I’ve written up a short UrlScan bypass article, made some edits, but overall this bypass has helped me identify and exploit SQL and XSS or other attacks that are URL based.
Identified after coming up with null for help with bypassing a WAF identified as UrlScan. After identifying that a web application was filtering and essentially dropping most attacks and their associated payloads a delve into how to bypass this was constructed. This is as simple as bypasses can possibly get, but at the same time is unique enough to warrant writing about.
We all understand that WAFs are in place to identify and block malicious requests before the reach the application, so in effect I need to figure out exactly what makes it tick or how to make it tick for us. To that regard the development of this came after exhaustive research into UrlScan and trying to see if anyone had run across this in the professional or unethical realm. Well, guess you can figure out how well that went.
Anyways… I’ve identified the IIS module “UrlScan” running on a IIS6 machine (Note this can be IIS 7.5, 6, 5, etc.), I’ve identified that the application is filtering certain characters, but I’m stuck because whatever malicious requests I send get dropped or filtered by UrlScan anyways.
So after perusing developer forums, Microsoft technical documentation, and various SQLmap documentation and tamper methods I had learned that appending %00 (Null Byte) can be performed by the tamper script “appendnullbyte”. This however refused to work, and the UrlScan module picked it up right away. So what was I supposed to do? Well apparently UrlScan doesn’t know how to handle or what to do when the prefix is the nullbyte, and as far as I’m aware there are no SQLmap tamper scripts that would perform this bypass.
So after formatting the sqlmap command with the real value of a parameter I know that this page exists and has dynamic content depending on the “users” integer value
I quickly learn that the appendnullbyte tamper script only modifys the payload like this (note this is just generic payload)
http://somesite.com/blog?users=3389%00′ WAITFOR DELAY ‘0:0:10′–
So after analyzing the responses from the application it seemed like it wouldn’t take it at all… However, not all is lost and after performing more research into the matter the solution became apparent after some random dev was complaining about UrlScan filter rule of basically crashing their application. So what I can extrapolate from that is that is a null value (obvious), but more so than that I can deduce that is not only not interpreted by UrlScan, but its completely overlooked because of the null value where it expects something.
Imagine if you will, you are an application looking for a value of anything greater than 0, but then you encounter 0. Would you simply stop interpreting it because to you there is nothing there? Well if you said yes, your in the same boat as UrlScan’s logic apparently.
So I’ve come to the conclusion that UrlScan expects some value to interpret or inspect whether it be in the Url, Url parameter, or POST body. I’ve also come to understand that if the value is %00 or Null. UrlScan simply ignores it on the basis that there is nothing to inspect, thus giving us the path towards carnage.
So here I am, at the end of the road… Will it work or not?
http://somesite.com/blog?users=3389%00’ WAITFOR DELAY ‘0:0:10′–
IT WORKS! This little null value gave me the ability to perform SQL injection where SQL had failed time and time before.
This also allowed XSS to any arbitrary parameter I wanted…
REMEMBER NULL should be PERCENT ZERO ZERO after the valid data, but before the actual payload.
In the end I’ve learned the following…
1. UrlScan’s logic is flawed in the manner of interpreting null values – Expects 1, but gets 0 and does not continue inspection
2. allows us to bypass UrlScan’s logic to perform XSS and SQL injection where it would normally fail
3. filtering within UrlScan breaks applications for whatever reason
4. SQLmap does not have a tamper script with which to bypass UrlScan, only has the ability to append to end of payload where instead requires it be prepended to the payload