Improving smart home security; integrating behaviour prediction into smart home

Abstract

The paper highlights various security issues in existing smart home technology and its inhabitant behaviour prediction techniques and proposes a novel behaviour prediction algorithm to improve home security. The algorithm proposed in this work identifies legitimate user behaviour and distinguishes it from attack behaviour. The work also identifies the parameters necessary to predict user behaviour during the seven week learning period. The paper identified three factors namely time parameter, light parameter, user's key placement behaviour to successfully predict user behaviour. The algorithm learned normal and suspicious user behaviours during the seven week training period and naive Bayesian network was designed based on the knowledge. The newly developed security algorithm was implemented in a studio apartment for a period of two weeks which was accessed 24 times generating two warnings and one alarm.