Conditional Probability Voting Algorithm Based on Heterogeneity of Mimic Defense System
Shuai Wei, Huihua Zhang, Wenjian Zhang, Hong Bo Yu · IEEE Access · 2020
In recent years network attacks have been increasing rapidly, and it is difficult to defend against these attacks, especially attacks at unknown vulnerabilities or backdoors. As a novel method, Mimic defense architecture has been proposed to solve these cyberspace security problems by using heterogeneous redundant variants to perform the same task. How to choose appropriate variants and voting algorithm according to heterogeneities of these variants become the key issue of designing mimic defense architecture. Most of current researches are based on the 2-level similarity of variants, but the results are not accurate enough. This article presents an attack model based on mimic defense architecture, abstracts binary division vector and relevant indexes to describe the heterogeneity of these variants, and innovatively proposes conditional probability voting algorithm, which is different from classic majority voting algorithm. This article also analyzes the system failure probability and scalability of these voting algorithms, experiment results show that conditional probability voting algorithm is the best, both in system failure probability and scalability.