Securing Smart Homes Using Hybrid Stacking Ensemble Deep Learning-Enabled Framework

The added deployment of Internet-of-Things (IoT) devices in various environments has led to people increasingly adding this technology to their homes to make them more comfortable and have a more convenient living environment. Despite the promise and the growth potential of the smart home market, one main concern is the security and privacy of smart home devices. This is because smart home devices equipped with IoT technology are prone to various network-based attacks. With smart home devices having limited computation and memory capabilities, many IoT manufacturers often overlook the security aspect by focusing more on the functionality. To address this issue, the concepts of machine learning (ML) and deep learning (DL) have emerged as viable and promising solutions. To that end, this paper proposes a hybrid stacking ensemble DL-enabled framework that combines both classical ML models with sophisticated DL models. The goal is to make use of the simplicity and interpretability of the classical ML models with the superior time-dependent patterns that DL models can detect. The proposed framework's performance is evaluated using multiple datasets representing different smart home devices. Experimental results show that the proposed framework achieved better detection accuracy, precision, recall, and F-score.

I. Introduction

The added deployment of Internet-of-Things (IoT) devices in various environments has led to the emergence of the concept of “Smart Communities”. A pillar of these smart communities is the “Smart Homes” paradigm. People are increasingly adding IoT technology to their homes to make them more comfortable and have a more convenient living environment [1], [2]. This added interest in adopting IoT devices in smart homes is supported by the recent estimates conducted by Zion Market research stating that the global smart home market will reach approximately $53.45 billion by 2022 [3].