Enhancing Phishing Detection: A Machine Learning Approach with Feature Selection and Deep Learning Models

Abstract

Phishing attacks continue to pose significant threats to online security, with the potential to steal sensitive user data and compromise systems. To address this issue, this project introduces a machine learning-based approach for enhancing phishing detection by integrating feature selection techniques and deep learning models. The project first utilizes a dataset of phishing and legitimate URLs, containing various features such as URL length, number of dots, and subdomain levels. Through feature selection methods like SelectKBest and mutual information, the most relevant features are identified, which improve the model’s ability to distinguish between phishing and legitimate sites. Subsequently, Multi- Layer Perceptron (MLP) models are employed to learn complex relationships between the features and accurately classify URLs as phishing or legitimate. The MLP architecture consists of multiple layers of neurons, enabling it to model non- linear patterns in the data, which enhances its performance for phishing detection. The proposed model is trained and evaluated using standard performance metrics, including accuracy, precision, recall, and F1 Score. Experimental results demonstrate the effectiveness of combining feature selection with MLP-based deep learning for phishing detection, yielding high detection accuracy and low false-positive rates. This method offers an enhanced solution for real-time phishing threat detection and can be integrated into web browsers or security systems to mitigate phishing attacks and protect user data. This approach underscores the potential of machine learning and deep learning in cybersecurity, paving the way for more adaptive and efficient phishing-detection systems.