Deep Learning-Based Intrusion Detection Framework for Securing IoT-Enabled Smart Homes

Abstract

The explosion of Internet of Things (IoT) technologies has allowed smart home devices to cooperate and offer effortless access to many benefits. With everyone more connected these days, criminals have a larger chance to attack your smart home, since it makes them more likely to get into network breaches, be targeted by malware, face denial-of-service attacks and have their data stolen. Most existing IDS systems which depend on fixed rules or known signatures, have difficulty dealing with the fast changes and limited resources typical in IoT networks. To overcome these problems, deep learning introduced an automated way for detecting intrusion attempts that greatly boosted both the accuracy and flexibility of the tools. This study gives a detailed look at recent deep learning IDS frameworks built for IoT smart homes. It sorts existing methods into three categories: those for centralized architecture, distributed architecture and those depending on both host and network deployment; the list of deep learning models included are Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Long Short-Term Memory networks (LSTM), Auto encoders and Generative Adversarial Networks (GANs). The study also investigates the following challenges in IoT: shortage of datasets, difficulty in detecting events at the right moment, a lack of tools to interpret how IoT algorithms function and the small amount of processing power that the devices have. Besides, it assesses frequently encountered datasets, the main performance measures used and the various methods of optimization employed for practical uses. This paper hopes to offer a starting point for those who want to build efficient, strong and flexible deep learning systems for detecting attacks in smart home environments.