Continuous Authentication on a Smartwatch

The purpose of this work is to leverage two types of sensors, motion and optical, to create a continuous authentication system for smart devices such as smartwatches. The proposed solution is based on an Android application that uses the accelerometer and gyroscope to measure movements and to classify them in normal and session-endangering classes. If suspicious movements are identified, then the app enacts a second decision level and activates the heart or body detection sensor to check if the watch is actually still on the user’s wrist. The two-level architecture tries to optimize energy consumption. To validate our system, various measurements were carried out with the aim of mapping the typical gestures of users who wear a smartwatch. The goal is therefore to be able to recognize certain movements, limit checks involving the optical sensors that are extremely energy hungry, and, thus, achieve a better battery recharge cycle.