In recent years, IoT (Internet of Things) devices and their attendant solutions have become more prevalent, particularly in manufacturing, supply chain management, healthcare, transportation, and numerous other "smart" settings. Thanks to IP (Internet Protocol) addresses designated to IoT devices, cyber-physical systems are now able to communicate with one another automatically. Inadequate protection on these ultimate platforms has resulted in attacks such as denial-of-service, Botnets, identity theft, and information theft. Mirai, Torii, Emotet, Dridex, Trickbot, Gluteba, and QBots are examples of Internet of Things-related threats. With the aid of AI, we can at last create cyber-physical systems that are both trustworthy and secure. Machine learning and deep learning approaches combat cyberattacks by identifying and blocking Botnets. This paper investigates Botnet attacks, which are prevalent in IoT devices due to a lack of security requirements during manufacturing or a lack of user security awareness. Anomaly detection is a potential weapon in the hands of machine learning or Deep Learning for identifying and preventing cyberattacks on Internet of Things devices. In this study, we present a Botnet detection system capable of detecting an attack on live traffic. Next, we employ the Aposemat IoT-23 dataset to evaluate and compare a variety of machine learning as well as deep learning techniques for detecting Botnets based on their conventional characteristics. Gated Recurrent Unit (GRU) is a deep learning model with a detection accuracy of 99.87% for identifying Botnets. Thirdly, we use the Wireshark program to investigate the Aposemat IoT-23 dataset's raw packet captured (pcap) files for attacks. Then, we employ GRU, a deep learning model, to identify malware infections with a 99.89% success rate and reduced time complexities.