Phishing URLs Detection Using Sequential and Parallel ML Techniques: Comparative Analysis.

Area of Science:

• Cybersecurity
• Computer Science
• Artificial Intelligence

Background:

• Cyber-attacks, particularly phishing, exploit Uniform Resource Locators (URLs) to steal sensitive data and cause financial/reputational damage.
• Artificial intelligence (AI) techniques like machine learning (ML) and deep learning (DL) are effective for phishing detection.
• Sequential ML models can be slow and inefficient for real-time detection of vast cyber threats.

Purpose of the Study:

• To investigate the efficiency of parallel computing techniques in accelerating ML and DL model training for enhanced phishing attack detection.
• To compare the performance of sequential versus parallel execution for training various ML and DL models.

Main Methods:

• Implemented multiprocessing and multithreading techniques in Python for parallel model training.
• Trained four models: Random Forest (RF), Naïve Bayes (NB), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM).
• Utilized a dataset of 54,000 records for training and 12,000 for testing, conducting five experiments comparing sequential and parallel execution.

Main Results:

• Parallel execution using Python's multiprocessing and multithreading significantly reduced computation time for ML and DL models.
• Achieved excellent results and notable speedup in phishing detection model training.
• Comparative analysis confirmed the effectiveness of parallel techniques over sequential methods.

Conclusions:

• Parallel computing offers a robust and efficient approach to developing precise ML and DL models for real-time phishing attack detection.
• Python's parallel backend and multiprocessing capabilities enhance the scalability and speed of cybersecurity threat detection.
• The study validates the use of parallel ML/DL for building effective, faster, and more robust cyber-attack defense systems.