Online Library

VOX-Pol Blog post.

With the spread of social networks and their unfortunate use for hate speech, automatic detection of the latter has become a pressing problem. In this paper, we reproduce seven state-of-the-art hate speech detection models from prior work, and show that they perform well only when tested on the same type of data they were trained on. Based on these results, we argue that for successful hate speech detection, model architecture is less important than the type of data and labeling criteria. We further show that all proposed detection techniques are brittle against adversaries who can (automatically) insert typos, change word boundaries or add innocuous words to the original hate speech. A combination of these methods is also effective against Google Perspective – a cuttingedge solution from industry. Our experiments demonstrate that adversarial training does not completely mitigate the attacks, and using character-level features makes the models systematically more
attack-resistant than using word-level features.

With the spread of social networks and their unfortunate use for hate speech, automatic detection of the latter has become a pressing problem. In this paper, we reproduce seven state-of-the-art hate speech detection models from prior work, and show that they perform well only when tested on the same type of data they were trained on. Based on these results, we argue that for successful hate speech detection, model architecture is less important than the type of data and labeling criteria. We further show that all proposed detection techniques are brittle against adversaries who can (automatically) insert typos, change word boundaries or add innocuous words to the original hate speech. A combination of these methods is also effective against Google Perspective – a cutting edge solution from industry. Our experiments demonstrate that adversarial training does not completely mitigate the attacks, and using character-level features makes the models systematically more attack-resistant than using word-level features.

Content analysis of the posts from eleven Facebook accounts of known Dutch and Belgian foreign fighters (in Dutch). English summary is available here: http://leidensafetyandsecurityblog.nl/articles/foreign-fighters-on-social-media-an-analysis-of-11-facebook-accounts

It is no surprise to come across information or video on social or mainstream media that was posted by a terrorist organisation like Al-Shabaab. In this regard, researchers have attempted to answer the question of what terrorist organisations aim to achieve by gaining a strong foothold in cyberspace. This article explores the evolution of Al-Shabaab in terms of their media usage and presence—from a local insurgency using magazines and radio stations, to what can be described as a media mogul in Africa. The author explores how and why this group chooses to pursue a strong cyber presence, and what, if anything, Africa and the international community can do about it.

The so-called Islamic State (IS) has a sophisticated media strategy (Winter 2017), an important part of which has been its English-language online magazine Dabiq. Launched in July 2014, a total of fifteen issues of Dabiq in the two years that followed. These issues were disseminated in a variety of ways, including archive sites (Bodo and Speckhard 2017), web forums and file-sharing networks (Gambhir 2016), the dark web (Stacey 2017), and even via an attempt to sell the freely available magazines for profit through the online retailer Amazon (Masi 2015). One of the most important forums for dissemination, however, was the social media platform Twitter (Bodo and Speckhard 2017; Gambhir 2016; Cunningham, Everton and Schroeder 2015; Shaheen 2015). Released in July 2016, the theme of Dabiq issue 15 was ‘Break the Cross’. After referring to a number of attacks that had occurred in the preceding weeks, the issue’s foreword called on ‘pagan Christians’, ‘liberalist secularists’ and ‘sceptical atheists’ to ‘recognize their Creator and submit to Him’. In addition to regular features, including ‘Among the believers are men’ and ‘In the words of the enemy’ (which, in this issue’ focussed on Pope Francis), and advertisements for other IS media, issue 15 contained an 18 page feature article, also titled ‘Break the Cross’. Arguing that the Bible does not display the three hallmarks of a true divine text, the article discusses the doctrine of the Trinity, whether Jesus was crucified, and whether Paul’s New Testament teachings are authentic (repeatedly stating that Paul was a known liar). It then seeks to establish the authenticity of the Prophet Muhammad, before asking Christians rhetorically: ‘O People of the Scripture, follow the truth from your Lord, whom you claim to love. Would you follow your parents and ancestors if you knew they were walking into a fire?’ and concluding ‘Know well that our fight will continue until you are defeated and submit to the rule of your Creator, or until we achieve martyrdom. Allah has made our mission to wage war against disbelief until it ceases to exist, as he has ordered us to kill all pagans wherever they are found’. Drawing on an original dataset, in this article we examine the response to the release of Dabiq issue 15 on Twitter. We examine the response in two respects: first, the response from Twitter itself, in terms of suspension activity; and, second, the response from other users, in terms of their engagement with posts disseminating the new issue. Before presenting our findings, we begin by offering an overview of our methodology.