Comparison of Data Fluctuations that Lead to Cyber Security Attacks: A Difference between Surface, Deep and Dark Net

Abstract

The term "darknet" refers to the address space on the internet that is not being used, and users do not anticipate that this area will interact with their machines. Darknet is a source of cyber intelligence. In order to develop network security, it is necessary to conduct studies of the many dangers that comprise the network. In this research, we offer brand new machine learning classifiers that go by the name stacking ensemble learning. Their purpose is to evaluate and categorize darknet traffic. This novel approach employs predictions created by three different base learning techniques in order to deal with the issues relating to darknet attacks. The software was validated using a dataset that had more than 141,000 records and was derived from the CIC-Darknet 2020 database. The findings of the experiment indicated that the classifiers used in the investigation were able to easily differentiate between benign and malignant traffic. The classifiers have the ability to efficiently recognize known as well as unknown threats with a high degree of precision and accuracy that is greater than 99% in the training and 97% in the testing phases, with increments ranging from 4 to 64% based on the algorithms that are currently in use. As a consequence of this, the suggested system will become more reliable and accurate as more data is collected. Additionally, in comparison to other AI algorithms already available, the suggested system has the lowest standard deviation.