A feature selection technique based on rough set and improvised PSO algorithm (PSORS-FS) for permission based detection of Android malwares

Abstract

The set of permissions required by any Android app during installation time is considered as the feature set which are used in permission based detection of Android malwares. Those high dimensional feature set should be reduced to minimize computational overhead by choosing an optimal sub set of features. In recent times, selection of meaningful attributes is an inevitable step for mining of large dimensional data and the application of heuristic feature selection algorithms are the main research directions in this field. “Quality of classification” measure is inspired by rough set theory and can be combined with bio inspired heuristic search techniques (Particle swarm optimization, Genetic Algorithm etc.) in selecting optimal or near optimal subsets of features. In this work, a feature selection technique based on rough set and improvised particle swarm optimization (PSO) algorithm is proposed for selection of features in the permission based detection of Android malwares. The main contribution of this work is to recommend a new random key encoding method which is used in the proposed work (PSORS-FS) to convert classical PSO algorithm in discrete domain. It also reduces the issues related to maximum velocity of particles as well as sigmoid function which is related with binary PSO. PSORS-FS ensures diversity in the search process and it also reduces the tendency of premature convergence. Datasets of UCI, KEEL machine learning repository and two Android permission datasets have been used to evaluate the performance of the proposed method. Better classification performance has been yielded by proposed method over conventional filters and wrapper methods for most of the machine learning classifiers.