An Anomaly Based Network Intrusion Detection System Using LSTM and GRU

Today, billions of devices are connected to the internet and the count keeps on increasing. Most of these networked devices are vulnerable to security attacks. Various types of active and passive attacks will cause severe damage to the privacy and security of millions of users. Even though we have firewalls for security but they limit the access between the networks to prevent intrusion, but they do not alert when there is an attack inside a network. The proposed method uses NSL-KDD dataset, a benchmark dataset from the Canadian Institute for Cybersecurity. We created a network intrusion detector using this dataset, which is a prediction model capable of distinguishing between “bad” connections, which are subsequently categorised into the classifications DoS, Probe, and R2L, and “good” normal connections. This study offers a Deep Learning-based Network Intrusion System. After training the model we achieve a good accuracy and precision. The highest accuracy of about 96% for classifying the Probe Attack. We got 92% accuracy for DOS attack and 88% for R2L.