There's a hack that's beginning to be active that targets pages named "index.*". Actually it sounds rather like an old hack that is resurfacing. Since many ColdFusion sites use this convention for the home page this attack tends to hit quite a few ColdFusion sites that are vulnerable. The attack appends a script like this one to the bottom of each "index.*" page:
As you can see the script itself is pretty simple. It writes out an invisible Iframe to the bottom of the page. The target of the Iframe attempts to download a trojan or malware to the users machine. This attack is insidious and I have yet to discover the origin. But I do know a few things about it - and how to prevent it from continuing. One important thing to note, if you have this problem and Google indexes your sites and sees these pages they will flag your site. Browsers like Firefox use the Google service to throw up a big "malware" warning.
The following article details the attack and the notes I've gathered about it. Some day soon I hope to post a more definitive who, what, when and why post about it. To gather the following notes I'm indebted to the folks on the CF-Talk List (this thread), Nathan, Nick, Jason, Scott, Don and probably a few others I am forgetting. I can't give away too much info here - but please accept my thanks.
When we first started this process I scoured the internet for a common virus or malware that was doing the dirty deed. I assumed I would be able to scan the server with something and fix a vulnerability. None of our scanning found anything that was a smoking gun. We had a "find and replace" script (in Perl) that went through all the files and removed the malicious code each day, but it kept coming back and no malware was found on the server. Here are the clues we gathered from our specific problem.
We also gathered some clues from a few folks who had combated similar attacks - From Don we gathered the following clues from the attack that he combated (which also left scripts embedded on index.* pages):
Finally, 2 excellent blog post titled Hidden iframe injection attacks by Niyaz PK and a Swiss security blog post outlines what I think is the most likely origin of this attack as I'll explain. Niyak's take was that the malware does not reside on the server at all. Instead it resides on client computers who connect to the database or (in particular) FTP permissions. In Niyaz' view the malware piggybacks onto FTP clients, steals credentials, and then launches regular attacks that buzz through the folders available to it altering index.* files.
Now in regards to the attack I was troubleshooting I can see evidence of both of these vectors. We had additional users and file permissions changed and we also had the iframe injection issues. Am I dealing with one attack or two? My belief at this point is that we are dealing with 2 separate complementary attacks (and I'll certainly change my mind if someone gives me more info). The attack related to extra users and file permissions seems to me to be a root kit or a "full control" attack. Past scan logs indicate the possible presence of the graybird trojan at some point (which was removed). This trojan can give full access to a malicious user. So the file permissions and users could have come from a root access type attack.
Ok, so if someone has full access they can change your files too right? Why do I think it is 2 attacks? For one thing, shutting down FTP on the server immediately stopped all further file changes. And there is one more clue. Like a lot of servers this one has a sort of convention for where the html files go. It is like (not exactly so don't think I'm giving anything away here).
There is one thing that I can't explain however. If the attack was a client based agent using existing FTP permissions to access files, such activity would look like legitimate FTP activity on the server. That means the server would log such activity. But the FTP log files were actually truncated - possibly to remove any signs of the attack. The FTP user did not have permissions to the logs directory. So if I am correct and the attack is a client FTP type attack, how did such an attack remain unlogged? That certainly gives my theory of 2 separate attacks a black eye. It seems more likely that this is indeed a multi-tiered attack as Don had surmised, and altering the files was just the end game. Still, if your goal is to deface web based files and you already have legitimate FTP credentials, why bother changing file permissions and adding users?
To summarize, lots of bad things are happening, but disabling FTP on the server put a stop to the file changes. So the first step here (if you are victim of this attack) is to disable FTP and force all FTP users to scan their computers for malware. If you are on a shared host with a control panel interface (where users can add pretty much any password) and lots of websites you are particularly vulnerable to this sort of attack. It's one of the downsides to commodity hosting that you are sometimes at the mercy of the weaknesses of others.
The really bad news is that as long as FTP is necessary there is little that can be done to prevent an attack that steals credentials on a client computer. Sure, you can make sure your servers are scanned and up to date, but how do you monitor every client?
In closing dear readers, I'm aware that many of you will likely be tempted to pipe in and tell me how bad FTP is and how it should never be used on a production server. Please refrain from doing that. FTP is a regrettable de facto standard in the shared hosting world. Folks victimized by this attack don't need kicking when they are down :) On the other hand, if you have additional clues or thoughts that would help us identify the actual agent or culprit involved by all means share away!
Nathan Bruer came up with this additional and fascinating information. He Writes: