- Do you work in a **cybersecurity company**? Do you want to see your **company advertised in HackTricks**? or do you want to have access to the **latest version of the PEASS or download HackTricks in PDF**? Check the [**SUBSCRIPTION PLANS**](https://github.com/sponsors/carlospolop)!
- **Join the** [**💬**](https://emojipedia.org/speech-balloon/) [**Discord group**](https://discord.gg/hRep4RUj7f) or the [**telegram group**](https://t.me/peass) or **follow** me on **Twitter** [**🐦**](https://github.com/carlospolop/hacktricks/tree/7af18b62b3bdc423e11444677a6a73d4043511e9/\[https:/emojipedia.org/bird/README.md)[**@carlospolopm**](https://twitter.com/carlospolopm)**.**
- **Share your hacking tricks by submitting PRs to the [hacktricks repo](https://github.com/carlospolop/hacktricks) and [hacktricks-cloud repo](https://github.com/carlospolop/hacktricks-cloud)**.
The CORS (Cross-origin resource sharing) standard is needed because it **allows servers to specify who can access its assets** and which **HTTP request methods are allowed** from external resources.
A **same-origin** policy, requiers that both the **server requesting** a resource and the server where the **resource** is located uses the same protocol ([http://),domain](http://\),domain) name (internal-web.com) and the same **port** (80). Then, if the server forces the same-origin policy, only web pages from the same domain and port will be able to access the resources.
The specification of `Access-Control-Allow-Origin` allows for **multiple origins**, or the value **`null`**, or the wildcard **`*`**. However, **no browser supports multiple origins** and there are **restrictions** on the use of the **wildcard**`*`.(_The wildcard can only be used alone, this will fail `Access-Control-Allow-Origin: https://*.normal-website.com` and it cannot be used with_ _Access-Control-Allow-Credentials: true_)
The **default** behaviour of cross-origin resource requests is for **requests** to be **passed without credentials** like cookies and the Authorization header. However, the cross-domain server can **permit reading** of the **response** when **credentials** are **passed** to it by setting the CORS **`Access-Control-Allow-Credentials`** header to **`true`**.
* includes a **non-standard HTTP method (HEAD, GET, POST)**
* includes new **headers**
* includes special **Content-Type header value**
{% hint style="info" %}
**Check** [**in this link**](https://developer.mozilla.org/en-US/docs/Web/HTTP/CORS#simple\_requests) **the conditions of a request to avoid sending of a pre-flight request**
{% endhint %}
the cross-origin request is preceded by a **request** using the **`OPTIONS`** **method**, and the CORS protocol necessitates an initial check on what **methods and headers are permitted prior to allowing the cross-origin request**. This is called the **pre-flight check**. The server **returns a list of allowed methods** in addition to the **trusted origin** and the browser checks to see if the requesting website's method is allowed.
{% hint style="danger" %}
Note that **even if a pre-flight request isn't sent** because the "regular request" conditions are respected, the **response needs to have the authorization headers** or the **browser****won't be able to read the response** of the request.
For **example**, this is a pre-flight request that is seeking to **use the `PUT` method** together with a **custom** request **header** called `Special-Request-Header`:
Note that usually (depending on the content-type and headers set) in a **GET/POST request no pre-flight request is sent** (the request is sent **directly**), but if you want to access the **headers/body of the response**, it must contains an _Access-Control-Allow-Origin_ header allowing it.\
**Therefore, CORS doesn't protect against CSRF (but it can be helpful).**
Notice that most of the **real attacks require `Access-Control-Allow-Credentials`** to be set to **`true`** because this will allow the browser to send the credentials and read the response. Without credentials, many attacks become irrelevant; it means you can't ride on a user's cookies, so there is often nothing to be gained by making their browser issue the request rather than issuing it yourself.
One notable exception is when the **victim's network location functions as a kind of authentication.** You can use a victim’s browser as a proxy to bypass IP-based authentication and access intranet applications. In terms of impact this is similar to DNS rebinding, but much less fiddly to exploit.
It is also true that a lot of developers want to **allow several URLs in the CORS**, but subdomain wildcards or lists of URLs aren't allowed. Then, several developers **generate** the \*\*`Access-Control-Allow-Origin`\*\*header **dynamically**, and in more than one occasion they just **copy the value of the Origin header**.
In other cases, the developer could check that the **domain** (_victimdomain.com_) **appears** in the **Origin header**, then, an attacker can use a domain called **`attackervictimdomain.com`** to steal the confidential information.
`null` is a special value for the **Origin** header. The specification mentions it being triggered by redirects, and local HTML files. Some applications might whitelist the `null` origin to support local development of the application.\
This is nice because **several application will allow this value** inside the CORS and any **website can easily obtain the null origin using a sandboxed iframe**:
If you found the domain _victim.com_ to be **whitelisted** you should check if _victim.com.**attacker.com**_ is **whitelisted also**, or, in case you can **takeover some subdomain**, check if _**somesubdomain**.victim.com_ is whitelisted.
Most of the regex used to identify the domain inside the string will focus on alphanumeric ASCII characters and `.-` . Then, something like `victimdomain.com{.attacker.com` inside the Origin header will be interpreted by the regexp as if the domain was `victimdomain.com` but the browser (in this case Safari supports this character in the domain) will access the domain`attacker.com` .
**For more information and settings of this bypass check:** [**https://www.corben.io/advanced-cors-techniques/**](https://www.corben.io/advanced-cors-techniques/) **and** [**https://medium.com/bugbountywriteup/think-outside-the-scope-advanced-cors-exploitation-techniques-dad019c68397**](https://medium.com/bugbountywriteup/think-outside-the-scope-advanced-cors-exploitation-techniques-dad019c68397)
One defensive mechanism developers use against CORS exploitation is to white-list domains that frequently requests access for information. However, this isn’t entirely secure, because if even **one** of the subdomains of the **whitelisted** domain is **vulnerable** to other exploits such as **XSS**, it can enable CORS exploitation.
Let us consider an example, the following code shows the configuration that allows subdomains of _requester.com_ to access resources of _provider.com_.
Assuming that a user has access to sub.requester.com but not requester.com, and assuming that `sub.requester.com` is vulnerable to XSS. The user can exploit `provider.com` by using cross-site scripting attack method.
If the stars are aligned we may be able to use server-side cache poisoning via HTTP header injection to create a [stored XSS](https://portswigger.net/web-security/cross-site-scripting/stored) vulnerability.
If an application **reflects** the **Origin header** without even checking it for illegal characters like , we effectively have a **HTTP header injection vulnerability against IE/Edge users as Internet Explorer and Edge view \r (0x0d) as a valid HTTP header terminator**:`GET / HTTP/1.1`\
This isn't directly exploitable because there's no way for an attacker to make someone's web browser send such a malformed header, but I can **manually craft this request in Burp Suite and a server-side cache may save the response and serve it to other people**. The payload I've used will change the page's character set to **UTF-7**, which is notoriously useful for creating XSS vulnerabilities.
You may have occasionally encountered a page with [reflected XSS](https://portswigger.net/web-security/cross-site-scripting/reflected) in a custom HTTP header. Say a web page reflects the contents of a custom header without encoding:
With CORS, we can send any value in the Header. By itself, **that's useless** since the response containing our **injected JavaScript won't be rendered**. However, **if Vary: Origin hasn't been specified** the response **may be stored in the browser's cache and displayed directly when the browser navigates to the associated URL**. I've made a fiddle to [attempt this attack on a URL of your choice](https://jsfiddle.net/3gk8u8wu/3/). Since this attack uses client-side caching, it's actually quite reliable.
```markup
<script>
function gotcha() { location=url }
var req = new XMLHttpRequest();
url = 'https://example.com/'; // beware of mixed content blocking when targeting HTTP sites
XSSI designates a kind of vulnerability which exploits the fact that, when a resource is included using the `script` tag, the SOP doesn’t apply, because scripts have to be able to be included cross-domain. An attacker can thus read everything that was included using the `script` tag.
This is especially interesting when it comes to dynamic JavaScript or JSONP when so-called ambient-authority information like cookies are used for authentication. The cookies are included when requesting a resource from a different host. BurpSuite plugin: [https://github.com/kapytein/jsonp](https://github.com/kapytein/jsonp)
Try to add a **`callback`** **parameter** in the request. Maybe the page was prepared to send the data as JSONP. In that case the page will send back the data with `Content-Type: application/javascript` which will bypass the CORS policy.
You can ask a web-application to make a request for you and send back the response. This will bypass the **`Access-Control-Allow-Origin`** but notice that the **credentials to the final victim won't be sent** as you will be **contacting a different domain** (the one that will make the request for you).
CORS-escape provides a **proxy** that **passes** on our **request** along with its **headers**, and it also **spoofs** the **Origin** header (Origin = **requested domain**). So the **CORS policy is bypassed**.\
Proxying is kinda like “passing on" your request, exactly as you sent it. We could solve this in an alternative way that still involves someone else making the request for you, but this time, **instead of using passing on your request, the server makes its own request, but with whatever parameters you specified.**
You can **bypass CORS checks** such as `e.origin === window.origin` by **creating an iframe** and **from it opening a new window**. More information in the following page:
Basically you make the **victim access your page**, then you change the **DNS of your domain (the IP)** and make it **points** to your **victims web page**. You make your **victim execute** (**JS**) something when the **TTL is****over** so a new DNS request will be made and then you will be able to gather the information (as you will always maintains **the user in your domain**, he won't send **any cookie** to the victim server, so this options **abuses the special privileges of the IP of the victim**).
Even if you set the **TTL very low** (0 or 1) **browsers have a cache** that will **prevent** you from **abusing** this for several seconds/minuted.
So, this technique is useful to **bypass explicit checks** (the victim is **explicitly performing a DNS request** to check the IP of the domain and when the bot is called he will do his own).
Or when you can have a **user/bot in the same page for a long time** (so you can **wait** until the **cache expires**).
If you need something quick to abuse this you can use a service like [https://lock.cmpxchg8b.com/rebinder.html](https://lock.cmpxchg8b.com/rebinder.html).
If you want to run your own DNS rebinding server you can use something like [**DNSrebinder**](https://github.com/mogwailabs/DNSrebinder)**,** then **expose** your **local port 53/udp**, create an **A registry pointing to it** (ns.example.com), and create a **NS registry** pointing to the **previously created A subdomain**(ns.example.com).\
Then, any subdomain of that subdomain (ns.example.com), will be resolved by your host.
As it was explained in the previous section, **browsers** have the IPs of domains **cached more time** than the one specified in the TTL. However, there is a way to bypass this defence.
_Blue is the first DNS request and orange is the flood._
### DNS Rebinding via **Cache**
As it was explained in the previous section, **browsers** have the IPs of domains **cached more time** than the one specified in the TTL. However, there is another way to bypass this defence.
You can **create 2 A records** (or **1 with 2 IPs**, depending on the provider) for the **same subdomain** in the **DNS provider** and when a browser checks for them he will get both.
Now, if the **browser** decides to **use** the **attacker IP address first**, the **attacker** will be able to **serve** the **payload** that will **perform HTTP requests** to the same **domain**. However, now that the attacker knows the IP of the victim, **he will stop answering the victim browser**.
When the browser finds that the **domain isn't responding** to him, it will **use the second given IP**, so he will **access a different place bypassing SOP**. The attacker can abuse that to **get the information and exfiltrate it**.
Note that in order to access localhost you should try to rebind 127.0.0.1 in Windows and 0.0.0.0 in linux.\
Providers such as godaddy or cloudflare didn't allow me to use the ip 0.0.0.0, but AWS route53 allowed me to create one A record with 2 IPs being one of them "0.0.0.0"
* If **internal IPs aren't allowed**, they might **forgot forbidding 0.0.0.0** (works on Linux and Mac)
* If **internal IPs aren't allowed**, respond with a **CNAME** to **localhost** (works on Linux and Ma
* If **internal IPs aren't allowed** as DNS responses, you can respond **CNAMEs to internal services** such as www.corporate.internal.
### DNS Rebidding Weaponized
You can find more information about the previous bypass techniques and how to use the following tool in the talk [Gerald Doussot - State of DNS Rebinding Attacks & Singularity of Origin - DEF CON 27 Conference](https://www.youtube.com/watch?v=y9-0lICNjOQ).
[**`Singularity of Origin`**](https://github.com/nccgroup/singularity) is a tool to perform [DNS rebinding](https://en.wikipedia.org/wiki/DNS\_rebinding) attacks. It includes the necessary components to rebind the IP address of the attack server DNS name to the target machine's IP address and to serve attack payloads to exploit vulnerable software on the target machine.
* [https://wicg.github.io/private-network-access/](https://wicg.github.io/private-network-access/): Proposal to always send a pre-flight request when public servers want to access internal servers
- Do you work in a **cybersecurity company**? Do you want to see your **company advertised in HackTricks**? or do you want to have access to the **latest version of the PEASS or download HackTricks in PDF**? Check the [**SUBSCRIPTION PLANS**](https://github.com/sponsors/carlospolop)!
- **Join the** [**💬**](https://emojipedia.org/speech-balloon/) [**Discord group**](https://discord.gg/hRep4RUj7f) or the [**telegram group**](https://t.me/peass) or **follow** me on **Twitter** [**🐦**](https://github.com/carlospolop/hacktricks/tree/7af18b62b3bdc423e11444677a6a73d4043511e9/\[https:/emojipedia.org/bird/README.md)[**@carlospolopm**](https://twitter.com/carlospolopm)**.**
- **Share your hacking tricks by submitting PRs to the [hacktricks repo](https://github.com/carlospolop/hacktricks) and [hacktricks-cloud repo](https://github.com/carlospolop/hacktricks-cloud)**.