Abstract :Internet grid and computer networks have already transformed the way we live. Today,
governments and commercial firms deployed their services to internet. As a result, we use internet-
connected devices in every aspect of life. Since modern life is intertwined with computers and connected
devices, maintaining the security of internet is very important and considered as a strategic
infrastructure. To maintain the stable and integral operation of cyber assets, many cybersecurity tools
and products have been developed and commercially available. Most of those cybersecurity products
make use of machine learning because classical signature-based methods are unable to detect zero-day
attacks or slightly deformed versions of known attacks. Moreover, machine learning algorithms can
also be trained to detect abnormal network traffic. In this work, we explain how machine learning, deep
learning, and transfer learning methods are being used in cybersecurity. Traditional machine learning
algorithms such as support vector machines, decision trees, etc., have been implemented to well-known
cybersecurity datasets for detecting intrusion attacks and deep neural networks have been implemented
to malware analysis. Another popular implementation of deep learning is to use models that were
trained on large image datasets in a transfer learning paradigm. In these models, malware binaries are
converted to images and pre-trained networks are fed with those images. The performance of the pre-
trained models is very promising, and they can detect malware software with high accuracy. Keywords : Cybersecurity, Machine Learning, Deep Learning, Deep Transfer Learning, Malware
Detection
