Identity deception detection: requirements and a model

This paper aims to describe requirements for a model that can assist in identity deception detection (IDD) on social media platforms (SMPs). The model that was discovered demonstrates the usefulness of the requirements. The aim of the model is to identify humans lying about their identity on SMPs.

The requirements of a model for IDD will be determined through a literature study combined with a study that identifies currently available identity related metadata on SMPs. This metadata refers to the attributes that describe a user account on an SMP. The aim is to restrict IDD to be only based on these types of attributes, as opposed to or combined with the contents of a single or multiple communications.

Data science experiments were conducted and in particular supervised machine learning models were discovered that indeed detects identity deception on SMPs with an area under the receiver operator characteristics curve (ROC-AUC) of 75.5 per cent.

SMPs allow any user to easily communicate with their friends or the general public at large. People can now be targeted at great scale, most often for malicious purposes. The reality is that many of these cyber-attacks involve some form of identity deception, where the attackers lie about who they are. Much focus to date has been on the identification of non-human deceptive accounts. This paper focuses on deceptive human accounts that target vulnerable individuals on SMPs.