5. Exploring Potential of Transfer Deep Learning for Malicious Android Applications Detection

Mal Image represents any type of malicious executable (Windows files, APKs) for using image-based features for building classifiers. In recent years, Mal Image-based malware classification is getting attention, which provides a new approach to malware research and addresses some of the bottlenecks of traditional approaches. With the advancement in computing capacities in recent years, neural network research has gained tremendous attention. As a result, deep learning-based image classification techniques report very high accuracy for different classification tasks such as face detection and recognition, object identification, etc. In this proposed work, the authors have combined these two evolving techniques to improve android malware detection. For this chapter, the research involved experiments with transfer learning techniques under deep learning models and android malware detection techniques. The experimental result of various pre-trained models in terms of accuracy is in the range of 75–80%, but this technique can overcome bottlenecks such as analysis obstacles and obfuscation of traditional methods.