Data scientific approach to detect the DoS attack, Probe attack, R2L attack and U2R attack

Abstract

Data reliability is compromised by various cyber-attacks. Computational infrastructure is completely disturbed, broken or guided by these attacks. The current status of cyberspace foretells uncertainty for the future of the Internet and its rising user base. Data collected by the sensors and other input devices can be easily stolen by the unidentified user. It is a severe threat to the programming environment and individual personals. It is necessary to take into account the advanced technologies to counterfeit these cyber-attacks. Existing algorithms are decoded over certain period of time. Because of this always important to adapt the new technology that can prevent cyber-attacks. In this paper, various cyber-attacks predictions are analyzed and combines as a group based on its features. After analyzing the various cyber-attacks and its classification, recent technologies which can prevent the cyber-attack is studied. One of the main technologies that can able to learn themselves is the machine learning. Networking environment must use advanced machine learning approaches to protect the Data. Machine learning technique is classified as supervised and unsupervised technique. Supervised machine learning technique uses features that can be extracted from the source dataset. The most effective machine learning algorithm for predicting the types of cyber-attacks has been determined through a comparison study of different algorithms. We categorize attacks into four categories: R2L attacks, DOS attacks, U2R attacks, and probe attacks. Various machine learning algorithms are applied to detect and rectify the cyber-attacks. Their performances are compared and analyzed in terms of accuracy, F1 score, precision and recall.