International Journal of Computer Applications |
Foundation of Computer Science (FCS), NY, USA |
Volume 186 - Number 66 |
Year of Publication: 2025 |
Authors: Christy Onoshokwue Isokpehi, P.O. Asagba |
![]() |
Christy Onoshokwue Isokpehi, P.O. Asagba . Fake Profile Detection and Stalking Prediction on Twitter (X) using Convolutional Neural Network (CNN). International Journal of Computer Applications. 186, 66 ( Feb 2025), 14-23. DOI=10.5120/ijca2025923771
This research is directly focused on fake profile detection and stalking prediction in twitter using convolution neural network. Twitter is a real-time social media application that has gained global popularity, and the use of twitter is also raising serious issues in the form of cyber-stalking. Stalking is a serious cyber-attack in which the attacker uses digital media to harass the victim or group through personal attacks and the disclosure of false or confidential information among other persons. Stalking is categorize as email-stalking, internet-stalking, and computer-stalking. The main challenging problems in twitter social network security is to recognize fake profiles and stalking activities by followers. These Fake profiles are a preferred means for malicious users to commit various cybercrimes such as cyber-stalking, disseminating misinformation and fake news, stigmatize someone's personality, capturing subscribers' credentials and generating malicious communications, misleading users towards counterfeit sites and impacting notoriety. This study detect fake profiles and predict stalking activities in twitter. Convolution neural network is used to train with the dataset of twitter profiles which gives the accuracy of the model. Predicted results are produced after training and evaluation of the models, which is able to distinguish between fake and real twitter profiles based on attributes like follower and friend counts, status updates, and more. The adopted methodology is Object Oriented Analysis and Design Method (OOADM). Python programming language is use for implementation. The application results show that the proposed convolution neural network model performs better prediction accuracy than other considered algorithms and correctly classified over 95% of the accounts with a low error rate.