Artificial Intelligence outflanks all other machine learning classifiers in Network Intrusion Detection System on the realistic cyber dataset CSE-CIC-IDS2018 using cloud computing

Abstract

Our paramount task is to examine and detect network attacks, is one of the daunting tasks because the variety of attacks are day by day existing in colossal number. The program proposed detects botnet attacks using the newest CSE-CIC-IDS2018 cyber dataset published by the Canadian Cybersecurity Establishment (CIC). The cyber dataset can be accessed on AWS (Amazon Web Services). The realistic network dataset consists of all the modern and existing attacks such as Brute-force attacks and password cracking, Heartbleed, Botnet, DoS (Denial of Service), DDoS also known as Distributed Denial of Service, Web attacks i.e. vulnerable web app attacks, and infiltration of the network from inside. The objective of the proposed research is to identify a classification of Botnet attacks. Botnet attack is a Trojan Horse malware attack that poses a serious security threat to the banking and financial sectors. Since a specific classifier could possibly work for such datasets it is crucial to finish a comparative examination of classifiers in order to achieve the most noteworthy execution in such basic detection of network attacks. The proposed framework is to incorporate different classifier methods such as KNearset Neighbor classifier, Naïve Bayes, Adaboost with Decision Tree, Support Vector Machine classifier, Random Forest classifier, and Artificial Intelligence to distinguish a portrayal of botnet attacks on the recent and realistic cyber dataset CSE-CIC-IDS2018. The results of the classification are given as precise precision for the specific classifiers. And furthermore, the proposed framework uses the Calibration curve as a standard approach in analytical methods which generates reliability diagrams to check the predicted probabilities of various classifiers are well-calibrated or not. Finally, the displayed graph proves how well the artificial intelligence technique outperforms all other classifiers which generates reliability diagrams to check the predicted probabilities of various classifiers are well-calibrated or not.