Graph-based technique for survivability assessment and optimization of IoT applications

The Internet of Things (IoT) has evolved from theoretical research to market deployment. The IoT will enable a plethora of new applications in various areas of human activity and will provide tremendous opportunities for societies around the world. However, implementation is hampered by fears that the societal costs of the IoT outweigh its benefits. Failure of, or hacking into, IoT applications can, for example, disable home security systems, destroy crops, and destabilize hospitals. To unlock the IoT's potential, it needs to provide application survivability. For this purpose, it needs a tradeoff between IoT resources and system survivability. In previous works, there has been a lack of quantitative methods considering this problem, which combines specificity of network topology, intrusion details, and properties of intrusion detection/prevention system. In this work, we offer a corresponding approach that combines graph theory and stochastic process-based models. The network topology is described as a probabilistic graph. To address the properties of intrusions and defense mechanisms, we use basic survivability models, that generate the probabilities for graph elements. Therefore, the criterion of system survivability is a function defined on the created graph. An approach for deduction and computation of this survivability metric is discussed. Survivability optimization problems are formulated. In some important practical cases, closed-form solutions are offered.