Distributed Denial of Service (DDoS) attacks on Web-based services have grown in both number and sophistication with the rise of advanced wireless technology and modern computing paradigms. Detecting these attacks in the sea of communication packets is very important. There were a lot of DDoS attacks that were directed at the network and transport layers at first. During the past few years, attackers have changed their strategies to try to get into the application layer. The application layer attacks could be more harmful and stealthier because the attack traffic and the normal traffic flows cannot be told apart. Distributed attacks are hard to fight because they can affect real computing resources as well as network bandwidth. DDoS attacks can also be made with smart devices that connect to the Internet, which can be infected and used as botnets. They use Deep Learning (D.L.) techniques like Convolutional Neural Network (C.N.N.) and variants of Recurrent Neural Networks (R.N.N.), such as Long Short-Term Memory (L.S.T.M.), Bidirectional L.S.T.M., Stacked L.S.T.M., and the Gat G.R.U.. These techniques have been used to detect (DDoS) attacks. The Portmap.csv file from the most recent DDoS dataset, CICDDoS2019, has been used to test D.L. approaches. Before giving the data to the D.L. approaches, the data is cleaned up. The pre-processed dataset is used to train and test the D.L. approaches. In the paper, we show how the D.L. approach works with multiple models and how they compare to each other.