In addition to the most recent advances in machine learning-based IoT security techniques, Machine learning algorithms can learn from data and adapt to new security risks, allowing IoT systems to defend against unanticipated cyber-attacks. ML-based security solutions are flexible and dynamically updateable. This is critical in dealing with emerging IoT based security threats. In IoT security, there are two types of ML: supervised learning and unsupervised learning. To detect anomalies and categorize security risks, the supervised learning technique employs various algorithms such as decision trees, support vector machines, and neural networks. On the other hand, the unsupervised learning approach includes clustering techniques, such as k-means clustering and hierarchical clustering to detect anomalies and identify unknown security threats. In addition, reinforcement learning (RL) has been employed to enhance the security of IoT systems by training the system to identify and respond to cyber-attacks in real time. Additionally, these solutions can provide real-time security monitoring and incident response, allowing for quicker and more effective mitigation of cyber-physical security risks in IoT systems. Henceforth ML-based security solutions have shown promise in addressing the cyber-physical security challenges that arise in IoT systems. These solutions can learn from data, adapt to new and evolving threats, and provide real-time security monitoring and incident response. As the number of connected devices in IoT systems continues to grow, it is essential to adopt modern and innovative security solutions to ensure the safety and privacy of IoT users.