Security best practices

If you plan on developing a plugin or contributing to Matomo (formerly Piwik) Core your code must be secure.

This guide contains a list of methods to combat certain vulnerabilities. Follow all of them when working on your plugin or contribution.

Preventing XSS

XSS is the injection of malicious scripts (e.g. JavaScript) into the user interface. It can allow attackers to gain control of the application or steal information.

Attackers can achieve that either by:

  • storing malicious scripts in data (like website names for Matomo)
  • passing malicious scripts as HTTP request parameter/data

Get request parameters via Common::getRequestVar()

In your PHP code, if you need access to a variable in $_GET or $_POST, always use Common::getRequestVar().

getRequestVar() will sanitize the request variable. If an attacker passes a string containing <script>...</script>, it will be sanitized to &lt;script&gt;...&lt;/script&gt;. This will help to avoid accidentally embedding unescaped text in HTML output.

For text you know may contain special characters or if you need to output text in a format that doesn't need XML/HTML sanitization (like JSON), call Piwik::unsanitizeInputValues() to undo the sanitization.

Note: You can sanitize text that isn't in a request parameter by using Piwik::sanitizeInputValues().

Use the correct Twig escaping strategy

It's important to use the correct escaping strategy depending on where you are printing a value. For example, when printing a value in an HTML attribute like "title" or "label", then we need to use an |e('html_attr') escaping strategy. Other escaping strategies are js, css, url and html (which is the default strategy).

Use |raw sparingly in Twig templates

When writing Twig templates, try to avoid using the |raw filter when possible. As an alternative, put the HTML you want to reuse in a separate template and {% include %} it.

If you do use |raw, make sure what you're inserting has been properly escaped.

If you have to use the |raw filter you may want to think about using the |rawSafeDecoded filter which may be more secure.

Be careful when using jQuery.html()

In your JavaScript, be careful when using the $.html method to insert HTML into the DOM. Make sure the string you are inserting came from Matomo and has been escaped.

If you know that the text you're inserting shouldn't be HTML, then do not use $.html(), instead use $.text() or $.val(). For example:

var ajaxData = getDataFromAjax();
$('#someLabel').text(ajaxData.labelToUse);

To escape strings in JavaScript you may use the helper method piwikHelper.escape for example:

var safeString = piwikHelper.escape( userInputUnsafeString );
$('#someLabel').text( safeString );

Use a restrictive Content Security Policy header

Content Security Policy is set by default in Matomo (since Matomo 4.6.0) and plugins can modify the CSP as needed using $this->securityPolicy methods in the controller, for example:

$this->securityPolicy->addPolicy('image-src', 'self');

Or using dependency injection by creating a plugins/MyPlugin/config/config.php like this to apply the change to every UI request instead of only a specific request:

return array(
    \Piwik\View\SecurityPolicy::class => DI\decorate(function ($previous) {
        /** @var \Piwik\View\SecurityPolicy $previous */

        if (!\Piwik\SettingsPiwik::isMatomoInstalled()) {
            return $previous;
        }

        $previous->addPolicy('default-src', 'my-cdn.example.com');
        return $previous;
    }),
);

Consider using the default, or a more restrictive CSP if possible which will protect against certain XSS attacks.

See Also

XSS on MDN

Preventing CSRF

CSRF attacks is where an attacker make a Matomo user perform an action unwillingly. To achieve that, the attacker sends a link to the user. The link could, for example, point to a Matomo controller method that changes the user's password, or delete a site.

This attack can be prevented with the following technique:

Check for the token_auth

In every controller method you create that changes Matomo settings, changes a user's settings or does some other admin level function, call the Controller::checkTokenInUrl() method. For example:

// method in a controller
public function doSomeAdminStuff()
{
    $this->checkTokenInUrl();

    // ...
}

In every API method that executes some admin level function, make sure to check for the proper user permissions by calling one of the Piwik::check... methods. For example:

// method in an API class
public function changeSettingsForUser($userLogin)
{
    Piwik::checkUserHasSuperUserAccessOrIsTheUser($userLogin);
}

token_auth in the browser

Your JavaScript should send the token_auth to controller and API methods that need it, but you should make sure the token_auth never appears in the URL. This way, it will never be saved or cached by the browser.

To keep the token_auth out of a browser cache, you can use POST requests.

See Also

CSRF on MDN

Preventing SQL Injection

SQL Injection is the manipulation of the application's SQL queries by injecting malicious SQL statements. Attackers can inject malicious SQL through inputs of the application: form fields, request parameters, …

For example, if an application builds an SQL query like this:

$sql = 'SELECT * from mytable where id = ' . $_GET['id'];

An attacker could pass "1 OR 1" for the id URL parameter. This would cause the following query to be executed: SELECT * from mytable where id = 1 OR 1, which would output every row of mytable.

SQL injection can be prevented by doing one thing:

Use SQL prepared statements

When writing SQL statements, use SQL prepared statements instead of directly inserting variables into your statement. SQL prepared statements means using placeholders in your SQL queries.

In other words, don't do this:

$idSite = Common::getRequestVar('idSite');
// DON'T DO THIS!!
$sql = "SELECT * FROM " . Common::prefixTable('site') . " WHERE idsite = " . $idSite;

$rows = Db::query($sql);

Instead, do this:

$idSite = Common::getRequestVar('idSite');
$sql = "SELECT * FROM " . Common::prefixTable('site') . " WHERE idsite = ?";

$rows = Db::query($sql, array($idSite));

Preventing Remote File Inclusion

Remote File Inclusion is the inclusion and execution of source code that is not part of the webapp. It happens in PHP with include or require statements that use a path determined by the user.

In Matomo, the best way to prevent remote file inclusion attacks is to just never require/include files using data from the user. Instead, put logic in classes that can be loaded by Matomo's autoloader and instantiate/use different classes based on data obtained from the user. In other words, don't do this:

$clientToUse = Common::getRequestVar('seoClient');

// DON'T DO THIS!!
require_once PIWIK_INCLUDE . '/plugins/MyPlugin/Clients/' . $clientToUse . '.php';

$client = new $clientToUse();

// ... use $client ...

Instead, do this:

$clientToUse = Common::getRequestVar('seoClient');

if ($clientToUse == 'mySeoProvider') {
    $client = new Clients\MySeoProvider();
} else if ($clientToUse == 'myOtherSeoProvider') {
    $client = new Clients\MyOtherSeoProvider();
} else {
    throw new Exception("Invalid SEO provider client: $clientToUse!");
}

// ... use $client ...

Other Coding Guidelines

Here are some other coding guidelines that will help make your code more secure:

  • PHP files should start with a <?php tag that is never closed.

  • Use the .php extension for all your PHP scripts.

  • Avoid executing php code using one of the following functions: eval, exec, passthru, system, popen or preg_replace (with the "e" modifier).

  • Make sure that accessing your files directly doesn't execute any code that could have an impact on your Matomo install.

  • Make sure your code doesn't rely on register_globals set to On. Note: PHP5 sets register_globals to Off by default.

  • For timing attack safe equal comparisons use Common::hashEquals() method

  • If your plugin has admin functionality (functionality only an administrator or the super user can use) then your plugin's Controller must extend Piwik\Plugin\ControllerAdmin.

Checklist

By no means is below a complete checklist. You'll always be required to still think of any other case something can be used in a way we don't want it to be used. It's only a list for the most common checks.

  • Authorisation checks: Any controller action or api method has an access/permission check
  • CSRF checks: Any form or action or api that changes data has a CSRF nonce or token check
  • SQL injection checks: Database parameters use bound parameters or cast values to int
  • XSS checks: User input is escaped (also in JavaScript) see XSS section in this guide. Ensure the correct escaping strategy is used
  • Timing attack checks: For sensitive equal comparisons Common::hashEquals is used
  • Exposing of data: No sensitive data is exposed (any tokens or passwords eg in the HTML as part of a system check or logs or so)
  • Secure data storing: Passwords or similar data like a sessionId is stored securely hashed
  • Brute Force Attack: If there is any token or password like authentication then we use our brute force prevention feature
  • External links have a rel="noreferrer noopener" (although newer browser versions add this automatically). The noopener is needed for security reasons. The noreferrer for privacy reasons. See rel=noreferrer and rel=noopener
  • Email notifications are sent for critical actions related to login credentials, tokens, 2FA etc and ideally also for any other crticial configuration change such as brute force changes etc.
  • Unsafe methods: Methods to watch out for:
    • Insecure deserialization: unserialize is unsafe for user input. Consider using Common::safe_unserialize or for user input possibly JSON instead
    • file_get_contents can be unsafe when eg a protocol phar:// can be passed to it
    • require|include(_once) should not use any user input unless it's whitelisted
    • Also see eg phpdangerousfuncs.md
  • Think of other ways such as RCE etc
  • Password confirmation for critical actions related to login credentials, tokens, 2FA etc and ideally also other critical/sensitive changes like installing/activating plugins etc.
  • Exception stack traces should only be shown in the UI when it was specifically enabled (Piwik_ShouldPrintBackTraceWithMessage()===true). Be aware that echoing an \Exception or using it as a string when logging (echo $exception) can cause the exception to be indirectly appear with a stack trace.

Learn more