Identity - Attribute Inference in Online Social Network(s) Using Bio-Inspired Algorithms and Machine Learning Approaches

Abstract

Twitter is one of the most popular social networking sites today, and it has become a critical tool for gathering data from numerous individuals throughout the world. The platform hosts a variety of debates spanning from current events and news to entertainment, advertising, and technology. In contrast to earlier approaches, the proposed work employs the concept of both direct (via tweets) and indirect stance detection (via homophily elements) to infer sensitive attributes. Along with attribute-based inference, the proposed work also matches user profiles across cross platforms via user-generated posts. Unlike prior efforts, usernames are not included in the feature set here since they are a bit of a giveaway. Bio-inspired algorithms are used along with ensemble methods to extract the best set of features.