Efficient Intrusion Detection System Using Convolutional Long Short Term Memory Network

In recent research, network intrusion detection systems have been improved using machine learning techniques. The rising usage of computer networks has made it extremely difficult to maintain their availability, confidentiality, and integrity (CIA). As a result, it is now crucial to detect intrusions in order to spot and address vulnerabilities in computer systems and networks as well as unauthorised access attempts. Utilising deep learning techniques, the current study aims to construct an intrusion detection and prevention system. The main objective is to develop a system that is capable of identifying and thwarting numerous attacks, which involves Denial of Service (DOS), R2L, Probe, and U2R. In contrast to signature-based identification, which is based on previously known attack patterns, anomaly detection use statistical methodologies to identify behaviour that differ from normal activity. The Convo-LSTM network is a persuasive approach combining Convolutional Neural Networks (CNNs) with Long Short-Term Memory (LSTM) networks. CNNs, which have been created to find spatial patterns, do exceptionally well at spotting signature-based attacks by identifying certain patterns or signatures connected to earlier attacks.When an attack is not part of the training dataset, LSTM networks are capable of identifying it as strange or suspicious. The implementation of a CNN-LSTM architecture in intrusion detection has a lot of potential since it can achieve high accuracy while minimising false alarms. The NSL-KDD dataset (Network Security Lab-KDD Cup 1999) is used to properly train the intrusion detection algorithm. The present dataset, considered an enhanced variant of the KDDCup99 dataset, has a broader set of characteristics that provide the basis for the model's accuracy in detecting and averting network intrusions.