Abnormal Events Detection for Infrastructure Security using Key Metrics
Abstract
This paper presents a detection process which utilizes various sensors (camera, card readers, movement detector) for detecting automatically abnormal events. The detection process strengthens current security systems to identify attackers in the context of building and office. Key metrics are proposed to describe people’s behavior in critical zones of the building. They are built using measures from the sensors, which provide information about the person, the position, and the instant. These metrics are used to classify abnormal behaviors from regular ones, based on a statistical classifier. This technique is tested on both simulated data and real data, in which an attacking scenario was prepared by security experts. Results show that abnormal events from the scenario have been successfully detected. The experiments demonstrate that the proposed key metrics are relevant and the proposed detection scheme is appropriate for infrastructure surveillance.