Virtual machines (VMs) in the form of cloud computing are now mainstream for the delivery of information technology (IT) services. However, the environment in which these VMs run is complex and prone to malicious attack, potentially impacting service reliability.

Current support models for cloud environments are typically simple extensions of older models for supporting physical machines with questionable applicability to virtual environments. Joseph and Mukesh propose an autonomic model to predict VM security problems and effectively self-recover and self-heal a VM, thus providing an effective tool for reliable cloud services. The authors begin by introducing the issues faced by cloud computing, and then describe the decision-making model they propose. The discussion covers predictive modeling processes, methods for training data, machine learning algorithms, a comparison of snapshot techniques, and the components of autonomic secure predictive models.

The authors describe the proposed architecture--basically, a VM monitor reports to an autonomic manager that analyzes behavior against a knowledge base. A decision maker then uses predictor algorithms to determine whether a VM is under attack and if so, recovers using a recent snapshot. The authors explain machine learning training and testing, and discuss their experimental analysis and results. Thorough references are cited.

While not detracting from the paper’s interest and current relevance, of minor annoyance were the many typographical errors throughout the paper that could have benefited from more thorough editorial oversight. A novel and interesting proposal for seamlessly delivering continuous, reliable VM cloud services to clients.