Attack Detection and Alarming System on IOT Facilities Using Random Forest Enabled-Correlation Based Clustering (RF-CBC) Technique

Abstract

In the past decade, Internet Of Things (IOT) technology has become one of the fastest-growing and most widely used technologies and is rapidly becoming a basic feature of global civilization. However, the high connectivity and diversity of these IOT devices make them complex and vulnerable to both visible and invisible security threats that are capable of causing irrecoverable damage. To alleviate these challenges, this work presents a novel and analytical hybrid machine learning model that suitably combines the Random Forest with Correlation-Based Clustering techniques, in order to report and detect potential attacks on IOT facilities. This work also showcases the development of the Single Threshold Boxplot Outlier-Based feature scaling method (STBO). The (STBO) method is used to scale down the number of attributes in order to select the best feature at the pre-processing stage of the attack detection procedures. The implementation of the present system is accomplished with the aid of an in-house Python program using XAMP/Apache HTTP as the hosting server with MySQL application for database development and management. A comparative analysis of the present model alongside ANN, Traditional Random Forest, Naïve Bayes, and the traditional Clustering method shows that the proposed system outperformed the baseline methods studied, with precision rates and attack detection quality equal to or greater than 75% in most cases, and is therefore capable of providing a useful, faster, efficient, and accurate anomalous detection online and on a real-time basis consistently with low false positive and negative rates.