Faber, Lennart Niels (2021) Recognising Darknet market vendors using author verification methods. Master's Thesis / Essay, Artificial Intelligence.
|
Text
mAI_2021_FaberLN.pdf Download (2MB) | Preview |
|
![[img]](https://fse.studenttheses.ub.rug.nl/style/images/fileicons/text.png) |
Text
toestemming.pdf Restricted to Registered users only Download (121kB) |
Abstract
The dark web has provided criminals with new ways of selling their illegal goods. Law enforcement agencies are continuously trying to lift some of the anonymity provided by onion routing in order to run investigations and prosecute these individuals. One approach to this challenge is to find links between multiple profiles managed by a single person in the hope of discovering revealing information on one of them. We will support this endeavor by applying machine learning techniques to link profiles by means of author verification. Several classification algorithms trained on handcrafted stylometric features are compared with Siamese neural networks relying on two types of word embeddings. To assess the capabilities of these models, they are evaluated using data scraped from the dark web, created by various numbers of authors, as well as texts provided for a well-known public shared task on author verification. Additionally, the benefit of adding meta-features to the input of some models is experimented with. A support vector machine trained on handcrafted features performed best when dealing with texts written by a small set of authors. The Siamese neural network-based approach delivered the highest scores in many-author scenarios and comes with an additional benefit when applied in production. Furthermore, meta-features have shown to be significantly improving results.
| Item Type: | Thesis (Master's Thesis / Essay) |
|---|---|
| Supervisor name: | Wiering, M.A. |
| Degree programme: | Artificial Intelligence |
| Thesis type: | Master's Thesis / Essay |
| Language: | English |
| Date Deposited: | 01 Jul 2021 09:54 |
| Last Modified: | 01 Jul 2021 09:54 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/24864 |
Actions (login required)
 |
View Item |