Online Library

Since the inception of the World Wide Web, security agencies, researchers, and analysts have focused much of their attention on the sentiment found on hate-inspired web-forums. Here, one of their goals has been to detect and measure users’ affects that are expressed in the forums as well as identify how users’ affects change over time. Manual inspection has been one way to do this; however, as the number of discussion posts and sub-forums increase, there has been a growing need for an automated system that can assist humans in their analysis. The aim of this paper, then, is to detect and measure a number of affects expressed in written text on Stormfront.org, the most visited hate forum on the Web. To do this, we used a machine learning approach where we trained a model to recognize affects on three sub-forums: Ideology and Philosophy, For Stormfront Ladies Only, and Stormfront Ireland. The training data consisted of manual annotated data and the affects we focused on were racism, aggression, and worries. Results indicate that even though measuring affects is a subjective process, machine learning is a promising way forward to analyze and measure the presence of different affects on hate forums.

The growing penetration of online social networks (OSNs) has emerged as a strong operational tool for spreading online radicalization by adding intensity and acceleration to the operations of radical groups. The efficiency of ways and methods proposed to measure the dynamics of radicalization are limited by the availability of models for modeling the intricacies of human dynamics and heterogeneous time varying interactions. Most of the developed algorithms and tools mainly focus on the static network structures and tend to ignore the “origin” of a new radical idea or propagation of a “weak” topic which might find its relevance in the security domain later on with the passage of time. In this article we propose a mathematical model for measuring dynamic propagation of radicalization in OSNs using Discrete Time Markov Chains by discretizing the continuous time domain. We also discuss our experimental results obtained through the information network of Twitter. The approach presented here is scalable and can easily be extended for modeling diffusion of general topics in OSNs.

Researchers have previously explored how right-wing extremists build a collective identity online by targeting their perceived “threat,” but little is known about how this “us” versus “them” dynamic evolves over time. This study uses a sentiment analysis-based algorithm that adapts criminal career measures, as well as semi-parametric group-based modeling, to evaluate how users’ anti-Semitic, anti-Black, and anti-LGBTQ posting behaviors develop on a sub-forum of the most conspicuous white supremacy forum. The results highlight the extent to which authors target their key adversaries over time, as well as the applicability of a criminal career approach in measuring radical posting trajectories online.

Terrorist groups like the Islamic State of Iraq and Syria (ISIS) have exploited social media such as Twitter to spread their propaganda and to recruit new members. In this work we study the extent to which ISIS is able to spread their message beyond their immediate supporters. Are they operating in their own sphere with limited interaction with the overall community? Or are they well rooted among normal users? We find that three-quarters of the interactions ISIS received on Twitter in 2015 actually came from eventually suspended accounts raising questions about the potential number of ISIS-related accounts and how organic ISIS audience is. Towards tackling these questions, we have created a unique dataset of 17 million ISIS-related tweets posted in 2015. This dataset is available for research purposes upon request.

VOX-Pol Blog post.

We are delighted to announce the launch of the final report of the Mechachal project, one of first academic studies to contextually examine how hate speech emerges and disseminates in social media.

Focusing on Ethiopia, and in collaboration with Addis Ababa University, the research team examined thousands of comments made by Ethiopians on Facebook during four months around the time of the country’s general election. Hate speech’ –defined as statements to incite others to discriminate or act against individuals or groups on grounds of their ethnicity, nationality, religion or gender – was found in just 0.7% of overall statements in the representative sample. The paper says the findings may have wide implications for the many countries trying to address growing concerns about the role played by social media in promoting radicalisation or violence.

Ethiopia represented an exceptional case study because of its distinct languages, which allowed the research team to gain a realistic sample of the overall online debates focused on one country. The research team analysed Facebook statements made by Ethiopians, both in their homeland and abroad, in the run-up to and just after the general election on 24 May 2015. Fans or followers, rather than people with any real influence online, were found to be mainly responsible for the violent or aggressive speech that appeared on Facebook pages in the sample.

It appears these individuals use Facebook to vent their anger against more powerful sections of society. Around 18% of total comments in the sample were written by fans or followers compared with 11% of comments made by highly influential speakers (the owners of web pages). One fifth (21.8%) of hostile comments were grounded in political differences, only slightly higher than the overall average of 21.4% of all conversations containing hostile comments. Religion and ethnicity provoked fewer hostile comments (10% and 14% of overall comments in sample respectively).