Darknet traffic analysis, and classification system based on modified stacking ensemble learning algorithms

Darknet, a source of cyber intelligence, refers to the internet’s unused address space, which people do not expect to interact with their computers. The establishment of security requires analyses of the threats characterizing the network. New machine learning classifiers known as stacking ensemble...

Full description

Saved in:
Bibliographic Details
Published in:Information systems and e-business management Vol. 23; no. 1; pp. 209 - 240
Main Author: Almomani, Ammar
Format: Journal Article
Language:English
Published: Heidelberg Springer Nature B.V 01.03.2025
Springer Berlin Heidelberg
Subjects:
Algorithms
Artificial intelligence
Dark web
Ensemble learning
Machine learning
Original
Threat evaluation
Traffic analysis
ISSN:1617-9846, 1617-9854
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Darknet, a source of cyber intelligence, refers to the internet’s unused address space, which people do not expect to interact with their computers. The establishment of security requires analyses of the threats characterizing the network. New machine learning classifiers known as stacking ensemble learning are proposed in this paper to analyze and classify darknet traffic. In dealing with darknet attack problems, this new system uses predictions formed by 3 base learning techniques. The system was tested on a dataset comprising more than 141,000 records analyzed from CIC-Darknet 2020. The experiment results demonstrated the study’s classifiers’ ability to distinguish between the malignant traffic and benign traffic easily. The classifiers can effectively detect known and unknown threats with high precision and accuracy greater than 99% in the training and 97% in the testing phases, with increments ranging from 4 to 64% by current algorithms. As a result, the proposed system becomes more robust and accurate as data grows. Also, the proposed system has the best standard deviation compared with current A.I. algorithms.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:1617-9846
1617-9854
DOI:10.1007/s10257-023-00626-2