Uninformed Adversarial Community Detection

Abstract

An increasingly popular branch of graph theory is the concept of community detection. However, the field of adversarial community detection currently has very little scientific literature. Initial tests suggest that an attacker in an adversarial community detection situation may not need specific information about a network as a whole in order to effectively mask a single community. In some cases, the attacker can achieve this with even random attacks on the network. However, some community detection algorithms are far more robust than others, and so the results of these tests vary greatly as a result.