Network Attack Prediction By Random Forest : Classification Method
M. K. Krishna Prasath, B. Perumal · 2019 3rd International conference on Electronics, Communication and Aerospace Technology (ICECA) · 2019
Proposed system comprised of 32 honeypots with reports from 17M login attempts provided by many countries from 6000 different source IP addresses. Increasing number of attacks for blocking network connections in switch level was handled by Software Defined Network because of decoupled data in control plane. The Motive of software defined networks was defining rules based on the SDN controller for blocking unauthorized network connections. An historical network attack detects a data, blocks the untrusted connections. Though, each attacker provides solutions cannot effectively against chain attacks that contain many IP address utilized. In this paper, a machine learning algorithm on historical network attack data is trained for detecting potential unauthorized connections and potential attack destinations. Decision Tree (DT) and Naïve-Bayes is used predicted host attacked depending on the historical data. Software attack pattern was predicted to identify unauthorized user that is same as Identifying intrusion in system. The Proposed system shows that average accuracy prediction using Bayesian networks are 91.68%.