By 2023, the number of IoT devices is expected to reach 3.5 billion, taking data production to over 79.4 zettabytes by 2025. This explosion in connectivity multiplies security challenges, particularly in smart homes entirely dependent on IoT. Moreover, the inherent gaps in IoT infrastructure make it even more difficult to detect cyber threats, especially distributed denial-of-service (DDoS) attacks, which often evade traditional security protocols such as intrusion detection systems (IDS). For this reason, we aim in this study to provide a comprehensive review to evaluate the effectiveness of current models based on machine learning (ML) and deep learning (DL) approaches for detecting cyber-attacks in real time. Our research seeks to assess the strengths and limitations of current approaches while identifying key gaps in the existing literature. We also plan to explore datasets used in the IoT environment that could enhance detection accuracy by offering relevant data for model training. By addressing these challenges, we aim to propose promising research directions that could lead to more effective detection systems and improve security in the IoT landscape.