Is it Sarrah Rahamah? A supervised classification model to detect fake identities on Facebook within the Sudanese community

• Published in: Personal and ubiquitous computing Vol. 27; no. 1; pp. 107 - 118
• Main Author: Elhussein, Mariam
• Format: Journal Article
• Language: English
• Published: London Springer London 01.02.2023; Springer Nature B.V
• Subjects: Algorithms; Classification; Community; Computer Science; Decision trees; False information; Image verification; Literature reviews; Machine learning; Mobile Computing; Original Article; Personal Computing; Regression analysis; Research methodology; Social networks; Support vector machines; Ubiquitous computing; User Interfaces and Human Computer Interaction; Virtual communities; Fake identities; Fake profiles; Logistic regression; Social media; Naïve-Bayes; Sudan; Machine learning; Random forest; Support vector machine; Facebook; Decision tree; Fake accounts
• ISSN: 1617-4909; 1617-4917
• DOI: 10.1007/s00779-022-01664-2

Fake accounts that are run by fake identities are causing many problems within the Sudanese online community on Facebook. While research is focused on automatic and semi-automatic accounts, human fake accounts are often neglected. A general characterization of these accounts needs to be considered based on the cultural context they exist within. This research interviewed 250 Sudanese persons on Facebook who fell victim of eight of these accounts. Data was manually harvested for both confirmed fake accounts and confirmed real accounts. The dataset included 231 instances which was imbalanced and skewed toward the real accounts. Over-sampling with SMOTE was applied to treat the over-fitting problem of the machine learning models. Supervised classification algorithms achieved an accuracy of up to 89.7% and an AUC of 0.96 in detecting fake accounts. Furthermore, human-based methods, such as profile image verification and history, were identified by the interviewees as ways to assert the legitimacy of an account.