Side Channel Attacks in the Cloud

I saw this paper (by Yinqian Zhang, Ari Juels, Michael K. Reiter and Thomas Ristenpart in ACM CCS 2012) earlier this year, but thought it was a very specific threat model. In a  one line summary - it is possible to recover private keys when they are being used within a virtual machine, through observations of the activity of the virtual machine from the host machine. It is a very complex attack, and requires at least host access for these observations, so my initial thoughts were that this attack could only be carried out by extremely skilled admins of a cloud hosting provider; but the complexity would probably mean that there was no realistic threat in that regard.

With the NSA revelations of the past few months, this is an interesting approach that could be taken by an agency (such as the NSA) to recover private keys from cloud providers, without getting actual access to the servers themselves. Given that PRISM does provide such access to hosts, it is not inconcievable that systems that are hosted on public cloud services such as Amazon's EC2 could be monitored. However, given the description of events relating to Lavabit, it is likely that this type of attack hasn't been operationalised yet - but remains interesting on what could be achieved.