A Bi-directional Gated Recurrent Unit and Long Short-Term Memory based Fake Profile Identification System

Nowadays, fake profiles are programmatically created online individualities that mislead and influence others, which are done by frequently using stolen photos, fraudulent information, or fictional particulars. Hence, these fake profiles are used for malicious activities such as phishing, social engineering, catfishing, and scattering misinformation. The expansion of fake profiles is increased with triggering more damage than any other cyber-attacks. The solution has been recommended in various ways to detect fake profiles, but the existing models have more overfitting problems. To overcome these issues, a combined model of Bi-directional Gated Recurrent Unit (BiGRU) and Long Short-Term Memory (LSTM) algorithms is proposed. Initially, data preprocessing is done on TwiBot-20 dataset with Natural Language Processing (NLP) and tokenization to remove unnecessary characters. Next, a word embedding technique name called GloVe is used to extract the optimal features, Finally, from the extracted features, the proposed BiGRU-LSTM model gives a precise rate in detecting and classifying fake profiles as human and bot generated. The proposed BiGRU-LSTM gives better results than the other compared existing methods Convolutional Neural Network (CNN) in terms of accuracy, precision, recall, and f1 score of $99.73 \%, 98.23 \%, 99.56 \%$, and $99.63 \%$ respectively.