Evaluating Security Anomalies by Classifying Traffic Using Deep Learning
M.H. Samadzadeh, Najmeh Farajipour Ghohroud · 2023
Network traffic identification is an essential function for network domain systems, which facilitates accurate management through the classification of network traffic flows. In this research we used traffic separation using deep learning approach to detect security anomalies. The method proposed has several steps. Since many features are usually used to detect network security anomalies, in the first stage, feature selection was an optional step to select some of the most important features associated with the problem of detecting security anomalies in the network. Then the SMOTE balancing method was used to balance the data when the evaluated data set was unbalanced in class distribution. The results of balanced data and imbalanced data were obtained. Ultimately, the convolutional neural network was used to train the proposed model. The proposed model was tested and evaluated after training the model. The evaluation results indicated that in the mode of feature reduction and data balancing, the proposed CNN classifier showed the accuracy of 96.88% and 98.18% in feature reduction and data imbalance mode, when using no feature reduction and data balancing mode we reached the accuracy of 97.35% and 98.57% accuracy in feature reduction and non-balancing data.