A Semi-Markov Survivability Evaluation Model for Intrusion Tolerant Database Systems

Abstract

Survivability modeling and evaluation have gained increasing importance. Most existing models assume that the distributions for transitions between states are exponential. However, this assumption does not hold in many real cases. To address this problem, we propose a novel semi-Markov survivability evaluation model, which allows the transitions between states to follow nonexponential distributions. Novel quantitative measures are also proposed to characterize the capability of a resilient system in surviving intrusions. Model validation, which is possibly the most important step in the life cycle of model development, is largely overlooked in previous research. In this paper, a real intrusion tolerant database system ITDB is implemented to validate the proposed state-space models. Empirical experiments show that the semi-Markov model predicts the system behaviors with high accuracy. Furthermore, in this paper we evaluate the impact of intrinsic system deficiencies and attack behaviors on the survivability of intrusion tolerant database systems.