Malicious DNS Detection and Prediction Using SMOTE-ENN and Hybrid Artificial Neural Network

DNS (Domain Name Service) is a critical component of the internet’s operation. Malicious actors use domain names to direct customers to their services, much like corporations, to make their computing services more accessible. It is critical to distinguish between harmful and non-malicious web addresses since it enables the grant and denies access to external service providers, increasing the company’s and its users’ security. This paper suggested a deep neural network-based method that can detect and predict malicious domains and web addresses using 45000 domains chosen at random. Between September and November 2020, data
¹
was taken from Rapid7 Labs, an open-source data source, and SANS has a well-known hazardous list. Synthetic Minority Oversampling Technique (SMOTE) and Edited Nearest Neighbor(ENN) [SMOTE-ENN] and Hybrid Deep Artificial Neural Network, and perform oversampling, under-sampling, and select principal components using the Chi-Square method with an accuracy of 95%, Precision 97%, Recall 97%, F1 Score 94% and Roc 94%.