Network Intrusion Detection Through Stacked Machine Learning Models on UNSW-NB15 Dataset

As an essential component of system defence, intrusion detection systems (IDSs) are now attracting a lot of attention. IDSs employ the data they gather about network traffic from different locations inside a computer system or network to protect the network. The vast amount of data and intricate patterns involved make it difficult and time-consuming to distinguish between legitimate and malicious network activity. In the past, finding incursion sequences in network connections required an analyst to go over large datasets. As a result, effective techniques that take into accountthe state of network traffic are required to identify network intrusions. This research offers a novel machine learning technique for detecting intrusions systems (IDSs) that identifiesintrusion features using a stacking classifier. The method integrates multiple models with a stacking classifier to improve prediction accuracy, to generate rules through categorisation. These rules, which are used in the stacking classifier to detect intrusions and carry out preventative measures by rejecting intrusions, are able to recognise intrusion characteristics. Among the attack types that are categorized are Shellcode, Worms, Fuzzers, Reconnaissance, Analysis, Backdoors, DoS, Exploits, Generic, and Normal. The following methods were used to assess the recommended strategy's efficacy: The three types of errors are Root Mean Squared Error (RMSE), Mean Squared Error (MSE), and Mean Absolute Error (MAE). Through the use of group learning's advantages, this strategy seeks to improve intrusion detection systems' effectiveness(IDSs) and offer a more reliable and accurate means of detecting and preventing network intrusions.