Machine Learning Based Predictive Model for Intrusion Detection

Abstract

A software that examines network traffic and searches for inconsistencies is known as an Intrusion Detection System (IDS). Network changes that seem to be abnormal or unexpected could be evidence of fraud at any phase, from the beginning of an attempt through the end of an intrusion. Data sharing is required to be safe since it primarily relies on the internet. Encryption processes and verification are unsuitable for internet security, and firewalls are unable to recognize fragmented fake transmissions. Additionally, attackers frequently update their strategy, tools, techniques, and tactics, which can have bad consequences like productivity losses, financial harm, data loss, etc. Therefore, it is essential to set up a trustworthy IDS, which is an extremely difficult task. In this work, the accuracy of an IDS system is forecasted by using a variety of supervised Machine Learning (ML) algorithms, including Decision tree (DT), Random Forest (RT), K-Nearest Neighbor (KNN), and Logistic Regression (LR) models. For the analysis, the dataset is collected from Kaggle, and the method that produces the highest accuracy is recommended for making future forecasts of intrusion. Furthermore, the outcomes have resulted in accuracy, execution speed, precision, F-measure, and recall. Additionally, the random forest performed best with the highest accuracy of 98.65% which can be recommended for the enhanced dataset to be implemented for better results for an IDS.