Jul 2026· Journal of Computer Virology and Hacking Techniques· Vol 22· 0 citations· 51 references
Computer Science
TL;DR
A hierarchical binary classification framework that decomposes the multiclass task into a sequence of structured binary decisions and provides competitive and feature-efficient classification performance is proposed.
Malicious URLs remain a persistent attack vector for phishing, malware delivery and web defacement, but operational filters often reduce the problem to binary benign/malicious detection. This paper presents X-URL, a mathematically grounded explainable framework for multi-class malicious URL classification. The framework combines quantitative feature analysis (QFA), Shannon-entropy and information-gain-based priority coefficients, character-level bidirectional long short-term memory (BiLSTM), Bayesian hyperparameter optimization and SHAP-based explanation. The learning objective is formulated as weighted multi-class cross-entropy with class-imbalance coefficients, while priority feature fusion maps normalized lexical and structural indicators into an information-theoretic weighted vector. X-URL classifies URLs into benign, defacement, malware and phishing classes and produces analyst-readable reason codes. Experiments were conducted on 651,191 URLs using a stratified 80/20 split; the independent test set contained 130,236 samples. X-URL achieved 97.73% accuracy, 97.71% weighted precision, 97.73% weighted recall and 97.72% weighted F1-score. Additional analysis shows a macro F1-score of 96.68%, a balanced accuracy of 96.33%, and an approximate 95% confidence interval of 97.65-97.81% for test accuracy. Ablation results indicate that BiLSTM improves accuracy by 3.26 percentage points over vanilla LSTM, and QFA adds another 3.23 points with only 0.4 ms extra latency per URL. Algorithmic analysis shows that inference complexity is linear in URL length and quadratic in the BiLSTM hidden size, O(T H(H+d)+m), making X-URL suitable for high-throughput web filtering when combined with caching and slow-path enrichment..
Bao Minh Tran· Journal of Intelligent Decis...· 0 citations
Experimental evaluation demonstrates that the Hybrid CNN–LSTM model effectively classifies webpages as Safe, Suspicious, or Malicious, providing improved detection accuracy and faster prediction compared with conventional machine learning approaches.
Sinchana and Dr. Kruthi R· International Journal of Adv...· 0 citations
The internet has rapidly evolved in its communication, commerce and information sharing making it a huge platform for cyber threats, particularly malicious URLs. They pose a serious threat to individuals and to organisations. Phishing attacks, malware distribution and other types of cybercrime frequently are carried out through malicious URLs. In this research, we have created and tested the machine learning models to detect malicious URLs. The labeled URLs used were obtained from a public dataset with more than 651,000 labeled URLs. The dataset was prepared for classification by applying data pre-processing techniques like stratified sampling, label encoding and Term Frequency–Inverse Document Frequency (TF-IDF) vectorization. To train and test the algorithms, five machine learning were used: Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Naïve Bayes (NB), Random Forest (RF) and Extreme Gradient Boosting (XGBoost), which were trained and evaluated by the metrics of accuracy, precision, recall and F1 score. The results indicated that the Random Forest model had the highest classification accuracy (95%) as compared to the other models. Moreover, a web based malicious URL detection system was developed to demonstrate the actual application of the developed models in real time cyber security scenarios. The study provides a conclusion that the machine learning techniques, particularly ensemble learning techniques can be considered an effective and reliable technique to detect malicious URL.
Oludele Adeleke, Jimoh Abdulhakeem Kuranga, Samuel Adeolu Ogunbiyi et al.· American Journal of Interdis...· 0 citations
A feature-driven framework for phishing Uniform Resource Locator (URL) detection is introduced, emphasizing the design and evaluation of enhanced feature representations and highlighting that performance gains are primarily driven by feature design rather than model complexity.
Deniz Kaya, Murat Osmanoğlu· PeerJ Computer Science· 0 citations
The high rate of internet service expansion has posed a great challenge due to a high number of malicious sites that conduct internet activities that include phishing, malware downloads and online fraud. These threats must be detected at the initial stage and improve security to make people resistant to possible threats. This paper introduces a machine learning-based system to identify malicious websites based on the URL-based features. The suggested method is aimed at deriving both lexical and structural attributes of URLs, such as URL length, the use of special characters, subdomains and protocols. They are the characteristics based on which classification models that can differentiate between benign and malicious URLs are trained. A smart detection system, which is known as the Adaptive URL Feature Intelligence Algorithm (AUFIA), is proposed to enhance analyzing features with the use of feature weighting and normalization algorithms. The experimental assessment proves that the given approach has high detection accuracy and better classification performance than some of the existing methods. The results suggest that feature engineering and machine learning methods are effective in offering scalable and effective solutions to the problem of malicious websites.
Sampath Mnv, Nahita Pathania· 2026 4th International Confe...· 0 citations
Abstract— Phishing website detection using machine learning focuses on the design and implementation of an intelligent system for detecting malicious URLs using machine learning techniques. The system aims to classify URLs as either legitimate or malicious by analyzing various structural and statistical features extracted from the URLs. A dataset containing both benign and malicious URLs is used to train and evaluate the model. The proposed approach utilizes a Gradient Boosting Classifier due to its high accuracy and ability to handle complex patterns in data. Feature extraction plays a crucial role in the project, where attributes such as URL length, presence of special characters, domain age, use of HTTPS, and abnormal patterns are considered. These features are fed into the model, which learns to differentiate between safe and harmful URLs. The project involves several stages, including data collection, preprocessing, feature extraction, model training, and performance evaluation. Multiple machine learning algorithms such as Support Vector Machine (SVM), Decision Tree, Random Forest, and XG Boost are also explored and compared to identify the most effective model .The system is designed to work in real time, allowing users to input URLs and receive instant predictions regarding their safety.
Keywords— Phishing Website Detection, Machine Learning, Gradient Boosting Classifier, URL Feature Extraction, Cybersecurity, Malicious URL Detection, Web Security, Classification, Feature Engineering, Real-Time Detection.
V. B, K. Subba Rao, G. Prasad· International Scientific Jou...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
