A big data analytic identity management expert system for social media networks

A novel concept on risk identification of social media networks using Hidden Markov Models (HMMs) based on behavioural usage patterns is proposed. Unauthorized users are identified based on characteristic activity sequences. Here, a user behavioural model is built for the social media network clients using their detailed activity logs comprising of usage patterns over years. Thereafter, incorporating an HMM, the legitimacy of users is evaluated. This followed a successful training and testing of a Neuro-Fuzzy inferencing model that uses behavioural legitimacy and login time as inputs. The HMM user behavioural model alone gives an accuracy of 73.61% in detecting authorized users while, giving 71.33% accuracy in detecting unauthorized users. The Neuro-Fuzzy model gave an 84% accuracy to distinguish legitimate users while providing 62.5% accuracy to identify illegitimate users. This gives an overall accuracy of detecting valid users (legitimate or illegitimate) of 76.85% and a false positive rate of 10.65% and a false negative rate of 12.5% A Big Data Analytic model can be developed using this concept for user behaviour analysis at the social media network service provider's end as a dynamically evolving system. To our knowledge, this work has invented a novel mechanism for monitoring user behaviour in social media networks without violating their privacy rights.