Deep Learning Based Static Analysis of Malwares in Android Applications

Abstract

Android is a widely distributed mobile operating system developed especially for mobile devices with touch screens. It is an open source, Google-distributed Linux-based mobile operating system. Since Android is open source, it enables Android devices to be targeted effectively by malware developers. Third-party markets do not search for malicious applications in their databases, so installing Android Application Packages (APKs) from these uncontrolled market places is often risky. Without user’s notice, these malware infected applications gain access to private user data, send text messages that costs the user, or hide malware apk file inside another application. The total number of new samples of Android malware amounted to 482,579 per month as of March 2020. In this paper deep learning approach that focuses on malware detection in android apps to protect data on user devices. We use different static features that are present in an Android application for the implementation of the proposed system. The system extracts various static features and gives them to the classifier for deep learning and shows the results. This proposed system will assist users in checking applications that are not downloaded from the official market.

Smali opcode based Android Malware detection and Obfuscation Identification

“…To assess the effectiveness of their approach, they evaluated their models on different obfuscation techniques commonly employed by malware authors, such as encryption, reflection, and trivial obfuscation and scored the best accuracy of 98.6%. Nivedha et al [24] collected a diverse set of static features from Android applications, including permissions, Intent usage, API calls, and n-gram opcodes. To effectively classify Android applications based on these static features, they employed a deep learning model and got an accuracy of 99.37%.…”