Detecting Multi-class Hostile Traffic in the Network Using a Hybrid Model

Hostile cum malicious traffic poses a significant threat to network systems and the security of sensitive data. By staying vigilant, organizations can better protect themselves from these threats and minimize the potential impact in their network. This paper presents a hybrid model for the detection of hostile traffic on a network system. For detecting multi-class of hostile traffics on the network system, we utilized a dataset that comprise 9 types of malicious attacks on a network system. In order to have an efficient model, we conduct exploratory data analysis on the dataset. The exploratory data analysis was used in checking missing values, correlated features, data imbalance, and also important features. The results of the exploratory data analysis show that the dataset is imbalanced which will lead to overfitting if not resolved. We resolved the data imbalance by performing random oversampling by utilizing the Random Over-Sampling technique in python. After resolving the data imbalance, we used a random forest classifier to extract important features of the data. Here, we extracted ten features based on the ranking of the random forest model. The extracted features were used in training the proposed model for the detection of hostile traffic in a network system. The results of the model show a better accuracy for the detection of hostile traffic in a network system with high accuracy and F1-score.