In the era of information technology, Botnets have become the most persistent cyber threat, capable of launching large-scale attacks like Distributed Denial of Service. stealing sensitive information and disturbing online services. Botnets have evolved from simple networks to complicated distributed networks including IoT devices, making them pervasive, harder to track, and destroy. Machine learning and Deep learning based models are emerging to detect bot attacks by analyzing large datasets and detecting patterns and anomalies. The state of the art methodologies for detecting bot infection are reviewed deeply and compared based on adopted methodologies, dataset and feature selection mechanism. The paper further discusses the pros and cons of existing methodologies. Finally, research gaps are presented to help future research on enhancing bot detection.