A Deep Transfer Learning Framework for Robust IoT Attack Detection: A Review

Abstract

Our lives have been significantly altered due to the digital revolution, and the Internet of Things (IoT) has played a significant part in this transformation. However, the fast expansion of the IoT into almost every aspect of life has resulted in various new cybersecurity dangers. As a result, detecting and preventing possible attacks on IoT networks have lately garnered significant attention from the academic and business worlds. Machine learning (ML)-based techniques, intense learning (DL), have shown considerable promise among the many different approaches to attack detection. This is because they can identify attacks at an early stage. However, for these DL algorithms to be effective, gathering substantial data from IoT devices, including label information, is necessary. On the other hand, the labeling process is often resource-intensive and time-consuming; hence, it may not be able to accommodate rapidly growing IoT threats in the real world. The introduction of DL methods to the IoT datasets is the main emphasis of this study, which also reviews the newest advancements in security measures for threat detection. This review aims to examine DL techniques and continuing breakthroughs in approaches that may be used to produce enhanced attack detection models for IoT frameworks. This is the objective of this review. When applying DL to IoT security, we address the benefits and research gaps associated with each strategy.