⭐Open Source

Malicious URL Detection using CNN

›2019 (Final Year Project)

›Overview

A deep learning-based security solution that detects malicious URLs using Character-level Convolutional Neural Networks. This project addresses the growing threat of web-based attacks by identifying malicious patterns in URL structures.

›Challenge

Traditional signature-based detection methods struggle with new and evolving attack patterns. The challenge was to create a model that could learn to identify malicious URLs based on their character-level patterns without relying on predefined signatures.

›Solution

Implemented a Character-level CNN architecture using Python and Keras that analyzes URL strings at the character level. The model was trained on comprehensive datasets of malicious and benign URLs, learning to identify patterns indicative of XSS attacks, SQL injection attempts, and directory traversal exploits.

›Technology Stack

PythonKerasTensorFlowDeep LearningCNNMachine Learning

›Key Features

✓Character-level CNN architecture

✓Multi-class classification for different attack types

✓Training on real-world security datasets

✓URL preprocessing and feature extraction

✓Model evaluation and performance metrics

✓Python implementation with Keras/TensorFlow

✓Modular design for easy extension

✓Detailed documentation and examples

›Impact & Results

▸Effective detection of XSS, SQL injection, and directory traversal attacks

▸Model trained on comprehensive security threat datasets

▸Character-level analysis providing robustness against obfuscation

▸Open source contribution to security research

▸Academic project demonstrating ML application in cybersecurity

›Technical Highlights

◆Character-level Convolutional Neural Network

◆Deep learning with Keras and TensorFlow

◆Feature extraction from URL strings

◆Multi-label classification

◆Training pipeline with data augmentation

◆Model optimization and hyperparameter tuning

◆Evaluation metrics (accuracy, precision, recall)

◆Python implementation with NumPy and Pandas

Deployment:Open source on GitHub

¯ Project loaded successfully

✓ 8 features documented

✓ 6 technologies used

› Ready for review