Manufactured Narratives: On the Potential of Manipulating Social Media to Politicize World Events

Social media platforms are as popular as ever. Billions of users worldwide log in daily to one or more of them to connect with friends, share their day-to-day experiences, keep up with the news, and scout potential employers/employees. News organizations, either directly or through the help of citizen journalists, regularly cite posts and comments made on social media for news stories. A subset of these stories is focused on individuals responsible for radical, sometimes violent actions. It is not uncommon for stories to appear on news platforms after a citizen has committed a crime, attempting to link their posts to their actions, in order to ascertain how to best establish a baseline of radical behavior online, and prevent violent extremism in the future. Things, however, are not always as simple. In this paper, we look into the previously ignored possibility that the posts cited are not genuine, or that the profile linked to an individual does not necessarily belong to them altogether. We select 11 social media platforms, and show that by performing seemingly innocuous actions on them, we can craft a fake narrative meant to disinform. We propose two distinct threat models, each tied to a different type of adversarial behavior, and measure the susceptibility of each social media platform across 18 dimensions, through custom experiments mapping on- and off-platform behavior to these threat models. Through these experiments, we find that 10 out of the 11 social media platforms are susceptible to our threat models, and that the majority of them (8/11) embed links in posts in a way that allows attackers to change the final destinations of these links, reframing posts and ultimately manufacturing narratives.