Detection Of E-mail Phishing Attacks Using Machine Learning And Deep Learning
Kanchan Waghmare
, Prof.S.Khan
Email phishing detection; machine learning; deep learning; LSTM; convolutional neural network; random forest; SVM; ensemble classification; feature engineering; cybersecurity; real-time detection; NLP.
Malicious email campaigns that impersonate trusted entities to harvest credentials, disseminate malware, and perpetrate financial fraud continue to impose severe economic and reputational damage on individuals and organizations worldwide. Conventional rule-based and blacklist-driven countermeasures demonstrate an inherent inability to generalize beyond catalogued threat signatures, leaving systems unprotected against emergent and zero-day phishing variants. This paper introduces PhishGuard AI, a novel hybrid ensemble framework that synergistically combines four classical Machine Learning (ML) classifiers — Random Forest (RF), Support Vector Machine (SVM) with Radial Basis Function kernel, Decision Tree (DT), and Logistic Regression (LR) — with two Deep Learning (DL) architectures — Long Short-Term Memory (LSTM) and one-dimensional Convolutional Neural Network (CNN). A structured pipeline extracts twenty-six discriminative features spanning five semantic dimensions: URL structural patterns, HTML document composition, email content semantics, sender behavioral metadata, and textual statistical properties. Individual model probability estimates are aggregated through a calibrated arithmetic ensemble mechanism to yield a continuous phishing risk score, which drives a three-tiered policy enforcement layer (LOW, MEDIUM, HIGH) integrated directly into a production Flask web application. Experiments conducted on a benchmark dataset of 18,650 labeled email records yield 97.80% accuracy, 98.10% precision, 97.50% recall, and 0.9982 area under the ROC curve for the Random Forest component; the LSTM model achieves 95.60% accuracy. The ensemble reduces the total misclassification rate to 2.07%, outperforming all constituent single-model baselines. The real-time operational deployment, comprehensive phishing log database, and interpretable alert generation distinguish this system from prior research-only evaluations, addressing a critical gap in practical cybersecurity tooling.
"Detection Of E-mail Phishing Attacks Using Machine Learning And Deep Learning", IJSDR - International Journal of Scientific Development and Research (www.IJSDR.org), ISSN:2455-2631, Vol.11, Issue 5, page no.c192-c197, May-2026, Available :https://ijsdr.org/papers/IJSDR2605227.pdf
Volume 11
Issue 5,
May-2026
Pages : c192-c197
Paper Reg. ID: IJSDR_310095
Published Paper Id: IJSDR2605227
Downloads: 000266
Research Area: Science and Technology
Country: Udgir, Maharashtra, India
ISSN: 2455-2631 | IMPACT FACTOR: 9.15 Calculated By Google Scholar | ESTD YEAR: 2016
An International Scholarly Open Access Journal, Peer-Reviewed, Refereed Journal Impact Factor 9.15 Calculate by Google Scholar and Semantic Scholar | AI-Powered Research Tool, Multidisciplinary, Monthly, Multilanguage Journal Indexing in All Major Database & Metadata, Citation Generator
Publisher: IJSDR(IJ Publication) Janvi Wave
