In the past, I've described how mass SQL injection worms took the Web completely by storm. Two years ago, SQL injection attacks evolved from sentient, one-off, targeted data-stealing exploits, like in the breaches at Hannaford Brothers and Heartland, to fully automated, unauthenticated mass-scale Website malware-injecting robots. The robots' mission, rather than extracting sensitive data, is implanting malware-laced iFrames into back-end databases where they are later served up to unsuspecting Web visitors as content. And they've been extremely successful.According to Websense, 70 percent of the top 100 popular Websites either have hosted malicious content or redirected unsuspecting victims to malicious sites. If a user's machine gets a virus/worm, chances are the infection originated from some Web page.
As bad as things are now, we could be due for another evolution.
Some of the first evolutions we saw in mass SQL injection worms was their ability to disguise their payload from intrusion detection systems, Web application firewalls, and other application input validation routines. Clever encoding schemes helped bypass security controls, and funneling attacks over HTTP POST would hamper exploit logging and forensics.
The good news is that today's mass SQL injection attacks are NOT scanning while authenticated: They don't register accounts, and they don't log in (username/password), so they can't see or exploit any Website functionality with those preconditions. Basically, none of the Deep Web is touched. Only the functionality wide open to the public is within scope, leaving out a ton of Websites and attack surface. But what if mass SQL injection worms evolved again? What if they found a generic way to log in?
At first, that concept sounded rather remote. I mean, talk about some serious AI. Not even Google can do this kind of Deep Web crawling capability (at least I don't think so). Consider that the worms would first need to figure out that they needed to register an account, then find and fill out the registration form with valid data (including CAPTCHA), potentially activate their accounts via email, log-in with their username/password, and then finally start attacking the Website. The more I broke down the requirements, the more feasible this sounded.
Personal experience and cursory research reveals that Web spammers already know how to fingerprint popular blogging systems, message boards, and Web mail providers. Spammers are completely capable of automatically registering accounts, bypassing CAPTCHA protection, responding to activation email, logging-in, and spamming the hell out of us. We've all seen the crappy backlinks, malware URLs, and other spammy messages. But again, these activities are fairly limited to well-known platforms.
What Web spammers don't do, to my knowledge, is how to use this capability to perform SQL injection attacks post-login. So we're left to speculate: How hard would it really be for them to make these technology pieces more generic? All the pieces of the puzzle seem possible, perhaps already built, just not yet culled together. We also don't know when, if ever, this evolution might take place.
Are we ready if it does? Some are. But the rest are going to have a really bad year.
Jeremiah Grossman is CTO and founder of WhiteHat Security. Special to Dark Reading Jeremiah Grossman, Chief of Security Strategy, SentinelOne, Professional Hacker, Black Belt in Brazilian Jiu-Jitsu, & Founder of WhiteHat Security. Jeremiah Grossman's career spans nearly 20 years. He has lived a literal lifetime in computer security to become one of the ... View Full Bio