A Robust Multiple Network Attacks Detection Method Based on Artificial Neural Network

During the last decade of the development of computer networks, it is more and more important to identify multiple network attacks to improve computer security. This paper based is on NSL-KDD datasets to achieve the purpose of identifying network attacks. This research not only focuses on improving the accuracy that got from training datasets but also manages to improve the accuracy that gets from official test datasets which is more similar to real life. To get the best accuracy, we applied Random Forest, which is the best model previously. In this model, we use several data reduction methods to improve model performance. Next, we propose a model that has not been used before, which is Artificial Neural Network. According to the accuracy we get from ANN, we found that this model has better performance than traditional models, which increase test accuracy from 0.759 to 0.825. The results showed that ANN has entirely satisfactory performance in intrusion detection.