Artificial Neural Networks for Cybersecurity: A Comprehensive Review

Abstract

Cybersecurity is a significantly growing field that primarily focuses on safeguarding systems,
networks, and data from digital attacks. With the advancement of the Internet and cyber
attacks, developing new cybersecurity tools has been vital, especially in the Internet of things
networks. A systematic review of the application of deep learning approaches for
cybersecurity is presented in this paper. This paper explains the short description of DL
methods used in cybersecurity, such as deep belief networks, generative adversarial networks,
recurrent neural networks, etc. Secondly, this review shows the difference between shallow
learning and DL. This paper then provides a discussion of the most common cyber-attacks in
IoT and other networks and whether DL methods are effective in mitigating them.
Furthermore, this paper presents examples of some Studies showing the DL method
cybersecurity utilized, and the datasets literature source. The last review in this paper is the
DL system’s feasibility analysis for malware detection and classification, IDS (intrusion
detection system), and other most common cyber-attacks, such as file type, spam, and
network traffic. We found that the RBM can reach a classification accuracy of 99.72% on a
custom dataset, while the LSTM on KDDCup99 attains accuracy of 99.80%. Finally, this
paper concludes by discussing the significance of cybersecurity to achieve dependable and
viable, IoT, and data-driven healthcare.