Online Library

Since the outbreak of the Syrian civil war in 2011 close to 30 000 foreign recruits from more than 100countries have migrated to the area of Iraq and Syria in support of the terrorist organization this thesis will refer to as ISIS. Among those traveling is a historically unprecedented number of women. Why women are drawn to violent Islamic extremist groups is rather unexplored. Through a qualitative text analysis of official ISIS-propaganda, this thesis investigates what promises the organization makes to women, examining pull-factors derived from social media studies of female migration to ISIS-held territories. The thesis concludes that women are promised the possibility to fulfill their religious duty, become important state builders, experience deep and meaningful belonging and sisterhood, to live an exciting adventure and find true romance, as well as being increasingly influential is also promised. Official propaganda does not make explicit promises to women of exerting violence. A secondary purpose of the thesis is to assess the potential risk that ISIS-affiliated women returning to the West, pose to society. This thesis further concludes that women who gain limited knowledge of handling weapons and explosives in ISIS-territory are not probable participants in armed terrorist attacks directed towards the West. However, through increased social networks acquired while in Syria or Iraq, women may play an important supporting role in the process of planning, crowdfunding and executing attacks. Based on these findings the thesis provides some gender-specific policy proposals intended to counter the recruitment of women to ISIS.

Online social networks have become a powerful venue for political activism. In many cases large, insular online communities form that have been shown to be powerful diffusion mechanisms of both misinformation and propaganda. In some cases, these groups users advocate actions or policies that could be construed as extreme along nearly any distribution of opinion and are thus called Online Extremist Communities (OECs). Although these communities appear increasingly common, little is known about how these groups form or the methods used to influence them. The work in this thesis provides researchers a methodological framework to study these groups by answering three critical research questions:
• How can we detect large dynamic online activist or extremist communities?
• What automated tools are used to build, isolate, and influence these communities?
• What methods can be used to gain novel insight into large online activist or extremist communities?
This group members social ties can be inferred based on the various affordances offered by OSNs for group curation. By developing heterogeneous, annotated graph representations of user behavior I can efficiently extract online activist discussion cores using an ensemble of unsupervised machine learning methods. I call this technique Ensemble Agreement Clustering. Through manual inspection, these discussion cores can then often be used as training data to detect the larger community. I present a novel supervised learning algorithm called Multiplex Vertex Classification for network bipartition on heterogeneous, annotated graphs. This methodological pipeline has also proven useful for social botnet detection, and a study of large, complex social botnets used for propaganda dissemination is provided as well. Throughout this thesis, I provide Twitter case studies including communities focused on the Islamic State of Iraq and al-Sham (ISIS), the ongoing Syrian Revolution, the Euromaidan Movement in Ukraine, as well as the alt-Right.

The ever-increasing use of the internet by terrorist groups as a platform for the dissemination of radical, violent ideologies is well documented. The internet has, in this way, become a breeding ground for potential lone-wolf terrorists; that is, individuals who commit acts of terror inspired by the ideological rhetoric emitted by terrorist organizations. These individuals are characterized by their lack of formal affiliation with terror organizations, making them difficult to intercept with traditional intelligence techniques. The radicalization of individuals on the internet poses a considerable threat to law enforcement and national security officials. This new medium of radicalization, however, also presents new opportunities for the interdiction of lone-wolf terrorism. This dissertation is an account of the development and evaluation of an information technology (IT) framework for detecting potentially radicalized individuals on social media sites and Web fora. Unifying Collective Action Framing Theory (CAFT) and a radicalization model of lone-wolf terrorism, this dissertation analyzes a corpus of propaganda documents produced by several, radically different, terror organizations. This analysis provides the building blocks to define a knowledge model of terrorist ideological framing that is implemented as a Semantic Web Ontology. Using several techniques for ontology guided information extraction, the resultant ontology can be accurately processed from textual data sources. This dissertation subsequently defines several techniques that leverage the populated ontological representation for automatically identifying individuals who are potentially radicalized to one or more terrorist ideologies based on their postings on social media and other Web fora. The dissertation also discusses how the ontology can be queried using intuitive structured query languages to infer triggering events in the news. The prototype system is evaluated in the context of classification and is shown to provide state of the art results. The main outputs of this research are (1) an ontological model of terrorist ideologies (2) an information extraction framework capable of identifying and extracting terrorist ideologies from text, (3) a classification methodology for classifying Web content as resonating the ideology of one or more terrorist groups and (4) a methodology for rapidly identifying news content of relevance to one or more terrorist groups.

Moderators of online discussion forums often struggle with controlling extremist comments on their platforms. To help provide an efficient and accurate tool to detect online toxicity, we apply word2vec’s Skip-Gram embedding vectors, Recurrent Neural Network models like Bidirectional Long Short-term Memory to tackle a toxic comment classification problem with a labeled dataset from Wikipedia Talk Page. We explore different pre-trained embedding vectors from larger corpora. We also assess the class imbalance issues associated with the dataset by employing sampling techniques and penalizing loss. Models we applied yield high overall accuracy with relatively low cost.

The growing use of the Internet as a major means of communication has led to the formation of cyber-communities, which have become increasingly appealing to violent extremists due to the unregulated nature of Internet communication. Online communities enable violent extremists to increase recruitment by allowing them to build personal relationships with a worldwide audience capable of accessing uncensored content. This research presents methods for identifying and forecasting the recruitment activities of violent groups within extremist social media websites. Specifically, these methods employ techniques within supervised learning and natural language processing for automatically: (1) identifying forum posts intended to recruit new violent extremist members, and (2) forecasting recruitment efforts by tracking changes in an online community’s discussion over time. We used data from the western jihadist website Ansar AlJihad Network, which was compiled by the University of Arizona’s Dark Web Project. Multiple judges manually annotated a sample of these data, marking 192 randomly sampled posts as recruiting (Yes) or non-recruiting (No). We observed significant agreement between the judges’ labels; the confidence interval of Cohen’s kappa was (0.5,0.9) at p=0.01. We used naive Bayes models, logistic regression, classification trees, boosting, and support vector machines (SVM) to classify the forum posts in a 10-fold cross-validation experimental setup. Evaluation with receiver operating characteristic (ROC) curves show that our SVM classifier achieves 89% area under the curve (AUC), a significant improvement over the 63% AUC performance achieved by our simplest naive Bayes model (Tukey’s test at p=0.05). The forecasting task uses a time series regression analysis to model the daily count of extremist recruitment posts. Evaluation with mean absolute scaled error (MASE) shows that employing latent topics as predictors can reduce forecast error compared to a naive (random-walk) model and the baseline time series model. To our knowledge, these are the first results reported on these tasks, and our analysis indicates that automatic detection and forecasting of online terrorist recruitment are feasible tasks. This research could ultimately help identify the impact of violent organizations, like terrorist groups, within the social network of an online community. There are also a number of important areas of future work including classifying non-English posts and measuring how recruitment posts and current events change membership numbers over time.

Criminologists have generally agreed that the Internet is not only a tool or resource for right-wing extremists to disseminate ideas and products, but also a site of important identity work, accomplished interactively through the exchange of radical ideas. Online discussion forums, amongst other interactive corners of the Web, have become an essential conduit for the radical right to air their grievances and bond around their “common enemy.” Yet overlooked in this discussion has been a macro-level understanding of the radical discussions that contribute to the broader collective identity of the extreme right online, as well as what constitutes “radical posting behavior” within this context. Drawing from criminal career measures to facilitate this type of analysis, data was extracted from a sub-forum of the most notorious white supremacy forum online, Stormfront, which included 141,763 posts made by 7,014 authors over approximately 15 years. In study one of this dissertation, Sentiment-based Identification of Radical Authors (SIRA), a sentiment analysis-based algorithm that draws from traditional criminal career measures to evaluate authors’ opinions, was used to identify and, by extension, assess forum authors’ radical posting behaviors using a mixed-methods approach. Study two extended on study one by using SIRA to quantify authors’ group-level sentiment about their common enemies: Jews, Blacks, and LGBTQs. Study three further extended on studies one and two by analyzing authors’ radical posting trajectories with semi-parametric group-based modeling. Results highlighted the applicability of criminal career measures to study radical discussions online. Not only did this mixed-methods approach provide theoretical insight into what constitutes radical posting behavior in a white supremacy forum, but it also shed light on the communication patterns that contribute to the broader collective identity of the extreme right online.