Cyber Deception in the Internet of Battlefield Things: Techniques, Instances, and Assessments

The Internet of Battlefield Things (IoBT) is an emerging application to improve operational effectiveness for military applications. The security of IoBT is one of the more challenging aspects, where adversaries can exploit vulnerabilities in IoBT software and deployment conditions to gain insight into their state. In this work, we look into the security of IoBT from the lens of cyber deception. First, we formulate the IoBT domain as a graph learning problem from an adversarial point of view and introduce various tools through which an adversary can learn the graph starting with partial prior knowledge. Second, we use this model to show that an adversary can learn high-level information from low-level graph structures, including the number of soldiers and their proximity. For that, we use a powerful n-gram based algorithm to obtain features from random walks on the underlying graph representation of IoBT. Third, we provide microscopic and macroscopic approaches that manipulate the underlying IoBT graph structure to introduce uncertainty in the adversary’s learning. Finally, we show our approach’s effectiveness through analyses and evaluations.