Original URL: http://www.theregister.co.uk/2012/09/19/zero_day_ie/
Inside the guts of a fiendish Internet Explorer 0-day attack
Bite down, matey, surgeon Pott's got his chopper out
Sysadmin blog The latest use-after-free vulnerability in Microsoft's Internet Explorer is being actively exploited in the wild.
While this comes as a surprise to nobody, we all get to live in fear for a while until the software giant manages to get a patch out the door. In the interests of attempting to overcome a particularly nasty bout of insomnia, I decided to find out how exactly how the exploit worked.
The selection event handler script then writes a character to the web page. This, for reasons best known to Microsoft, causes the browser to delete the original command object and free its memory.
However, the "select all" command has not finished executing - remember, it was interrupted. The selection event handler is still running, and rewrites 100 HTML image tags with extra data. This is a textbook attempt to force the browser to reuse the memory that was previously holding the "select all" command object and fill it instead with the data in the updated tags. With a lot of luck and testing on the hackers' part, it's possible to carefully overwrite the memory that held the command object with new data. The event handler then exits.
The browser returns to handling the "select all" command, and it tries to use what it thought was its object - instead the object's memory had been freed and now contains data set by the hackers. The object holds the location of an internal browser function that needs to be called, and this value can be changed to point to malicious code - thus hijacking the flow of execution and putting hackers on the road to compromising the machine. It is a classic use-after-free bug.
The malicious code runs with the same privileges as the Internet Explorer process, and can be placed in memory prior to the attack. Once the processor is executing the hackers' code, it can be instructed to download and run a second executable, such as a rootkit, worm or whatever.
Inside the child HTML script is where the magic happens. There are two functions; one "selects all" and the other prints a single character "L" and changes the "src" value of the image object inside the parent HTML's array to the special string that mimics the command object in memory. This is the meat of the exploit. Note that the Metasploit example doesn't have to do this 100 times.
To trigger the whole process, a couple of simple additions to the body tag are used: on loading the malicious page, the child HTML calls the function that selects all the text; this triggers the function assigned to the "on select" handler, the function that writes the changes.
Throwing a fuzzer at IE and looking at memory dumps, or calling functions in a way that Microsoft never intended, until you identify specific code that allow you to do things like this is intensive work. No matter how well-written your application is, there are always going to be bugs like this that someone can take advantage of.
What should give systems administrators the heebie jeebies on this one is just how easy it is to use, and to extend past the basic exploit to do really creative things. ®