An Intrusion Detection System (IDS) is a form of network security that monitors for malicious activity. In the cloud, an IDS might operate on individual hosts or over an entire network. DDoS attacks are so named because they involve multiple computers working together to disable a service so that it can be taken offline entirely. By diminishing victim resources, the assaults prevent legitimate users from accessing the targeted services. In a classification problem, the classes are already defined and do not overlap before the method is applied, making classification itself a very straightforward problem to solve. Back Propagation (BP), Multi-Layer Perceptron Neural Network (MLP-NN), Levenberg-Marquardt (LM) method, MLPNN-BP, MLPNN-LM, and Radial Basis Function (RBF) were all proposed in this paper (RBF). It is possible that the BP algorithm can be employed successfully for training Neural Networks (NN). After the learning phase, the MLP-NN is able to generalise to completely new data. During the training phase, the weights are optimised to minimise the occurrence of any specified error functions. The weight configuration used as a starting point by the LM algorithm is used by all MLP-NNs in this combination method. After the LM training is complete, MLP-NN will use the validation data's MSE to compute. RBF was developed to address the interpolation issue with multiple variables. This is an example of a local approximation of a NN, and it has a high approximation power, a high classification accuracy, and a fast learning rate.