Published on 09/03/2022 byNevulo
⚠️ This post is purely for explanatory purposes—you should not attempt to perform any type of security attack on a website unless you have approval.
Cross-site scripting is an attack performed on websites, where an attacker can inject some malicious code or scripts that get executed to modify the behaviour of that website.
This is ultimately a security vulnerability, and can be very dangerous if your website accepts and shows user input, but isn’t properly protected.
Cross-site scripting is an old attack, dating back to the 90s, so the name is a bit outdated. The scope of code injection has increased since then, and so “cross-site” isn’t as relevant these days – you can often perform most XSS attacks directly within the website.
But it doesn’t make it any less important today. The
There are two main types of cross-site scripting:
Reflected (non-persistent) XSS, is where the malicious input comes from the current HTTP request, in the form of
The malicious request reflects off the web server, where the response reflects the attack back to the user’s browser. Only works for that request, or anyone else who makes the same request.
Stored (persistent) XSS, is where the malicious input gets stored in a permanent place like a database. Each time the page gets requested, the malicious input gets requested from the database and included in the response, causing it to be rendered as raw HTML.
Ultimately, there’s no single way to perform an XSS attack, as there are many vectors to attack and considerations per website that can change what code will be required.
A common example includes something like this; let’s say there’s a website with a search box that displays my search input back to me:
I search for “nevulo”, sending a request to the server with my query (usually through query parameters, but somewhere in the request)
I go to a new webpage containing my search results, including the search term I used The important thing to understand is that the DOM contains your user input.
The problem starts to creep in when you consider that users can enter
The browser doesn’t know any different – it will just execute whatever HTML it sees as long as it’s valid. If I type in:
1<div style="background: blue; width: 32px; height: 32px;"></div>
It ends up showing we searched for a blue square. This is an example of reflected XSS.
If the website takes the input from those request parameters and returns a new page containing your input as-is, the renderer doesn’t know the difference between the real website code and the injected code.
So, whatever code you write in the input box in this vulnerable website, it will render the output as if the page as it would with your extra code.
But, this is dangerous ⛔️
Let’s say you’ve just built a new social media platform, and your platform gives users the ability to post little pieces of content on their personal, public wall. I’ll explain how you introduce this vulnerability in a site, but for now, just know your site is vulnerable to XSS.
I’ll sign up for your social platform because it’s wonderful. Now that I’ve got my account, I want to post my first piece of content to my wall! But, I’ve got a sneaky trick up my sleeve.
When I go to the input box to post on my wall, I’m going to type this in:
XSS is a dangerous vulnerability because if you can inject any valid HTML into a website, you can inject malicious elements like
All of these examples and more pose a serious threat to users of your website if vulnerable.
There are different types of XSS which can be used to attack websites, so there’s no one single way to introduce the vulnerability.
Reflected XSS is one of the most common types of attacks, and it typically becomes a vulnerability on most websites when the website accepts user input (typically query parameters) which is not sanitised properly. Then, that user input gets rendered in the response.
Imagine if a site had search functionality – it might look something like this:
searchquery parameter to the final HTML to be sent back to the user without any sanitisation:
1<html>2 <body>3 <p><?php echo 'You searched for: ' . $_GET["search"]; ?></p>4 </body>5</html>
scripttag gets executed, causing an alert to appear for the user.
Stored XSS, on the other hand, happens when some user input (say, for a comment on a blog), is stored in a permanent location like a database, and then that user content gets rendered back on the page.
If the input does not get sanitised properly, and it contains HTML code, it will load the user-generated content and render the content as HTML.
Stored XSS might look something like this in action:
A user sends a request to a server to update/create data that will be rendered on some page. This contains malicious HTML input.
At some point, a user makes a request and the server responds with the relevant data that the user tainted with raw HTML, and that data gets used directly within the HTML response and sent back to the user.
On most sites, you don’t want users to enter raw HTML code and return the user's input as code to be executed on your page.
Below is a non-exhaustive list of options for preventing XSS attacks on your website if your website accepts user input, and displays it back to them in any way (through an input box, query parameters, etc):
The first line of defence that can give some assurance that users won’t be able to enter certain characters you don’t want, like
> angled brackets that are used in HTML.
This is obviously not fool-proof, there are many workarounds for entering HTML without strictly using angled brackets.
But, if you’re just setting out to store somebody’s age and then display it back to them on their profile, it’s best to stop any malicious input at the start if you know the stored input will never need to use HTML characters.
Validation must be present on the backend, and not just the frontend. It’s easy for users to bypass the frontend validation and send bogus/malicious data to your backend. Your backend acts as a source of truth, since users cannot manipulate your backend logic.
If you’re returning a HTML document to users where the final result can be modified based on input parameters, before you return the document, you should escape any user input used within the document.
1<html>2 <body>3 <h1>Your name is <?php4 $new = htmlspecialchars($userInput, ENT_QUOTES);5 echo $new;6 ?></h1>7 </body>8</html>
If the input is:
The final output will look something like this:
Your name is <script>alert(1)</script>
This is a lot better rather than running the malicious code in the users' browser!
Generally, when escaping content, you’ll be passing the user input you want to show on the page through a function that will return the same input, just without any special characters used in HTML.
This will be different depending on which templating system you’re using (some do this by default, automatically). Make sure to look up a recent tutorial for escaping user input with whatever templating system you’re using.
Since XSS attacks rely on the attacker being able to run scripts on a vulnerable website, either by injecting scripts directly within the webpage, or getting the browser to download scripts from an untrusted third-party.
The content security policy works through “directives” that allow you to create an allowlist of sources of trusted content, and instructs the browser to only execute or render resources from those sources.
Content-Security-Policy to your website through the HTTP response headers looks like this:
1Content-Security-Policy: script-src 'self'
You can also specify the
Content-Security-Policy through the
meta tag in your HTML document:
1<meta http-equiv="Content-Security-Policy" content="script-src 'self'" />
In the policy above, there is one directive
script-src which controls a set of script-related privileges, and
self (the current page) is the only valid source. By default, it will also disable the ability to use inline scripts, which is one of the biggest security implications of using the policy.
This effectively means that any attacker attempting to inject code in your site will be unsuccessful and see an error.
Subscribe to the Nevuletter so you never miss new posts.
You need to be signed in to comment