Enhancing Fake Account Detection on Digital Platforms Using Deep Learning Models

This research addresses the growing challenge of detecting fake accounts on digital platforms, such as social media and professional networks, which pose significant threats to user trust, platform security, and public discourse. The proposed framework integrates several powerful architectures, including Attention-Based LSTM, Bidirectional GRU, Bidirectional LSTM, LSTM with Flatten Layer, and Bidirectional LSTM with Flatten Layer, each designed to capture complex patterns and dependencies in the user profile data. These models are evaluated based on key performance metrics, such as F1-score, precision, recall, accuracy, and loss value, to assess their effectiveness in fake account detection. The results show that the LSTM with Flatten Layer model achieved the highest accuracy (95.83%), precision (94%), and recall (98%), making it the most effective model in distinguishing fake profiles. The Bidirectional LSTM with Flatten Layer (AdamW) also exhibited strong results, with an accuracy of 94.91%. The results of this study contribute to the development of more robust and accurate systems for detecting fraudulent profiles, offering valuable guidance for addressing the challenges posed by fake accounts in digital environments.