Towards predictive analysis of android vulnerability using statistical codes and machine learning for IoT applications

Abstract

Recently, the Internet of Things (IoT) technology is used for several applications for exchanging information among various devices. The intelligent IoT based system utilizes an Android operating system because it is also primarily used in mobile devices. One of the main problems for different IoT applications is associated with android vulnerability is its complicated and large size. To overcome the main issue of IoT, the existing studies have proposed several effective prediction models using machine learning algorithms and software metrics. In this paper, we are focused on conducting android vulnerability prediction analysis using machine learning for intelligent IoT applications. We conducted an empirical investigation for examining security risk prediction of 1406 Android applications with varying levels of risk using a metric set of 21 static code metrics and 6 machine learning (ML) techniques. It is observed from results that ML algorithms have different performances for predicting security risks. RF algorithm performs better for Android applications of all risk levels. By analyzing the findings of the conducted empirical study, it is suggested that developers may consider object-oriented metrics and RF algorithm in the software development process for android based intelligent IoT systems.