Deep Learning-based Android Malware Detection with SMOTE

In recent decades, Android has become the most widely used platform for smartphones due to low cost, numerous apps, and ease of use. Android carries many vulnerabilities that attract malware attackers to spread various malicious software in Android devices. Therefore, Android malware detection has turned into a challenging task for researchers. This research work illustrates an advanced Android malware detection approach using deep learning techniques such as Long Short-Term Memory (LSTM), GRUClassifier, and Multi-Layer Perceptron (MLP). The proposed work uses two significant Android malware datasets, Drebin and CICMALANAL2017, to demonstrate categorization between benign and malware classes by analyzing complex patterns in Android app behaviors. Further, SMOTE (Synthetic Minority Over-sampling Technique) is used to ensure a balanced dataset. The methodology includes data collection, feature engineering, model architecture design, and validation. In comparison to other deep learning techniques, the LSTM model demonstrates superior performance by capturing long-range dependencies in sequential data, achieving an accuracy of 98%. This research addresses critical gaps in Android security, including the need for balanced datasets, transfer learning applications, and model optimization for real-time detection on resource-constrained devices.