Comparison of Advanced Classification Algorithms Based Intrusion Detection from Real-Time Dataset
R. Aswanandini, C. Deepa · Automatic Control and Computer Sciences · 2023
Abstract Advanced machine learning (ML) and deep learning (DL) methods have provided efficient intrusion detection with minimum probability of false positives. Most existing IDS models have been evaluated over out-dated benchmark cyber security IDS datasets such as KDDCUP 1999 or NSL-KDD since the collection and extraction of attack features from the real-time network traffic data is a complicated process. This paper aims at generating a real-time dataset for IDS evaluation and testing. The real-time data were obtained by analyzing the real traffic analysis in a computer network with an Internet connection to extract the attack and normal traffic features which were referenced from popular datasets such as KDDCUP 1999 and NSL-KDD. In this study, the real-time data is collected with two attack classes (denial-of-service (DoS) and shellcode) and normal class data for evaluation. Advanced classification approaches of hyper-heuristic support vector machines (HH-SVM), hyper-heuristic improved particle swarm optimization based support vector machines (HHIPSO-SVM), and hyper-heuristic firefly algorithm based convolutional neural networks (HHFA-CNN) along with fuzzy optimized independent component analysis (FOICA) dimensionality reduction technique are evaluated using this dataset. Experimental results showed the effectiveness of the advanced classification methods with and without dimensionality reduction technique with the HHFA-CNN with FOICA achieving high accuracy and reduced time complexity for the real-time dataset.