Deep Learning Approaches on Image Representations of Android Malware: A Review

Creative Commons License

Sezer O., Doğru İ. A., Kılıç K.

Artificial Intelligence Studies, cilt.8, sa.2, ss.170-176, 2025 (Hakemli Dergi)

Özet

This review emphasizes the transition from conventional static analysis to graph-driven, visionbased, and transformer-focused methodologies in deep learning-based and hybrid multimodal Android malware detection from 2023 to 2025. This review utilizes various pertinent studies from Elsevier, IEEE Xplore, and ScienceDirect to analyze ten representative methodologies selected for their methodological diversity and their contributions to visual, hybrid, and interpretable detection frameworks. End-to-end image models that turn DEX bytecode into grayscale or RGB matrices, graph-attention and multimodal fusion techniques that combine structural and semantic features, and deeper architectures like 3D-CNNs and Vision Transformers that can find multiscale contextual patterns are all examples of these kinds of methods. The works that were looked at also include a number of explainable AI methods that use SHAP or Grad-CAM, as well as lightweight learning frameworks that try to make models less complicated. There is a clear change from manually made static features to automatically learned image-based representations. For example, graph-based models can be up to 99.5% accurate, while CNN models like MADRF-CNN can be between 96% and 98% accurate. The remaining issues are adversarial robustness, computational cost, and dataset imbalance. Lightweight CNN-Transformer hybrid detectors, the development of balanced benchmarks with adversarial and obfuscated samples, and a more thorough integration of explainability within multimodal learning for real-time usability and robustness enhancement are the main focuses of emerging research directions. The idea that the next generation of scalable and robust Android malware defense systems has been propelled by the convergence of visual computing, graph reasoning, and explainable AI is supported by these trends taken together