Detecting Low Rating Android Apps Before They Have Reached the Market

TL;DR

This paper introduces Sextant, a static analysis and machine learning-based method to detect low-rating Android apps from their APK files before they reach users, helping app markets protect their reputation.

Contribution

The paper presents a novel static analysis and machine learning approach, Sextant, for early detection of low-rating Android apps from APK files, reducing reputation risk.

Findings

Achieves 90.50% precision in detection

Achieves 94.31% recall in detection

Effective in preventing low-rating apps from reaching users

Abstract

Driven by the popularity of the Android system, Android app markets enjoy a booming prosperity in recent years. One critical problem for modern Android app markets is how to prevent apps that are going to receive low ratings from reaching end users. For this purpose, traditional approaches have to publish an app first and then collect enough user ratings and reviews so as to determine whether the app is favored by end users or not. In this way, however, the reputation of the app market has already been damaged. To address this problem, we propose a novel technique, i.e., Sextant , to detect low rating Android apps based on the .apk files.With our proposed technique, an Android app market can prevent from risking its reputation on exposing low rating apps to users. Sextant is developed based on novel static analysis techniques as well as machine learning techniques. In our study, our proposed approach can achieve on average 90.50% precision and 94.31% recall.