Secure email systems: phishing detection with cryptographic verification

The work introduces a system that combines phishing detection with secure email transmission to strengthen the security of digital communication. The proposed solution integrates a stacked machine learning model that leverages Logistic Regression, Random Forest, and Gradient Boosting, with Random Forest serving as the meta-classifier. The model identifies phishing emails by analyzing both text-based and stylometric features. To safeguard message confidentiality, the system incorporates multiple encryption methods—Advanced Encryption Standard (AES), Rivest–Shamir–Adleman (RSA), and ChaCha20—applied appropriately to email content and attachments. The implementation uses a Flask-based interface with an underlying SQLite database, offering features such as user-driven message search and secure key management. Experimental evaluation with both balanced and imbalanced datasets shows that the model achieves high detection accuracy, reaching up to 98.37%, while ensuring reliable encryption and decryption without information leakage. With its focus on both protection and usability, this system provides an effective and practical approach to securing sensitive email communication.