International Journal of Science, Engineering and TechnologyAuthors: Guru Angel Daisy M
Abstract: Due to the growing number and level of sophistication of cyber attacks, there is an urgent need for intelligent, adaptive and automated systems for detection and reaction on the threats. In this paper we propose a new Artificial Intelligence based Intrusion Detection System (AI-IDS), which uses a hybrid architecture with integration of ensembles of machine learning and deep learning algorithms in order to increase network security. Our method uses Random Forest, Support Vector Machine and Convolutional Neural Network algorithms to analyze traffic data and detect the normal traffic and the U2R attack with frequency less than one percent with accuracy of 98% and 94.3% respectively. According to our experiments on the NSL-KDD dataset, the performance of our system is significantly higher with false positive rate equal to 0.001%, precision equal to 0.99 and real time detection latency of 0.2 milliseconds. Multi class classification in our system helps us to classify DoS, Probe, R2L and U2R attacks much better than signature-based solutions do.
DOI: http://doi.org/10.5281/zenodo.21676208
