Presentation Title

Comparing Machine Learning Algorithms for Automated Detection of Jihadi Webpages

Presenter Information

Sam ChurchFollow

Advisor Information

Magie Hall

Location

Dr. C.C. and Mabel L. Criss Library

Presentation Type

Poster

Start Date

2-3-2018 10:45 AM

End Date

2-3-2018 12:00 PM

Abstract

The increased use of social media by jihadist groups such as The Islamic State in Iraq and the Levant (ISIL) has grabbed international attention. Though ISIL was not the first group to use the internet to further their ideological message, the group has arguably been the most successful. ISIL and others like it use a variety of open services, such as Twitter, YouTube, and JustPaste.it, to publish their content online. Each service has its own characteristics as to what type of content can be shared, how it can be shared, and how users can interact with the content. These differences make detection and removal increasingly difficult to orchestrate. This study focuses on content published to the site Justpaste.it, a simple open web service for publishing text and image based messages. We compare three different machine learning algorithms performance based on accuracy of classification. Our results indicate that while not all algorithms perform the same, machine learning has potential to automate the process of detecting jihadi material. In the future, automated tools could assist human analysts, enabling them to be more effective and saving a considerable amount of effort.

This document is currently not available here.

COinS
 
Mar 2nd, 10:45 AM Mar 2nd, 12:00 PM

Comparing Machine Learning Algorithms for Automated Detection of Jihadi Webpages

Dr. C.C. and Mabel L. Criss Library

The increased use of social media by jihadist groups such as The Islamic State in Iraq and the Levant (ISIL) has grabbed international attention. Though ISIL was not the first group to use the internet to further their ideological message, the group has arguably been the most successful. ISIL and others like it use a variety of open services, such as Twitter, YouTube, and JustPaste.it, to publish their content online. Each service has its own characteristics as to what type of content can be shared, how it can be shared, and how users can interact with the content. These differences make detection and removal increasingly difficult to orchestrate. This study focuses on content published to the site Justpaste.it, a simple open web service for publishing text and image based messages. We compare three different machine learning algorithms performance based on accuracy of classification. Our results indicate that while not all algorithms perform the same, machine learning has potential to automate the process of detecting jihadi material. In the future, automated tools could assist human analysts, enabling them to be more effective and saving a considerable amount of effort.