Lightweight Deep Learning Model and Genetic Algorithm Based Optimal Phishing Website Detection

Phishing attacks, exploiting human vulnerabilities to steal sensitive information, pose a persistent threat in cybersecurity. Traditional detection methods, often computationally intensive, struggle to keep pace with evolving cybercriminal tactics. Our study presents a novel detection approach using a genetic algorithm for optimal feature selection and a lightweight deep learning model for classification. Leveraging the DEAP library, the algorithm reduced 31 features to 9 critical ones, boosting model efficiency. The model achieved high precision rates - 0.97 for normal and 0.81 for phishing sites - highlighting the effectiveness of integrating genetic algorithms with deep learning for enhanced, efficient phishing detection.