Jadwal Sholat

Memuat jadwal sholatโ€ฆ

Ilmu Komputer & AI editorial

Open AccessOA2023

Phishing Detection System Through Hybrid Machine Learning Based on URL

This study proposes a hybrid machine learning model (LSD) that combines logistic regression, support vector machine, and decision tree with soft and hard voting to detect phishing URLs. Using a dataset of over 11,000 websites and canopy feature selection, the model outperforms individual classifiers in accuracy, precision, recall, -score, and specificity.
Abdul Karim; Mobeen Shahroz; Khabib Mustofa; S. Belhaouari; S. Jogaยท IEEE Accessยท 2023ยท DOI 10.1109/ACCESS.2023.3252366

The core problem

Phishing attacks, first identified in 1996, have become one of the most severe and dangerous forms of cybercrime on the internet. These attacks typically use email distortion and fake websites to deceive individuals into revealing sensitive information. Despite numerous studies on prevention, detection, and awareness, a complete solution remains elusive. Machine learning offers a promising defense against phishing. This study focuses on phishing URL detection using a dataset of over 11,000 websites, containing both phishing and legitimate URL attributes in vector form. The goal is to develop a robust detection system that can accurately classify URLs as phishing or legitimate, thereby protecting users from such attacks.

Innovation

The proposed approach employs a hybrid model called LSD, which combines logistic regression (LR), support vector machine (SVC), and decision tree (DT) using soft and hard voting mechanisms. The dataset was preprocessed and then used to train several machine learning models: decision tree (DT), linear regression (LR), random forest (RF), naive Bayes (NB), gradient boosting classifier (GBM), K-neighbors classifier (KNN), support vector classifier (SVC), and the proposed LSD model. Feature selection was performed using the canopy technique with cross-fold validation, and hyperparameter optimization was done using Grid Search. Evaluation metrics included precision, accuracy, recall, -score, and specificity. The hybrid model leverages the strengths of individual classifiers to improve overall detection performance.
Introduction
Phishing attacks, first identified in 1996, have become one of the most severe and dangerous forms of cybercrime on the internet. These attacks typically use email distortion and fake websites to deceive individuals into revealing sensitive information. Despite numerous studies on prevention, detection, and awareness, a complete solution remains elusive. Machine learning offers a promising defense against phishing. This study focuses on phishing URL detection using a dataset of over 11,000 websites, containing both phishing and legitimate URL attributes in vector form. The goal is to develop a robust detection system that can accurately classify URLs as phishing or legitimate, thereby protecting users from such attacks.

Why it matters

The study confirms that hybrid machine learning models can significantly improve phishing URL detection compared to single classifiers. The LSD model's success is attributed to the complementary strengths of its constituent algorithms: logistic regression for probabilistic outputs, support vector machine for high-dimensional space handling, and decision tree for interpretability. The soft and hard voting mechanisms allow the model to make robust decisions by aggregating predictions. However, the study acknowledges limitations such as the dataset's size and the need for real-time detection. Future work could explore deep learning approaches and larger, more diverse datasets to further enhance detection capabilities. Overall, the proposed approach provides a practical and efficient solution for defending against phishing attacks.

Who should read this

Pembaca Research Digest dan praktisi rumpun ini.

Opening member contentโ€ฆ