The Future Roadmap for Cyber-attack Detection

Abstract

Cyber-attacks can cause delays in world operations and substantial economic losses. Therefore, there is a greater interest in cyber-attack detection (CAD) to accommodate the exponential increase in the number of attacks. Various CAD techniques have been developed, including Machine Learning (ML) and Deep Learning (DL). Despite the high accuracy of the deep learning-based method when learning from large amounts of data, the performance drops considerably when learning from imbalanced data. While many studies have been conducted on imbalanced data, the majority possess weaknesses that can lead to data loss or overfitting. However, Generative Adversarial Networks can help solve problems such as overfitting and class overlapping by generating new virtual data similar to the existing data. This paper provides a comprehensive overview of the current literature in CAD methods, thus shedding light on present research and drawing a future road map for cyber-attack detection in different applications.