A Lightweight Explainable Malware Detection Framework Using Hybrid Features and Adaptive Machine Learning Models

Cybersecurity and Machine Learning

The emergence of sophisticated malwares and the extensive use of mobile and IoT devices has driven the development of high-precision and effective detection approaches. In this study (2022–2024), we review ten recent works on malware detection using machine learning, in the context of static, dynamic, hybrid, and federated learning approaches. Some other techniques, including image-based classification, opcode analysis, PCA, RFE, as well as light model like Random Forest, XGBoost, and federated deep learning, were employed on data sets such as Drebin, CICMalDroid and AndroZoo achieving impressive results. However, most of the current models suffer from the problems of poor explainability, heavy computation and lack of protection to zero-day attacks. We propose a light-weight hybrid framework, integrating adaptive learning approaches with interpretable static and dynamic features, to address these issues. The method is intended to supply real- time, open, and scalable malware detection suitable for mobile and Internet of Things (IoT) contexts.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.