urlbert-tiny-v4 Open-source Malicious Website Classifier - Quickly and Accurately Identify Four Types of Websites, Including Benign and Phishing Ones

Urlbert Tiny V4 Malicious Url Classifier

Developed by CrabInHoney

A lightweight BERT model, specifically fine-tuned for URL classification tasks, capable of categorizing URLs into four types: benign, phishing, malware, and tampered.

Text Classification

Transformers

Supports Multiple LanguagesOpen Source License:Apache-2.0 #Lightweight BERT #URL security detection #Phishing identification

Downloads 54

Release Time : 3/30/2025

Model Overview

This is a lightweight model based on the BERT architecture, specifically designed for malicious URL detection and classification. It efficiently and accurately identifies whether a URL is malicious and distinguishes between different types of malicious URLs (phishing, malware, tampered).

Model Features

Lightweight and efficient

The model has only 3.69 million parameters, compact yet high-performing, suitable for deployment in resource-limited environments.

High accuracy

Achieves an overall accuracy of 99.22% on the test set, with excellent F1 scores across all categories.

Multi-category classification

Can distinguish between four types of URLs: benign, phishing, malware, and tampered.

Specialized optimization

Specifically optimized and fine-tuned for URL text features.

Model Capabilities

Malicious URL detection

URL classification

Cybersecurity analysis

Phishing website identification

Malware URL identification

Website tampering detection

Use Cases

Cybersecurity

Enterprise cybersecurity protection

Integrated into corporate cybersecurity systems to automatically block malicious URLs.

Effectively reduces the risk of employees accessing malicious URLs.

Browser security plugin

Serves as the core detection engine for browser extensions.

Provides real-time warnings for potentially dangerous URLs accessed by users.

Email security filtering

Detects malicious links in emails.

Prevents phishing attacks from spreading via email.

Security research

Threat intelligence analysis

Batch analysis of malicious URLs in URL libraries.

Rapidly identifies and categorizes large volumes of suspicious URLs.

🚀 URLBERT-Tiny-v4 Malicious URL Classifier

This is a lightweight version of BERT, specifically fine-tuned for classifying URLs into four categories: benign, phishing, malware, and defacement. It provides an efficient solution for identifying malicious URLs in the field of cybersecurity.

🚀 Quick Start

The URLBERT-Tiny-v4 model is designed to classify URLs into four distinct categories: benign, phishing, malware, and defacement. It offers a lightweight and efficient solution for detecting malicious URLs in the realm of cybersecurity.

✨ Features

Lightweight Design: With only 3.69M parameters, it's highly efficient.
High Accuracy: Achieves excellent classification results on the test set.
Multi - category Classification: Capable of classifying URLs into four different threat levels.

📦 Installation

To use this model, you need to install the transformers library. You can install it using the following command:

pip install transformers

📚 Documentation

Model Details

Property	Details
Model Type	URL classification model
Model size	3.69M parameters
Tensor type	F32
Model weight size	14.8 MB
Base model	CrabInHoney/urlbert-tiny-base-v4
Training Data	Malicious URLs Dataset

Model Evaluation Results

The model was evaluated on a test set with the following classification metrics:

Metric	Model V3	Model V4 (this model)
Overall Accuracy	0.9837	0.9922
F1 - score (Benign)	0.9907	0.9955
F1 - score (Defacement)	0.9937	0.9984
F1 - score (Malware)	0.9741	0.9845
F1 - score (Phishing)	0.9444	0.9734
Weighted Average F1 - score	0.9836	0.9922

💻 Usage Examples

Basic Usage

Below is an example of how to use the model for URL classification using the Hugging Face transformers library:

from transformers import BertTokenizerFast, BertForSequenceClassification, pipeline
import torch

# Определение устройства (GPU или CPU)
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print(f"Используемое устройство: {device}")

# Загрузка модели и токенизатора
model_name = "CrabInHoney/urlbert-tiny-v4-malicious-url-classifier"
tokenizer = BertTokenizerFast.from_pretrained(model_name)
model = BertForSequenceClassification.from_pretrained(model_name)
model.to(device)

# Создание pipeline для классификации
classifier = pipeline(
    "text-classification",
    model=model,
    tokenizer=tokenizer,
    device=0 if torch.cuda.is_available() else -1,
    return_all_scores=True
)

# Примеры URL для тестирования
test_urls = [
    "wikiobits.com/Obits/TonyProudfoot",
    "http://www.824555.com/app/member/SportOption.php?uid=guest&langx=gb",
]

# Маппинг меток на понятные названия классов
label_mapping = {
    "LABEL_0": "benign",
    "LABEL_1": "defacement",
    "LABEL_2": "malware",
    "LABEL_3": "phishing"
}

# Классификация URL
for url in test_urls:
    results = classifier(url)
    print(f"\nURL: {url}")
    for result in results[0]: 
        label = result['label']
        score = result['score']
        friendly_label = label_mapping.get(label, label)
        print(f"{friendly_label}, %: {score:.4f}")

Example Output:

URL: wikiobits.com/Obits/TonyProudfoot
benign, %: 0.9996
defacement, %: 0.0000
malware, %: 0.0000
phishing, %: 0.0003

URL: http://www.824555.com/app/member/SportOption.php?uid=guest&langx=gb
benign, %: 0.0000
defacement, %: 0.0001
malware, %: 0.9998
phishing, %: 0.0001

📄 License

This model is released under the Apache - 2.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご