Fraudulent account detection in social media using hybrid deep transformer model and hyperparameter optimization.

Area of Science:

• Computer Science
• Artificial Intelligence
• Cybersecurity

Background:

• Social media growth fuels fake account proliferation, risking user privacy and platform integrity.
• Detecting fake accounts is challenging due to imbalanced, high-dimensional, and sequential user activity data.
• Existing methods struggle with complex patterns and risk overfitting, necessitating advanced detection models.

Purpose of the Study:

• To propose a novel deep learning architecture for scalable and precise social media fraud detection.
• To address data imbalance and dimensionality challenges in fake account detection.
• To optimize model performance and efficiency through advanced techniques.

Main Methods:

• A deep learning architecture integrating Temporal Convolutional Network (TCN) for sequence modeling.
• Generative Adversarial Network (GAN)-based data augmentation to address class imbalance.
• Autoencoder for dimensionality reduction and Seagull Optimization Algorithm (SOA) for hyperparameter tuning.

Main Results:

• The TCN-GAN-SOA framework achieved high performance on benchmark datasets (Cresci-2017, TwiBot-22).
• Achieved ROC-AUC scores of 0.96 and 0.95, with superior precision-recall and F1-scores compared to state-of-the-art models.
• Demonstrated computational efficiency and robustness in handling diverse fraudulent behaviors.

Conclusions:

• The proposed framework offers a scalable, reliable, and accurate solution for social media fraud detection.
• The integration of TCN, GAN, and SOA provides a powerful approach to combatting fake accounts.
• This methodology enhances the integrity and trustworthiness of social media platforms.