Online Library

The article describes a systematic approach used by the Swiss Intelligence Service (SAP). The procedure described is illustrated by the authentic case of messages among extremists going on the Website www.islamic-minbar.com, a case that was prosecuted and brought to verdict by the Attorney General of the Swiss Confederation.

Over the last four years as the Syrian uprising has grown into a full-blown civil war, a sinister parallel conflict has been fought out in cyberspace, with combatants wielding bytes and software rather than guns as they have battled for supremacy on Syria’s internet frontline.

But the consequences of this secret cyber war have been real and deadly – particularly for opponents of the Assad regime who have been targeted for arrest and torture as a consequence of personal information gleaned from their email traffic.

In some cases even the military plans of crucial rebel offensives had been hacked. But the opposition has been busy too, leaking President Bashar al-Assad’s embarrassing personal correspondence and eavesdropping on government troop deployments amid much else.

As a consequence Syria’s civil war has become fertile ground for ‘hacktivists’ from both sides – egged on and in some cases assisted by governments and agencies from outside the region.

In this special investigation for People & Power , Juliana Ruhfus has been finding out why some experts believe Syria’s electronic armies have been drawing up the blueprints for all wars of the future, conflicts that transcend traditional physical boundaries but which can be just as significant as those fought with tanks and missiles.

This study explores the synergies between two complementary interventions for countering online extremism: deplatforming, a “hard” intervention that involves removing extremist users from online platforms, and inoculation, a “soft” intervention that exposes users to weakened forms of extremist arguments to build resistance. We developed an agent-based model to simulate the diffusion and control of online extremism, using data from systematic reviews and empirical studies. The model focuses on a population of Bayesian agents within an anti-immigration Facebook group, where these agents form beliefs about far-right content spread by extremists. We introduce both deplatforming and inoculation as interventions in the model to control the spread of extremism and find strong evidence that inoculation has a synergistic effect on deplatforming. This synergy stems from inoculation’s ability to stabilise opinion dynamics at the group level, leading to more effective containment of extremist beliefs. The findings suggest that a combined approach involving both deplatforming and inoculation is more effective than implementing either intervention in isolation. We deploy a cost–benefit analysis to show that the existence of synergies at a structural level enables the identification of efficient resource allocation. Our article demonstrates the potential of applying agent-based modelling to the study of counter-extremism policies.

Social media platforms have recently seen an increase in the occurrence of hate speech discourse which has led to calls for improved detection methods. Most of these rely on annotated data, keywords, and a classification technique. While this approach provides good coverage, it can fall short when dealing with new terms produced by online extremist communities which act as original sources of words which have alternate hate speech meanings. These code words (which can be both created and adopted words) are designed to evade automatic detection and often have benign meanings in regular discourse. As an example, “skypes”, “googles”, and “yahoos” are all instances of words which have an alternate meaning that can be used for hate speech. This overlap introduces additional challenges when relying on keywords for both the collection of data that is specific to hate speech, and downstream classification. In this work, we develop a community detection approach for finding extremist hate speech communities and collecting data from their members. We also develop a word embedding model that learns the alternate hate speech meaning of words and demonstrate the candidacy of our code words with several annotation experiments, designed to determine if it is possible to recognize a word as being used for hate speech without knowing its alternate meaning. We report an inter-annotator agreement rate of K=0.871, and K=0.676 for data drawn from our extremist community and the keyword approach respectively, supporting our claim that hate speech detection is a contextual task and does not depend on a fixed list of keywords. Our goal is to advance the domain by providing a high quality hate speech dataset in addition to learned code words that can be fed into existing classification approaches, thus improving the accuracy of automated detection.

VOX-Pol Blog Post.

VOX-Pol Blog post.