An Approach to Selecting an Informative Feature in Software Identification
Statement of Research. A need to reduce the increasing number of system vulnerabilities caused by unauthorized software installed on computer aids necessitates development of an approach to automate the data-storage media audit. The article describes an approach to identification of informative assembly instructions. Also, the influence of a chosen feature that is used to create a unified program signature on identification result is shown. Methods. Shannon method allowing a determination of feature informativeness for a random number of object classes and not depending on the sample volume of observed features is used to calculate informativeness. Identification of elf-files was based on applying statistical chi-squared test of homogeneity. Main Findings. Quantitative characteristics of informativeness for 118 assembly instructions have been obtained. The analysis of experimental results for executable files identification with 10 different features used to create program signatures compared by means of the chi-squared test of homogeneity at significance levels p = 0.05 and p = 0.01 has been carried out. Practical Relevance. The importance of using a particular feature in program signature creation has been discovered, as well as the capability of considering several executable file signatures together to provide a summative assessment on their belonging to a certain program.