Evading Community Detection Via Counterfactual Neighborhood Search

Community detection techniques are useful for social media platforms todiscover tightly connected groups of users who share common interests. However,this functionality often comes at the expense of potentially exposingindividuals to privacy breaches by inadvertently revealing their tastes orpreferences. Therefore, some users may wish to preserve their anonymity and optout of community detection for various reasons, such as affiliation withpolitical or religious organizations, without leaving the platform. In thisstudy, we address the challenge of community membership hiding, which involvesstrategically altering the structural properties of a network graph to preventone or more nodes from being identified by a given community detectionalgorithm. We tackle this problem by formulating it as a constrainedcounterfactual graph objective, and we solve it via deep reinforcementlearning. Extensive experiments demonstrate that our method outperformsexisting baselines, striking the best balance between accuracy and cost.