Analysis of Different Dimensionality Reduction Techniques and Machine Learning Algorithms for an Intrusion Detection System
T N Varunram, M B Shivaprasad, K H Aishwarya, Anush Balraj, S V Savish, S. Ullas
Abstract
Intrusion Detection Systems are of paramount importance in the present-day network security environment. By analyzing and predicting the behavior of incoming data, IDS helps in classifying it as either a benign or dangerous activity. In this paper, implementation of multiple Intrusion Detection Systems is done using different machine learning algorithms like KNN, AdaBoost, SVM, Logistic Regression, Random Forest, Artificial Neural Networks and Naive Bayes. The class imbalance in the dataset, caused by innate mannerisms of network analysis, is corrected using Synthetic Minority Oversampling Technique (SMOTE). The dataset is then reduced using three different Dimensionality Reduction Techniques namely PCA, t-SNE, and UMAP. The three datasets produced by these techniques are then used to build the Intrusion Detection System, using the machine learning algorithms. The models will implement binary classification to classify the incoming attacks between benign activity and DDoS Attacks.