A Comparative Study of Machine Learning and Neural Network Models for Phishing Detection

Phishing remains one of the most prevalent cybersecurity threats, particularly within communication systems, as it exploits email platforms to deceive users into disclosing sensitive information. This paper presents a comprehensive comparison of traditional machine learning (ML) models and advanced neural network (NN) architectures for phishing email detection. We evaluate models including Naive Bayes, Logistic Regression, Decision Trees, Random Forests, Recurrent Neural Networks (RNN), Long Short-Term Memory (LSTM), and Gated Recurrent Units (GRU) based on their ability to identify phishing attempts from sequential email data. Our study analyzes the trade-offs between model complexity, computational cost, and accuracy, focusing on scalability and generalization to real-world phishing attacks. Results show that while neural networks, particularly LSTM and GRU, can effectively capture complex patterns and dependencies in email content, simpler ML models such as Stochastic Gradient Descent (SGD) achieve competitive accuracy with significantly lower computational overhead. This balance between performance and resource efficiency makes ML models particularly suitable for large-scale, real-time phishing detection systems. The findings of this research offer valuable insights for implementing robust, adaptive, and intelligent phishing detection in secure communication environments.