An Approach of Trustworthiness Evaluation of Software Behavior Based on Multidimensional Fuzzy Attributes

In order to increase the accuracy of trustworthiness evaluation of software, the paper improves the traditional construction process of the expected behavior trace of software and proposes an approach of trustworthiness evaluation of software behavior based on multidimensional fuzzy attributes. First, training samples of the same monitoring point are clustered based on multidimensional fuzzy attributes to construct a more accurate expected behavior trace of software. Second, an improved weight distribution method of multidimensional fuzzy attributes is presented based on correlation coefficient and standard deviation integrated approach (CCSD) for weight distribution of attributes in multiple attribute decision making. The improved weight distribution method is suitable for one-class samples from monitoring point and it considers both the dispersion of fuzzy attribute’s value and the influence among these fuzzy attributes. Finally, experiments and analyses show that: 1 the expected behavior trace of software constructed by training after clustering is more accurate than without clustering; and 2 our improved weight distribution method of multidimensional fuzzy attributes has better effect of trustworthiness evaluation than CCSD and other methods of weight distribution for one-class samples.