Evaluasi Efektivitas Penggunaan FastText Embedding dan LSTM Networks dalam Deteksi Phishing Email

Abstract

This research evaluates the effectiveness of using FastText embedding and Long Short-Term Memory (LSTM) neural network model in detecting phishing emails. Phishing emails are a rapidly growing cyber threat and require accurate detection methods. FastText is used to represent words numerically, capturing sub-word information and morphological variations, while LSTM handles sequential data to recognize patterns in text. The research uses a secondary dataset from Kaggle and involves a series of stages, from data preprocessing (tokenization, text cleaning, and embedding), LSTM model training, to performance evaluation using metrics such as accuracy, precision, recall, and F1-Score. Experimental results show that the combination of FastText and LSTM is able to detect phishing with an accuracy rate of 92%. However, the model still faces challenges in handling class imbalance as well as high computational requirements. Further optimizations, such as model structure adjustment and data augmentation techniques, are suggested to improve the performance. This research provides insights into the development of automated solutions to efficiently detect phishing and is expected to contribute to the improvement of cybersecurity.