Android Malware Detection on Imbalanced Data Using Deep Learning

---

Abstract

This Android remains the most widely adopted operating system due to its open nature, enabling users to install applications from various origins. Malware causes severe problems for Android users and steals personal data by injecting malware into various applications. Many previous works used machine learning models to detect malware but suffer from data imbalance and high feature size. In this work, we perform three tasks: 1) Data pre-processing using the Synthetic Minority Oversampling Technique to Fix the Class Inequality Problem, and Standardization is used to normalize the features; 2) Principal Component Analysis is applied to the dataset to shrink the dimensionality of the feature vector; 3) Multilayer Perceptron is used for classification. We performed the proposed approach on Drebin, Malgenome, and Maldroid2020 and it achieved 98.27, 98.15, and 97.12 accuracy respectively. The results of the experiments demonstrate that the proposed strategy is more accurate than previous studies.

---

Keywords

Android, PCA, SMOTE, Multilayer Perceptron, Malware