Software-Defined Networking (commonly referred to as SDN) is a newer paradigm that develops the concept of a software-driven network by separating data and control planes. It can handle the traditional network problems. However, this excellent architecture is subjected to various security threats. One of these issues is the distributed denial of service (DDoS) attack, which is difficult to contain in this kind of software-based network. Several security solutions have been proposed recently to secure SDN against DDoS attacks. This paper aims to analyze and discuss machine learning-based systems for SDN security networks from DDoS attack. The results have indicated that the algorithms for machine learning can be used to detect DDoS attacks in SDN efficiently. From machine learning approaches, it can be explored that the best way to detect DDoS attack is based on utilizing deep learning procedures.Moreover, analyze the methods that combine it with other machine learning techniques. The most benefits that can be achieved from using the deep learning methods are the ability to do both feature extraction along with data classification; the ability to extract the specific information from partial data. Nevertheless, it is appropriate to recognize the low-rate attack, and it can get more computation resources than other machine learning where it can use graphics processing unit (GPU) rather than central processing unit (CPU) for carrying out the matrix operations, making the processes computationally effective and fast.