Anomaly based Malicious Traffic Identification using Kernel Extreme Machine Learning (KELM) Classifier and Kernel Principal Component Analysis (KPCA)

Objectives: The rapid growth of new vulnerabilities causes the network by Denial of Service attack (DoS). The DoSattack causes traffic flow in network. Therefore it increases the difficulties to detect the DoSattack in traffic by means of misuse detection. The behavior patterns are analyzed in anomaly Anomaly detection to identify the attack. Methods: In detection of unknown worms anomaly detection is more comfortable than misuse detection. In this paper, hybrid optimization and extreme machine learning classifier is proposed for anomaly detection. This approach detects the DoSattack by analyzing the profiles of traffic patterns. Findings: Principal Component Analysis (PCA) is adopted in this approach to extract the feature from the dataset. A short time window is utilized to gather all features from packet headers. Extreme learning machine based HGAPSO is used to classify the unknown attack. Improvement: thus the proposed system is implemented as real-time. Performance evaluation shows that this approach provides 1.016s time consumption and 95 % accuracy tan existing approach during detection of DoSin network traffic.