PHP has a remote code execution vulnerability (CVE-2026-115560). The exact affected versions are unspecified. The report does not name a fixed version.
What to do now: Check whether a PHP update with the fix is available and install it.
CVE-2026-115560
Vulnerabilities: 7190
The MCP PHP SDK (Composer package mcp/sdk) is the official Model Context Protocol SDK for PHP. In versions 0.5.0 through 0.7.0, the HTTP client transport reads a Server-Sent Events response stream incrementally and appends each chunk to an in-memory buffer with no upper bound. The buffer is only flushed when an SSE event delimiter, a double newline, is found, so a remote MCP server that streams response bytes without ever sending the delimiter causes the buffer to grow without limit. A malicious, compromised, or man-in-the-middle-controlled server that the client connects to over the HTTP transport can exploit this to exhaust the client proce
Winter CMS versions before 1.2.13 contain an incomplete fix for a Twig sandbox escape vulnerability in System\\Twig\\SecurityPolicy that allows authenticated backend users with template-editing permissions to bypass sandbox restrictions. Attackers can exploit method forwarding through Eloquent models and query builders using methods like saveQuietly(), deleteQuietly(), increment(), decrement(), and newQuery() to read and modify arbitrary database records, execute arbitrary SQL, and achieve remote code execution by injecting PHP into template code sections.
Winter CMS before 1.2.13 contains a local file inclusion vulnerability in the JavascriptImporter filter that allows authenticated users with cms.manage_assets permission to disclose arbitrary server-readable files by placing =include or =require directives in theme JavaScript assets. Attackers can reference files like .env outside the theme directory, and the combined output served through the combine route becomes readable by unauthenticated visitors, exposing application keys and database credentials.
Grav before 2.0.16 contains an incomplete default denylist in the Twig sandbox configuration that fails to block access to system configuration secrets. Attackers with page-edit permission can use config.get() or config.toArray() in Twig templates to retrieve sensitive values like system.cache.redis.password when config_access is enabled.
Grav before 2.0.16 allows sandboxed Twig templates to access sensitive User fields through allow-listed offsetGet() and offsetexists() methods that lack field filtering. Attackers with page-edit permissions can call offsetGet() on User objects to extract hashed passwords and 2FA secrets, enabling offline password cracking and authentication bypass.
Grav CMS before 2.0.16 contains an origin validation bypass in the Uri::referrer() and Pages::referrerRoute() methods, which validate the Referer header using an unanchored string prefix match (str_starts_with($referrer, $base)) with no trailing delimiter. An attacker who controls a domain that begins with the victim site's origin (e.g. https://example.com.attacker.tld) can send a request with such a Referer to be treated as same-origin, bypassing the Referer-based origin check.
Grav CMS before 2.0.16 contains a timing vulnerability in Utils::verifyNonce() that uses non-constant-time string comparison with the === operator instead of hash_equals() for CSRF nonce validation. Attackers can measure response timing differences to recover valid nonce values byte-by-byte through multiple requests, weakening CSRF protection below its intended security margin.
Grav CMS before 2.0.16 fails to filter system, site, and theme configuration arrays in sandboxed Twig renders, allowing content editors to read sensitive configuration values. Attackers with page-content edit access can access raw configuration arrays including secrets like cache credentials by using dot notation in Twig templates, bypassing the config_denied_paths restrictions.
Grav CMS before 2.0.16 contains a path traversal vulnerability in the media_directory() Twig function that fails to validate filesystem paths, allowing authenticated users to enumerate and access files outside intended scope. Attackers with page authoring privileges can supply arbitrary filesystem paths to media_directory() and use the allow-listed filepath accessor on Medium objects to read file contents of any file matching configured media extensions that the web server process can access.
Grav before 2.0.16 contains a path traversal vulnerability in MediaUploadTrait::deleteFile() that allows authenticated users with media management permissions to delete arbitrary files by supplying filenames with directory traversal sequences. The method validates only the basename portion of the filename while preserving unvalidated directory paths containing ../ sequences that are passed to unlink(), enabling deletion of files outside the intended media storage directory.
CakePHP is a rapid development framework for PHP. Prior to versions 4.5.12, 4.6.5, 5.1.8, 5.2.14, and 5.3.7 on their respective release lines, custom mail headers added with Message::setHeaders() or Message::addHeaders() do not have CRLF bytes removed, allowing header injection when user-controlled data is used in message headers. This issue is fixed in versions 4.5.12, 4.6.5, 5.1.8, 5.2.14, and 5.3.7.
CakePHP is a rapid development framework for PHP. Prior to versions 5.1.10, 5.2.15, and 5.3.7 on their respective release lines, FunctionsBuilder::jsonValue() with PostgresDriver is vulnerable to SQL injection when user-controlled data is supplied to the jsonPath parameter. This issue is fixed in versions 5.1.10, 5.2.15, and 5.3.7.
Filament is a collection of full-stack components for accelerated Laravel development. Prior to versions 4.12.0 and 5.7.0, incorrect challenge-form required-field handling allows app-based multi-factor authentication to be bypassed when recovery codes are enabled. Email-based multi-factor authentication is not affected. This issue is fixed in versions 4.12.0 and 5.7.0.
YOURLS is a self-hosted, customizable URL shortener written in PHP. From 1.5.1 until 1.10.4, YOURLS stores the HTTP Referer header through yourls_get_referrer(), yourls_sanitize_url_safe(), and yourls_log_redirect(), then aggregates the value in yourls-infos.php and passes the derived domain through yourls_get_domain(), yourls_stats_pie(), and yourls_google_array_to_data_table(). The chart builder concatenates labels into inline JavaScript without JavaScript-string escaping, so an unauthenticated attacker can poison the statistics of an existing short URL with a crafted referrer. When an administrator or public stats-page viewer opens the aff
Phalcon is a high-performance, full-stack PHP framework. In 5.15.0 and earlier, resolveFilter in phalcon/Mvc/View/Engine/Volt/Compiler.zep builds the join filter by inserting the raw separator and array token values into generated PHP without passing them through expression(). An attacker who can influence Volt template source can place quote-breaking content in a join argument, inject PHP into the compiled cache file, and execute it when Phalcon\Mvc\View\Engine\Volt::render() loads the template. This issue is fixed in version 5.16.0.
### Impact Affected versions of Winter CMS render the `search` query parameter without HTML encoding inside a `<script type="text/template">` block in the backend Table widget partial (`modules/backend/widgets/table/partials/_table.php`): ```php value="<?= get('search') ?>" ``` `<script>` is an HTML raw-text context, so the surrounding `value="…"` attribute quoting is not a parser boundary. A literal `</script>` in the query string terminates the template element early, and everything after it is parsed as ordinary markup in the backend document. Any backend page rendering a Table or DataTable widget is a sink. The value is read from the
### Impact Affected versions of Winter CMS allow a backend AJAX handler to be invoked by a plain top-level `GET` navigation with no CSRF token. `Backend\Classes\Controller::actionExists()` accepted any public method on a controller as a page action, so handler-shaped names were never reserved from URL dispatch: an authenticated and authorized request to `/backend/system/eventlogs/index_onEmptyLog` reached the handler of the same name and truncated the system event log. Backend paths are routed through `Route::any`, and CSRF validation is skipped for `HEAD`, `GET` and `OPTIONS` requests. The default `SameSite=Lax` session cookie is sent on t
### Impact Users with the `backend.manage_branding` ("Customize the back-end") or `backend.manage_editor` ("Manage global code editor preferences") permission can provide custom CSS through **Settings → Customize Backend → Styles** or **Settings → Editor Settings → Markup Styles** that is compiled through the LESS CSS parser and rendered on every backend page. v1.2.13 addressed CVE-2026-32257 and CVE-2026-32258 by applying `strip_tags()` to the compiled output of `BrandSetting::renderCss()` and `EditorSetting::renderCss()`. That fix was incomplete. Both methods cache the **raw** compiler output with `Cache::forever()` and applied `strip_tag
### Impact Affected versions of Winter CMS did not enforce the `ImportExportController` behavior's granular access control on the handlers that actually perform the work. The behavior supports per-operation access control through the `import[permissions]` and `export[permissions]` configuration keys, enforced by `userHasAccess()`. That check was applied only to the `import()` and `export()` page actions. `Backend\Classes\Controller::execAjaxHandlers()` dispatches AJAX handlers and returns *before* `execPageAction()` runs, and the behavior binds its import and export form widgets in its constructor on every request to the controller. The ha
### Impact `Backend\Controllers\MyAccount`, introduced in v1.2.13, declares an empty `$requiredPermissions` array so that any authenticated backend user can manage their own account. It implements the `FormController` behavior, which exposes three routable actions — `create`, `update` and `preview` — that each take a record id from the URL. `index()` passes the authenticated user's own id to the behavior, but the inherited actions were left routable and `formFindModelObject()` was not scoped, so a caller-supplied id resolved against an unscoped `Backend\Models\User` query: ``` GET /backend/backend/myaccount/preview/{other_user_id} ``` `pr
Check your installed stack and versions against the current vulnerability feeds.