A Review of Mitigation of Attacks in IoT using Deep Learning Models

Abstract

In current era, the proliferation of IoT devices has transformed our daily life to a new level and made our life easier. IoT devices have interconnected with each other for communing and sharing information to gateways or Access Points (APs) for further processing of data. However, this provides growth to cybersecurity and zero-day attacks in IoT networks. In this paper, we have reviewed the deep learning models and datasets which are used to detect malicious data in an IoT ecosystem. We have observed that the combination of Long Short-Term Memory (LSTM) and Convolution Neural Network (CNN), LSTM, and stacked auto-encoders have better accuracy and precision for detecting malicious packets in the IoT environment. Moreover, a detailed theoretical analysis of deep learning models and datasets is also performed. This review provides a pathway for the new researchers to conduct research in IoT security and privacy issues by making these findings as references.