Email Phishing Attack Detection Using Machine Learning and Deep Learning
Abstract
Email phishing is a major cybersecurity threat in which attackers impersonate trusted individuals or organizations to trick users into revealing passwords, banking credentials and personal data. Traditional filters based on blacklists and fixed rules often miss new phishing messages that change their wording, links or sender identity. This paper proposes a comparative framework for phishing email detection using Machine Learning (ML) and Deep Learning (DL). The framework analyzes the sender, subject, message body, URLs and linguistic cues of each email. Logistic Regression, Support Vector Machine, Random Forest and Naive Bayes are compared with Long Short-Term Memory (LSTM) and transformer-based models under one dataset, one preprocessing pipeline and one held-out test set. Performance is to be assessed with accuracy, precision, recall, F1-score and confusion matrices, with particular attention to false negatives and false positives. The system labels an email as legitimate or phishing and reports a risk score together with the indicators that influenced the decision. This paper presents the design of the framework and its evaluation protocol. No experimental results are reported here, and the effectiveness of the approach must be confirmed through experiments on representative data