David M. Beskow

David M. Beskow, Kathleen M. Carley

Social bots have emerged over the last decade, initially creating a nuisance while more recently used to intimidate journalists, sway electoral events, and aggravate existing social fissures. This social threat has spawned a bot detection algorithms race in which detection algorithms evolve in an attempt to keep up with increasingly sophisticated bot accounts. This cat and mouse cycle has illuminated the limitations of supervised machine learning algorithms, where researchers attempt to use yesterday's data to predict tomorrow's bots. This gap means that researchers, journalists, and analysts daily identify malicious bot accounts that are undetected by state of the art supervised bot detection algorithms. These analysts often desire to find similar bot accounts without labeling/training a new model, where similarity can be defined by content, network position, or both. A similarity based algorithm could complement existing supervised and unsupervised methods and fill this gap. To this end, we present the Bot-Match methodology in which we evaluate social media embeddings that enable a semi-supervised recursive nearest neighbors search to map an emerging social cybersecurity threat given one or more seed accounts.

Thomas Magelinski, David Beskow, Kathleen M. Carley

Neural networks are increasingly used for graph classification in a variety of contexts. Social media is a critical application area in this space, however the characteristics of social media graphs differ from those seen in most popular benchmark datasets. Social networks tend to be large and sparse, while benchmarks are small and dense. Classically, large and sparse networks are analyzed by studying the distribution of local properties. Inspired by this, we introduce Graph-Hist: an end-to-end architecture that extracts a graph's latent local features, bins nodes together along 1-D cross sections of the feature space, and classifies the graph based on this multi-channel histogram. We show that Graph-Hist improves state of the art performance on true social media benchmark datasets, while still performing well on other benchmarks. Finally, we demonstrate Graph-Hist's performance by conducting bot detection in social media. While sophisticated bot and cyborg accounts increasingly evade traditional detection methods, they leave artificial artifacts in their conversational graph that are detected through graph classification. We apply Graph-Hist to classify these conversational graphs. In the process, we confirm that social media graphs are different than most baselines and that Graph-Hist outperforms existing bot-detection models.