A Survey on Deep Learning Approaches for Malware Detection and Classification

Abstract

Malware comes in so many varieties and families, it can be extremely difficult to identify, it can be stealthy, it can propagate, and it can avoid traditional security solutions. It's a huge cybersecurity threat. So, malware detection (MD) and classification are essential components to detect malicious software and protect computer systems from unauthorized access, data stealing and disrupting systems. This survey provides a comprehensive review of deep learning techniques for malware detection and classification, including the fundamentals of malware, malware types and families, deep learning malware detection, data representation, feature extraction and applications. The limitations of the traditional detection techniques are discussed, namely signature-based, behavior-based, and heuristic-based approaches. Furthermore, the survey goes on to discuss the deep learning malware classification approaches, namely Convolutional Neural Networks (CNNs) and Bidirectional Long Short-Term Memory (BiLSTM) networks. CNN-based representation learning methods are better suited to learning spatial features from malware representations while BiLSTM models can learn sequential features from behavioral or code-based data. Moreover, the various detection techniques (static, dynamic and hybrid) are discussed so that their role in malware analysis can be understood. The survey identifies the current challenges and gaps in research and emphasizes the need for strong, scalable and adaptive deep-learning models to combat new malware threats and enhance cybersecurity protection.