Users of cloud computing infrastructures should be aware that their sensitive data could be potentially leaked, a group of university researchers say.

In a new research paper (PDF), several computer scientists from the University of California at San Diego (UCSD) and the Massachusetts Institute of Technology (MIT) say they have discovered soft spots in the cloud computing concept that could leave data vulnerable to attack.

"Overall, our results indicate that there exist tangible dangers when deploying sensitive tasks to third-party compute clouds," the paper says.

In a nutshell, the researchers argue that by taking the right steps, an attacker could place a malicious virtual machine (VM) in close proximity to a target server in a shared, "cloud" environment. From there, it would be possible to launch a "cross-VM attack" using a variety of different hacking strategies, they say.

Such methods might be employed by an attacker looking to access a specific target or server, or they might be used by hackers casting a wide net, searching for any vulnerable server, the paper says.

The paper asserts that the automated methods used to place VMs in a cloud environment might be exploited to allow an attacker to intentionally create VMs and place them near a likely target server. The malicious VM then could use "side channels" to learn more about the target server, eventually collecting enough "leaked" data to allow penetration.

The good news is that there are some approaches that might mitigate the risk of leakage, the researchers say. One solution might be to "let users request placement of their VMs on machines that can only be populated by VMs from their (or other trusted) accounts," the paper says. "In exchange, the users can pay the opportunity cost of leaving some of these machines under-utilized. In an optimal assignment policy (for any particular instance type), this additional overhead should never need to exceed the cost of a single physical machine.

"Ultimately, we believe that the best solution is simply to expose the risk and placement decisions directly to users," the researchers say. "A user might insist on using physical machines populated only with their own VMs and, in exchange, bear the opportunity costs of leaving some of these machines underutilized. Regardless, we believe such an option is the only foolproof solution to this problem and thus is likely to be demanded by customers with strong privacy requirements."

Have a comment on this story? Please click "Discuss" below. If you'd like to contact Dark Reading's editors directly, send us a message.