Real or Fake Identity Deception of Social Media Accounts using Recurrent Neural Network

Abstract

Identity fraud is a widespread issue across online social networks in recent days. Current research effort is directed to develop technologies to detect identity fraud. The effectiveness of the existing strategies is uncertain. We describe a study of detecting identity fraud by using clustering and classification techniques. We define traditional methodological shortcomings in detection of identity fraud for these methods and suggest ways that can enhance their efficacy in real-world contexts. Initially, we collect data from social media accounts and applied preprocessing and filtration techniques like Natural Language Process (NLP), vectorization, dimensionality reduction, data normalization, etc. Features are extracted, based on the behavioral analysis, and characteristics of each profile. The clustering approaches are used to detect each profile, either real or fake, and similar approach has been carried out for deep learning classification. The Recurrent Neural Network (RNN) has been used to categorize each profile based on module training and testing. In the experimental analysis, we show the system's effectiveness when applied in the real-world social media environment.