Graph embeddings have gained huge popularity in the recent years as a powerful tool to analyze social networks. However, no prior works have studied potential bias issues inherent within graph embedding. In this paper, we make a first attempt in this direction. In particular, we concentrate on the fairness of node2vec, a popular graph embedding method. Our analyses on two real-world datasets demonstrate the existence of bias in node2vec when used for friendship recommendation. We, therefore, propose a fairness-aware embedding method, namely Fairwalk, which extends node2vec. Experimental results demonstrate that Fairwalk reduces bias under multiple fairness metrics while still preserving the utility.

Social Networks and other social media systems are an ever popular medium that allow users to freely communicate and interact with their peers. Once a user shares a piece of information, however, the transitive propagation of information in such systems can allow this information to spread quickly throughout the whole system. Due to the potentially sensitive nature of the shared information, users naturally have an interest in controlling the propagation of information to ensure privacy. At the same time, users also have utility requirements in terms of users they want to share a certain piece of information with, which naturally causes a conflict with the privacy requirements.,,In this paper, we tackle the issue of controlling the propagation of information through a social network while at the same time maintaining utility requirements set by the user. We leverage continuous-time diffusion networks to model the global propagation behavior in social networks and define combined privacy and utility policies that allow us to enforce privacy under utility restrictions, and vice versa. We show that optimally satisfying such policies corresponds to solving a constrained submodular minimization problem, which, while NP-hard, allows for a constant factor approximation due to the structure of our objective function.