Hybrid Approach for Android Malware Detection and Analysis through Artificial Intelligence Techniques

Abstract

In the realm of cybersecurity, the Android platform has become a primary target for malicious software, posing significant challenges in detecting and preventing malware threats. Despite the advancement in malware detection techniques, the rapid evolution and increasing sophistication of Android malware present ongoing challenges. Traditional methods often struggle with the dynamic nature of malware, the vast diversity of Android applications, and the constraints of the existing detection methods. This thesis addresses these challenges by proposing novel methods for malware detection and classification using static, dynamic, and hybrid analysis. newlineRecognizing the limitations in current methodologies, such as inadequate feature selection and the inability to adapt to new malware types, this work introduces innovative approaches to progress detection accuracy and efficiency. The research begins with quotDeepMDFC,quot a Deep Learning (DL) based approach for static analysis using a unique dataset of applications from 2019 to 2021, demonstrating superior detection and classification capabilities. The dynamic analysis section explores the use of Artificial Neural Networks (ANN) with the CICMalDroid2020 dataset, highlighting the effectiveness of different ANN optimizers and the role of feature selection in enhancing performance. newlineIn hybrid analysis, the thesis presents a novel combination of static and dynamic feature selection methods, integrating Information Gain and Recursive Feature Elimination techniques. This approach successfully identifies the informative features for malware detection, outperforming existing methods in terms of accuracy and efficiency. The work culminates with an innovative integrated feature selection technique, applying it to the Kronodroid dataset and achieving substantial accuracy with a reduced feature set. newlineThis comprehensive study not only advances the landscape of Android malware detection (AMD) but also sets a new standard for future research. By effectively addressing the limitations of existin