Various fields were applied to machines and deep learning methods and found an effective tool. Cyber security is the most popular example of its applicability. Machine and deep learning, together with this technology, can disallow the network attacking and detect it at the earliest. It is also essential to take into consideration factors that ensure that deviations caused on the system might be indicative of an ongoing attack. These algorithms are a great tool for cyber security professionals since it improves the security of systems. This study attempts to build robust voting models with two machine learning algorithms (Decision Tree and Random Forest) and two deep learning (CNN and LSTM) using a well-known dataset, the CICIDS-2018. Before running the two algorithms, five feature selection methods were employed to come up with the important features from feature dataset. The four typical performance metrics are accuracy, precision, recall and f1-score in assessing the performance of these models. Among the CNN-LSTM models’ univariate feature selection methods, the one with the highest accuracy was given comparably better results compared to the others. The results illustrate that the detection of such attacks can be effectively implemented by these algorithms.