Publications

Abstract : Anomaly detection in smart homes is paramount in the prevailing information age as smart devices remain susceptible to sophisticated cyber-attacks. Hackers exploit vulnerabilities such as weak passwords and insecure, unencrypted data transfer to launch Distributed Denial of Service (DDoS) attacks. Sensible deployment of conventional security measures is jeopardized by the heterogeneity and resource constraints of smart devices. This article presents a novel approach that leverages the power consumption of Internet of Things (IoT) devices to detect anomalous behavior in smart home environments. We prototype a smart camera using Raspberry Pi and gather power traces for normal activity. Furthermore, we model DDoS attacks on the experimental setup and generate attack traces of power consumption. Besides, we compare the performances of several machine learning models for accurate prediction of the presence of anomalies. A deep feed-forward neural network model achieves an accuracy of 99.2% compared to other models. Empirical evaluations of the proposed concept affirm that power consumption is a promising parameter in detecting anomalies in smart homes. The proposed method is suitable for smart homes as it does not impose additional overhead on resource-constrained IoT devices.