The Internet of Things (IoT) connects billions of smart gadgets so that they may communicate with one another without the need for human intervention. With an expected 50 billion devices by the end of 2020, it is one of the fastest-growing industries in computer history. On the one hand, IoT technologies are critical in increasing a variety of real-world smart applications that can help people live better lives. The cross-cutting nature of IoT systems, on the other hand, has presented new security concerns due to the diverse components involved in their deployment. For IoT devices and their inherent weaknesses, security techniques such as encryption, authentication, permissions, network monitoring, \& application security are ineffective. To properly protect the IoT ecosystem, existing security solutions need to be strengthened. Machine learning and deep learning (ML/DL) have come a long way in recent years, and machine intelligence has gone from being a laboratory curiosity to being used in a variety of significant applications. The ability to intelligently monitor IoT devices is an important defense against new or negligible assaults. ML/DL are effective data exploration techniques for learning about 'normal' and 'bad' behavior in IoT devices and systems. Following a comprehensive literature analysis on Machine Learning methods as well as the importance of IoT security within the framework of different sorts of potential attacks, multiple DL algorithms have been evaluated in terms of detecting attacks as well as anomaly detection in this work. We propose a taxonomy of authorization and authentication systems in the Internet of Things based on the review, with a focus on DL-based schemes. The authentication security threats and problems for IoT are thoroughly examined using the taxonomy supplied. This article provides an overview of projects that involve the use of deep learning to efficiently and automatically provide IoT applications.