Detecting Adversarial BENDs in the Information Environment: Phase 2 Proposal

Social cybersecurity is an emerging subdomain of national security that will affect all levels of future warfare, both conventional and unconventional, with strategic consequences. Social cybersecurity isĀ ōan emerging scientific area focused on the science to characterize, understand, and forecast cyber-mediated changes in human behavior, social, cultural and political outcomes, and to build the cyber-infrastructure needed for society to persist in its essential character in a cyber-mediated information environment under changing conditions, actual or imminent social cyber-threatsö (Carley et al., 2018). ĀCataclysmic changes in how we communicate in cyberspace are dramatically altering our society, allowing information to spread faster, farther, and with less assurance of its accuracy, and enabling groups to form and recruit members on-line to foster social divides. AI technology within this information environment is enabling both state and non-state actors to manipulate the global marketplace of beliefs and ideas at unprecedented speed, with global impact.Ā This is changing the battlefield at all levels of war. Influence campaigns, often supported by AI technology referred to as bots or cyborgs, are used to sever trust in institutions, create fractures in society, polarize groups, and generate mass-hysteria, to alter elections and subvert policies, and to support criminal and terror activities. Identifying these influence campaigns, in particular the maneuvers used to conduct them, and the impact that they have on the individual or group being targeted requires reasoning about the information environment in which teams of humans and AI work together.Ā It requires sifting through and analyzing vast quantities of data.Ā We propose to address these problems by creating a novel technology for influence campaign assessment that supports identification of maneuvers, their source, their target, and their impact.Ā These technologies will employ multi-modal reasoning, and draw on advances in dynamic network assessment and stance identification. This proposal is to develop, test, and make operational a scientifically sound approach for assessing information influence campaigns using the BEND framework in both a web-based and server-based form. This novel technology will use the joint social networks and knowledge networks implicit in social media communications and process these using social network informed machine learning techniques.Ā The end result will be scalable solutions that can be applied to posts in a large variety of languages.