Abstract
Currently over 2 billion active devices are running Android operating system. At present more than 2.2 million applications are accessible for download from Android's application store called Google Play. Android is the most popular opensource mobile operating system, though its security is still challengeable. There are many reports of user's privacy being voilated due to vulnerable mobile applications. A report published by McAfee in 2016 showed that over 2.5 million new malwares were found just in the last quadrant of 2016. Various strategies have been proposed to recognize pernicious applications, some use sequences of permissions to determine the malicious nature of applications, while others look into different system calls triggered by applications during its execution.
In this paper, we propose the use of intents raised by applications as a metric to identify the malicious behavior of an application. For this purpose, we generated a dataset that contained more than 30,000 applications (15,000 malicious and 15,000 benign applications), which were used to train the proposed model with different machine learning algorithms using most common events. The results have shown acceptable detection rate of malicious behavior with the help of intents. We can deduce that, our proposed model, is a novel and smart way of detecting malicious behavior in Android applications.