Deep abstraction for Android malware detection and recent development on privacy-preserving deep learning

Abstract

In this dissertation, we present our study about deep learning in two parts. The first part is about leveraging deep learning for Android malware detection. The second part is about Privacy-Preserving Deep Learning (PPDL) for Machine Learning as a Service (MLaaS) Part I is focused on implementing feature learning for Android malware detection. The current Android malware detection method is limited to two kinds of methods, static and dynamic. Static method is easy to use but difficult to detect new kinds of malware. On the other hand, dynamic method is strong against a new malware but needs an expert skill to manipulate it. For the last decades, machine learning has advanced rapidly as a new malware detection method. We propose a modified feature learning method for malware detection, which is based on Deep Abstraction and Weighted Feature Selection proposed (DFES) for Intrusion Detection System. The methodology consists of a combination between Stacked Autoencoder (SAE) for feature extraction and weight based Artificial Neural Network (ANN) for feature selection and classification. The goal of this dissertation is to conduct a study to compare the performance of Modified DFES (mDFES), DFES, and a simple Feature Extraction and Selection (FES). Part II is focused on study about leveraging deep learning for privacy-preserving. The exponential growth of big data and deep learning has increased the data exchange traffic in society. MLaaS, which leverages deep learning techniques for predictive analytics to enhance decision-making, has become a hot commodity. However, the adoption of MLaaS introduces data privacy challenges for data owners and security challenges for deep learning model owners. Data owners are concerned about the safety and privacy of their data on MLaaS platforms, while MLaaS platform owners worry that their models could be stolen by adversaries who pose as clients. Consequently, PPDL arises as a possible solution to this problem.%Recently, several papers about PPDL for MLaaS have been published. However, to the best of our knowledge, no previous work has summarized the existing literature on PPDL and its specific applicability to the MLaaS environment. We present a comprehensive study of privacy-preserving techniques, starting from classical privacy-preserving techniques to well-known deep learning techniques. Additionally, we provide a detailed description of PPDL and address the issue of using PPDL for MLaaS. Furthermore, we undertake detailed comparisons between state-of-the-art PPDL methods. Subsequently, we classify an adversarial model on PPDL by highlighting possible PPDL attacks and their potential solutions. Ultimately, our study serves as a single point of reference for detailed knowledge on PPDL and its applicability to MLaaS environments for both new and experienced researchers.