Software Diversity Assessment Based on Input Space Decomposition
Lichun Chen, John H R May, Grant Douglas Hughes · Bristol Research (University of Bristol) · 2000
Based on the EL and LM models for coincident failures in multiversion software, a new approach to estimation of failure behaviours correlation is presented. The crucial technique is to utilize decomposition of the input space. The result is that the greater the number of common success subregions confirmed by testing, the less likely it is that the independence model is appropriate on the whole input space. Input space decomposition will benefit diversity assessment.