This document describes the investigation process for a Deep Learning algorithm selection, training, and classification, to create a preventive system against web denial of service attacks called Dique type IDS / IPS. Besides, the Dique design and construction integrate a graphical user interface with the classification model, which verifies the operation of the system against denial of service attacks. In order to prevent denial of service attacks, the Dique system classifies input packets into two types: Non-Malignant and Malignant, malignant means packets that system classifies as possible denial of service attacks. This classification uses the MultiLayer Neural Network Deep Feed Forward and, for training, used Dataset called CICDDoS2019, which contains twelve types of denial of service attacks, and in the end, obtained an accuracy of 0.994. An offensive system called Diluvio was created to demonstrate how the preventive system works. Diluvio contains seven types of DoS attacks and can be targeted to the webserver. Dique has a graphical interface that allows for change between detection mode and prevention mode and displays the packages information and classification.