Security games for virtual machine allocation in cloud computing

While cloud computing provides many advantages in accessibility, scalability and cost efficiency, it also introduces a number of new security risks. This paper concentrates on the co-resident attack, where malicious users aim to co-locate their virtual machines (VMs) with target VMs on the same physical server, and then exploit side channels to extract private information from the victim. Most of the previous work has discussed how to eliminate or mitigate the threat of side channels. However, the presented solutions are impractical for the current commercial cloud platforms. We approach the problem from a different perspective, and study how to minimise the attacker's possibility of colocating their VMs with the targets, while maintaining a satisfactory workload balance and low power consumption for the system. Specifically, we introduce a security game model to compare different VM allocation policies. Our analysis shows that rather than deploying one single policy, the cloud provider decreases the attacker's possibility of achieving co-location by having a policy pool, where each policy is selected with a certain probability. Our solution does not require any changes to the underlying infrastructure. Hence, it can be easily implemented in existing cloud computing platforms