DTW-KNN Implementation for Touch-based Authentication System

Abstract

Touch dynamic is a new biological recognition method based on one's touch pattern. However, less works have been done to explore the use of motion sensors embedded in smart phone for touch dynamic to enhance the password authentication. In this paper, we propose a touch-based authentication system to enhance the protection level of traditional authentication mechanism by adding touch pattern into the password mechanism. The proposed system leverages only three motion sensors (accelerometer, magnetometer and gyroscope) to monitor users' touch behavior. After performing pre-processing and a series of in-depth analysis on these sensor data, some touch features are extracted for users to characterize their unique touch patterns. Then the K-Nearest Neighbor (KNN) with Dynamic Time Warping (DTW) distance is employed to train the model for authentication. The experimental results show that the proposed system can achieve a lowest average false acceptance rate (FAR) of 4.85% and false rejection rate (FRR) of 4.98% with 1200+ data.