Network Infrastructures in the Dark Web

With the appearance of the Internet, open to everyone in 1991, criminals saw a big opportunity in moving their organisations to the World Wide Web, taking advantage of these infrastructures as it allowed higher mobility and scalability. Later on, in the year 2000, the first system appeared, creating...

Full description

Saved in:  
Bibliographic Details
Main Author: Teixeira, Octávio Miguel Silveira (Author)
Format: Electronic Book
Language:English
Published: 2021
In:Year: 2021
Online Access: Volltext (kostenfrei)
Check availability: HBZ Gateway
Description
Summary:With the appearance of the Internet, open to everyone in 1991, criminals saw a big opportunity in moving their organisations to the World Wide Web, taking advantage of these infrastructures as it allowed higher mobility and scalability. Later on, in the year 2000, the first system appeared, creating what is known today as the Dark Web. This layer of the World Wide Web became quickly the option to go when criminals wanted to sell and deliver content such as match-fixing, children pornography, drugs market, guns market, etc. This obscure side of the Dark Web, makes it a relevant topic to study in order to tackle this huge network and help to identify these malicious activities and actors. In this master thesis, it is shown through the study of two datasets from the Dark Web, that we are surrounded by capable technologies that can be applied to these types of problems in order to increase our knowledge about the data and reveal interesting characteristics in an interactive and useful way. One dataset has 10 000 relations from domains living in the Dark Web, and the other dataset has thousands of data from just 11 specific domains from the Dark Web. We reveal detailed information about each dataset by applying di↵erent analysis and data mining algorithms. For the first dataset we studied domains availability patterns with temporal analysis, we categorised domains with machine learning neural networks and we reveal the network topology and nodes relevance with social networks analysis and core-periphery model. Regarding the second dataset, we created a cross matching information web graph and applied a name entity recognition algorithm which ended in a tool for identifying entities within dark web’s domains. All of these approaches culminated in an interactive web application where we publicly not only display the entire research but also the tools developed along with the project (https://darkor.org).Com o surgimento da Internet, aberta a todos em 1991, os criminosos viram uma grande oportunidade em passar as suas organizações para a World Wide Web, aproveitando-se assim dessas infraestruturas que permitiam uma maior mobilidade e escalabilidade. Mais tarde, no ano 2000, surgiu o primeiro sistema, criando o que hoje é conhecido como a Dark Web. Essa camada da World Wide Web tornou-se rapidamente a opção a seguir quando os criminosos queriam vender e entregar conteúdo como combinação de resultados, pornografia infantil, mercado de drogas, mercado de armas, etc. Este lado obscuro da Dark Web, torna-a num tema relevante de estudo a fim de ajudar a identificar atividades e atores maliciosos. Nesta dissertação de mestrado é mostrado, através do estudo de dois conjuntos de dados da Dark Web, que estamos rodeados de tecnologias que podem ser aplicadas neste tipo de problemas de forma a aumentar o nosso conhecimento sobre os dados e revelar características interessantes de forma interativa e útil. Um conjunto de dados tem 10 000 relações de domínios que vivem na Dark Web enquanto que o outro conjunto de dados tem milhares de dados de apenas 11 domínios específicos da Dark Web. Neste estudo revelamos informações detalhadas sobre cada conjunto de dados aplicando diferentes análises e algoritmos de data mining. Para o primeiro conjunto de dados, estudamos padrões de disponibilidade de domínios com análise temporal, categorizamos domínios com o auxílio de redes neuronais e revelamos a topologia da rede e a relevância dos nós com análise de redes sociais e a aplicação de um modelo núcleo-periferia. Em relação ao segundo conjunto de dados, criamos um grafo da rede com cruzamento de dados e aplicamos um algoritmo de reconhecimento de entidades que resultou em uma ferramenta para identificar entidades dentro dos domínios da Dark Web estudados. Todas estas abordagens culminaram em uma aplicação web interativa onde exibimos publicamente não apenas todo o estudo, mas também as ferramentas desenvolvidas ao longo do projeto (https://darkor.org)