PEARSON CORRELATION COEFFICIENT BASED IMPROVED LEAST SQUARE - SUPPORT VECTOR MACHINE FOR CYBER-ATTACK DETECTION IN INTERNET OF THINGS

Abstract

The wide utilization of smart devices and enormous weakness of security in networks has maximized a count of cyber-attacks in Internet of Things (IoT). The detection and classification of malicious traffic is essential for ensuring a security of system in IoT. For detecting and classifying the vulnerable threats in IoT, proposed a Pearson Correlation Coefficient (PCC) based Improved Least Square - Support Vector Machine (ILS-SVM). The proposed algorithm provided high classification accuracy and detection rate with NSL-KDD dataset. The data are pre-processed by checking the missing values and normalization of data. Then the features are selected by using PCC and the selected features are classified by ILS-SVM method with high detection rate. The metrics taken for evaluating the proposed algorithm are accuracy, detection rate, precision and f1-score. The proposed PCC algorithm achieved 99.71% accuracy, 99.03% detection rate, 99.26% precision and 99.37% f1-score that is performed well than previous methods like Hybrid Feature Reduced method and Multi-class Support Vector Machine (SVM).