The Internet of Things (IoT) is a network that is ever-growing by connecting devices that can interact and perform with minimum human involvement. With the increasing use of IoT across various applications, such as smart cities, healthcare, industries, and transportation, the security of IoT devices has become very critical and essential. Traditional security techniques and frameworks are often incapable due to the complex, heterogeneous, and distributed nature of IoT ecosystems. The increasing volume and complexity of cyber threats, including denial-of-service (DoS) attacks, intrusion attempts, and unauthorized access, necessitate the adoption of intelligent, and adaptive security measures. This research study uses several ML-based classifier approaches on the RT-IOT 2022 dataset (publicly available), encompassing both legitimate and malicious IoT data. The dataset is pre-processed to ensure a balanced class representation. Our used models were further examined after removing highly correlated features and minimizing overfitting. Our study shows that the Custom Transformer Encoder has achieved the best accuracy (93%) and showcased robust performance.