A hybrid GA-PSO fuzzy system for user identification on smart phones

Abstract

The major contribution of this paper is a hybrid GA-PSO fuzzy user identification system, UGuard, for smart phones. Our system gets 3 phone usage features as input to identify a user or an imposter. We show that these phone usage features for different users are diffused; therefore, we justify the need of a front end fuzzy classifier for them. We further show that the fuzzy classifier must be optimized using a back end online dynamic optimizer. The dynamic optimizer is a hybrid of Particle Swarm Optimizer (PSO) and Genetic Algorithm (GA). We have collected phone usage data of 10 real users having Symbian smart phones for 8 days. We evaluate our UGuard system on this dataset. The results of our experiments show that UGuard provides on the average an error rate of 2% or less. We also compared our system with four classical classifiers -- Na¨1ve Bayes, Back Propagation Neural Networks, J48 Decision Tree, and Fuzzy System -- and three evolutionary schemes -- fuzzy system optimized by ACO, PSO, and GA. To the best of our knowledge, the current work is the first system that has achieved such a small error rate. Moreover, the system is simple and efficient; therefore, it can be deployed on real world smart phones.