学术文献库

Online manipulation of information has become more prevalent in recent years as state-sponsored disinformation campaigns seek to influence and polarize political topics through massive coordinated efforts. In the process, these efforts leave behind artifacts, which researchers have leveraged to analyze the tactics employed by disinformation campaigns after they are taken down. Coordination network analysis has proven helpful for learning about how disinformation campaigns operate; however, the usefulness of these forensic tools as a detection mechanism is still an open question. In this paper, we explore the use of coordination network analysis to generate features for distinguishing the activity of a disinformation campaign from legitimate Twitter activity. Doing so would provide more evidence to human analysts as they consider takedowns. We create a time series of daily coordination networks for both Twitter disinformation campaigns and legitimate Twitter communities, and train a binary classifier based on statistical features extracted from these networks. Our results show that the classifier can predict future coordinated activity of known disinformation campaigns with high accuracy (F1 = 0.98). On the more challenging task of out-of-distribution activity classification, the performance drops yet is still promising (F1 = 0.71), mainly due to an increase in the false positive rate. By doing this analysis, we show that while coordination patterns could be useful for providing evidence of disinformation activity, further investigation is needed to improve upon this method before deployment at scale.