Black Hat 2014: How to Hack the Cloud to Mine Crypto Currency

Cyber security researchers devise a hack to demonstrate the need for improved anti-botnet security measures

4 min read
Black Hat 2014: How to Hack the Cloud to Mine Crypto Currency
Illustration: iStockphoto

Using a combination of faked e-mail addresses and free introductory trial offers for cloud computing, a pair of security researchers have devised a shady crypto currency mining scheme that they say could theoretically net hundreds of dollars a day in free money using only guile and some clever scripting.

The duo, who are presenting their findings at this week’s Black Hat 2014 cyber security conference in Las Vegas, shut down their proof-of-concept scheme before it could yield any more than a token amount of Litecoins (an alternative to Bitcoin). The monetary value of both virtual currencies is based on enforced scarcity that comes from the difficulty of running processor-intensive algorithms.

Rob Ragan, senior security associate at the consulting firm Bishop Fox in Phoenix, Ariz., says the idea for the hack came to him and his collaborator Oscar Salazar when they were hired to test the security around an online sweepstakes.

“We figured if we could get 100,000 e-mails entered into the sweepstakes, we could have a really good chance of winning,” he says. “So we generated a script that would allow us to generate unique e-mail addresses and then automatically click the confirmation link.”

Once Ragan and Salazar had finished securing the sweepstakes against automated attacks, they were still left with all those e-mail addresses.

“We realized that … for about two-thirds of cloud service providers, their free trials only required a user to confirm an e-mail address,” he says. So the duo discovered they effectively had the keys to many thousands of separate free trial offers of cloud service providers’ networked storage and computing.

In other words, they had access to many introductory accounts at sites like Google’s Cloud Platform, Joyent, CloudBees, iKnode, CloudFoundry, CloudControl, ElasticBox and Microsoft Windows Azure.

Some of these sites, each offering their own enticement of free storage and free computing as a limited introductory offer, could be spoofed, the researchers discovered. Troves of unique e-mail addresses, using a non-discoverable automated process they developed, could be readily made on the fly and then used to get free storage and processor time.

A spoof e-mail address of course has two components, Ragan says, the local part (the stuff to the left of the “@“ sign) and the domain (to the right). To appear like a random stream of e-mail addresses signing up for any given service, Ragan says they scraped real local addresses from legit e-mail address dumps on sites like Pirate Bay. The domain side they set up using “FreeDNS” servers that attach e-mail addresses to existing domains, a service that can be exploited for domains that have poor security measures in place.

So, say there’s an address dump file on the Internet containing the legit e-mail addresses “CatLover290 at gmail” and “CarGuy909 at Yahoo.” Ragan and Salazar’s algorithm would attach “CatLover290” and “CarGuy909” to one of thousands of spoof URLs they’d set up through the FreeDNS sites. The original e-mail accounts would then be unaffected. But the resulting portmanteau e-mail addresses would appear to be coming from a random stream of humans on the Internet.

Thus, Ragan says, not even a human observer watching the e-mails registering for free cloud computing accounts—none appearing to be produced by a simple algorithm or automated process—would detect anything overtly suspicious. And to further throw off the scent of suspicious activity, they used Internet anonymizing software like TOR and virtual private networks to spoof where the trial account requests were coming from. (Ragan says that generating real-seeming names using name-randomizing algorithms would probably be good enough.)

“A lot of the e-mail confirmation and authentication features rely on the old concept that one person has one e-mail address—and that is simply not the case anymore,” Ragan says. “We’ve developed a platform that would allow anyone to have 30,000 e-mail addresses.”

So they signed up for hundreds of free cloud service trial accounts and, in the process, strung together a free, ersatz virtual supercomputer.

“We demonstrated that we could generate a high amount of crypto hashes for a high return on Litecoin mining, using these servers that didn’t belong to us,” Ragan says. “We didn’t have an electricity bill, and we were basically able to generate money for free out of thin air.”

Ragan says at their scheme’s peak, they had 1000 accounts that were each generating 25 cents per day: $250 of free Litecoin. He says they shut the system down before it generated any real monetary value or made any noticeable performance dent in the cloud service systems.

And Ragan stressed that the devious schemes he and Salazar developed are being disclosed in order to raise awareness of problems in security measures that real criminal elements around the world can and probably already are taking advantage of.

“Not planning for and anticipating automated attacks is one of the biggest downfalls a lot of online services are currently experiencing,” Ragan says.

One measure Ragan says he and Salazar wanted to see that would combat their scheme’s spoofing of cloud service providers was the introduction of random anti-automation controls. Captchas, credit card verification, and phone verification can all be spoofed, he says, if they’re at predictable places in the cloud service signup and setup process.

“Some services don’t want to add a Captcha, because it annoys users,” Ragan says. “But…there are compromises that can be [employed], like once an abnormal behavior is detected from a user account, they then prompt for a Captcha. Rather than prompting every user for a Captcha every time, they can find that balance. There’s always a balance to be made between security and usability.”

Ragan says that’s what he and Salazar want the takeaway from their talk to be: that a lot more consideration is given to how to better implement anti-automation controls and features.

The Conversation (0)