Edit in GitHubLog an issue

Cross-site scripting (XSS)

Cross-site scripting, or XSS, is a security vulnerability that can be found in web applications. This vulnerability allows attackers to inject malicious code/styles into a web page viewed by users. Extension developers should be aware of these vulnerabilities to avoid introducing them in their code.

There are three main types of XSS vulnerabilities:

  • Persisted XSS - In this type of vulnerability, the source of unvalidated data comes from the database or Backend permanent store.
  • Reflected (non-persistent) XSS - This type of vulnerability occurs when data provided by a web client is used immediately by server-side scripts to parse and display a page to a user without properly sanitizing the request.
  • DOM XSS - For this vulnerability, the malicious data does not touch the web server. Rather, it is being reflected by the JavaScript code, fully on the client side.

Preventing XSS

XSS vulnerabilities can be prevented by validating and sanitizing user input as well as sanitizing dynamic values when rendering the view (HTML, mobile).

Input Processing

Any request data can be manipulated by attackers and can contain malicious values such as:

  • form fields filled with control characters ("<", ">" etc)
  • headers contain false IDs
  • URIs contain fake parts/query parameters
  • tampered cookies

To combat this, developers must validate any value coming in from requests.

It is better to validate/sanitize values as close as possible to the view context because only then you can be sure of the restrictions you have to impose on dynamic values and you are not risking security requirements for business requirements.

There is no reason, from a business standpoint, to disallow < > symbols in your users' "About me" section. By escaping control symbols when rendering HTML, allowing these characters would not be problematic. "About me" data may be delivered via RESTful API, where {} could cause issues. If sanitized earlier, the user data would be damaged and contain HTML control symbols (< >).

Output Processing

Output processing involves sanitizing strings that may have come from external data sources before using it to render views. It is the main method of protecting your extension from XSS attacks.

The general rule is: Do not trust dynamic values.

PHTML templates

The \Magento\Framework\Escaper class is provided for .phtml templates and PHP classes responsible for generating HTML. It contains HTML sanitization methods for a variety of contexts.

The $escaper local variable is available inside the .phtml templates. See the product listing template as example of $escaper usage in .phtml templates.

See Template guide to read more about templates in Magento.

When using the \Magento\Framework\Escaper or $escaper:

  • If a method indicates that the content is escaped, do not escape: getTitleHtml(), getHtmlTitle() (the title is ready for HTML output)
  • Type casting and the php function count() do not need escaping (for example echo (int)$var, echo (bool)$var, echo count($var))
  • Output in single quotes does not need escaping (for example echo 'some text')
  • Output in double quotes without variables does not need escaping (for example echo "some text")
  • For all other cases, escape the data using context-specific escape functions.

The following code sample illustrates XSS-safe output in templates:

Copied to your clipboard
<?= $block->getTitleHtml() ?>
<?= $block->getHtmlTitle() ?>
<?= $escaper->escapeHtml($block->getTitle()) ?>
<?= (int)$block->getId() ?>
<?= count($var); ?>
<?= 'some text' ?>
<?= "some text" ?>
<a href="<?= $escaper->escapeUrl($block->getUrl()) ?>"><?= $block->getAnchorTextHtml() ?></a>

When to use Escaper methods:

Case: JSON inside an HTML attribute

Escaper method: escapeHtmlAttr

Copied to your clipboard
<div data-bind='settings: <?= $escaper->escapeHtmlAttr($myJson) ?>'></div>

Case: JSON inside script tag

Escaper method: no sanitization needed

Copied to your clipboard
<script>
let settings = <?= $myJson ?>
</script>

Case: HTML tag content that should not contain HTML

Escaper method: escapeHtml

Pass in an optional array of allowed tags that will not be escaped.

If a tag is allowed, the following attributes will not be escaped: id, class, href, style and title. Any other attribute for that allowed tag will be escaped.

embed, iframe, video, source, object, audio, script and img tags are not allowed, regardless of the content of this array.

Copied to your clipboard
<span class="label"><?= $escaper->escapeHtml($block->getLabel()) ?></span>
// Escaping translation
<div id='my-element'>
<?= $escaper->escapeHtml(__('Only registered users can write reviews. Please <a href="%1">Sign in</a> or <a href="%2">create an account</a>', $block->getLoginUrl(), $block->getCreateAccountUrl()), ['a']) ?>
</div>

Case: URL inside certain HTML attributes

Escaper method: escapeUrl

Certain attributes like a.href accept URIs of various types and must be sanitized.

Copied to your clipboard
<a href="<?= $escaper->escapeUrl($myUrl) ?>">Click me</a>
<div attr-js-extracts="<?= $escaper->escapeHtmlAttr($myOtherUrl) ?>"></div>

Case: All JavaScript inside attributes must be escaped by escapeJs before escapeHtmlAttr:

Escaper method: escapeJS

Copied to your clipboard
<div
onclick="<?= $escaper->escapeHtmlAttr('handler("' . $escaper->escapeJs($aString) . '", ' . $escaper->escapeJs($aVar) .')') ?>">
My DIV
</div>

Case: JavaScript string that must not contain JS/HTML

Escaper method: escapeJS

Copied to your clipboard
<script>
let phrase = "Hi, my name is <?= $escaper->escapeJs($myName) ?>";
//Do not use HTMl context methods like escapeUrl
let redirectUrl = "<?= $escaper->escapeJs($myUrl) ?>";
location.href = redirectUrl;
</script>

Case: JavaScript variable that must not contain JS/HTML

Escaper method: escapeJS

Copied to your clipboard
<script>
let <?= $escaper->escapeJs($dynamicVariable) ?> = <?= $myJson ?>;
settings.<?= $escaper->escapeJs($myProperty) ?> = true;
</script>

Knockout templates

In knockout templates, you can bind a UI component's property/function as the 'inner HTML' of an element. Such properties may contain dynamic data and must be sanitized within componenets. See binding syntax to learn more about UI component templates.

In order to notify developers that these properties/function results may contain HTML, the application requires (with the help of a static test) that you name such properties/functions using "UnsanitizedHtml" suffix.

Copied to your clipboard
<div data-bind="html: propUnsanitizedHtml"></div>
<p html="returnUnsanitizedHtml()"></p>

Dynamically created DOM elements

When using variables that are not supposed to contain HTML, the safest way to generate DOM elements is to create them programatically using the appropriate API, instead of using the innerHtml property or jQuery's .html() function.

Copied to your clipboard
let newDiv = document.createElement("div");
newDiv.innerText = userName;
newDiv.setAttribute("custom-attribute", myAttribute);
parentElement.appendChild(newDiv);

UI component data providers

UI component data providers pass dynamic (user generated) data to UI components. The data they return is rendered in order to support component dynamic linking. Since user data is supposed to be treated as a literal value which is not refering to any other component, rendering of these properties must be disabled. See UI component data providers to read more about data providers and component linking.

Copied to your clipboard
$uiData = ['linkProperty' => '${ $.otherComponent.value }'];
$uiData['customer'] = $customer->getData();
//Customer data will be taken as is, linkProperty will be retrieved from otherComponent and taken as is
$uiData['__disableTmpl'] = ['customer' => true, 'linkProperty' => 1];
return $uiData;

Static Test

To check your .phtml template for XSS vulnerabilities, use the Magento2.Security.XssTemplate sniff from Coding Standard. This sniff finds all echo calls in PHTML-templates and determines if the output is properly escaped. It covers the following cases:

  • /* @noEscape */ before output. Output does not require escaping. Test is green.
  • /* @escapeNotVerified */ before output. Output escaping is not checked and should be verified. Test is green.
  • Methods which contain "html" in their names (for example echo $object->{suffix}Html{postfix}()). Data is ready for the HTML output. Test is green.
  • AbstractBlock methods escapeHtml, escapeHtmlAttr, escapeUrl, escapeJs are allowed. Test is green.
  • Type casting and php function count() are allowed (for example echo (int)$var, (bool)$var, count($var)). Test is green.
  • Output in single quotes (for example echo 'some text'). Test is green.
  • Output in double quotes without variables (for example echo "some text"). Test is green.
  • Other of previously mentioned. Output is not escaped. Test is red.
  • Privacy
  • Terms of Use
  • Do not sell or share my personal information
  • AdChoices
Copyright © 2024 Adobe. All rights reserved.