Building Comprehensive Dataset: Android File and Unstructured Data Collection from APKs for Enhanced Vulnerability Detection

Abstract

Google's Android platform grew to become one of the world's largest mobile systems. An Android application can use any functionality that the platform provides. Malware may also be present in Android apps. That is why our objective in this research is to collect files from wide malware families and use them to help us and other researchers run vulnerability detection tests. From the most well-known Android malware projects, we gathered both benign and malicious files. This process yielded 20832 malicious files and 10856 benign files. In this study, we also disclosed the Android file collection step as well as our vulnerability identification technique. Machine learning approaches are frequently used to identify whether an APK file is tainted, with the goal of detecting malicious apps. We have been concentrating on machine learning approaches to discover the unknown vulnerability. To identify malware, the malware researcher must create his or her own dataset. As part of our dataset production process, we collect Android files, do dynamic analysis, and then extract their characteristics. We described the technique of producing data sets in this paper. The android file contains a lot of unformatted data in the form of text or XML files, which is difficult to analyze and store. Our objective in this context is to provide a dynamic analysis of these obtained Android files, allowing you to access the underlying information, such as system calls, network traffic, and permission requests made by specific apps. Using Dynamic Analysis, we are attempting to manage massive amounts of data and assure correct processing in the context of Android files. In this study, we offer MalwareDefender, a dynamic analytical tool that handles the challenging issues of evaluating and processing massive volumes of data.