신체규모 네트워크에서의 이상현상 탐지

Abstract

As the average human life expectancy increases, more people and governments are interested in healthcare and ICT convergence in a means to improve their life and reduce costs. Thus, e-healthcare is regarded as one of the most promising research areas. To provide e-healthcare service, Wireless Body Area Networks (WBAN) is a common way to remotely and ubiquitously monitor patient’s health states and react to symptoms of a disease. WBAN deals with sensitive biological information of a patient and anomaly can lead to application malfunction. Therefore, automatic anomaly detection in WBAN is highly important to promote such systems.This thesis proposes a two-level anomaly detection method in WBAN which considers not only biological information but also network information of WBAN. In the first step, the proposed method detects outliers by using boundary of normal value range and time slot. In the second step, the method detects anomaly by using data correlation between biological parameters and network information. This thesis also describes the test environment, WBAN, which is implemented to collect real data of biological and network information.Finally, we evaluate the proposed anomaly detection method with the recorded data and sample data by showing detection result at each step. The test results show the feasibility of our proposed model. Recorded data sets of both physiological and network traffic information are made available on the Internet.