Transparent and trustworthy CyberSecurity: an XAI-integrated big data framework for phishing attack detection

Area of Science:

• Cybersecurity Analytics
• Machine Learning
• Big Data Technologies

Background:

• Modern digital infrastructures generate vast, high-velocity cybersecurity data, challenging traditional threat detection methods.
• Sophisticated cyber-attacks necessitate scalable Big Data Analytics and advanced Machine Learning (ML) techniques.
• The 'black-box' nature of many ML models hinders interpretability, trust, and regulatory compliance in critical security applications.

Purpose of the Study:

• To propose an integrated framework for accurate, transparent, and real-time phishing attack detection.
• To combine Big Data technologies, ML models, and Explainable Artificial Intelligence (XAI) for enhanced cybersecurity.
• To address the limitations of traditional methods in processing large-scale, high-velocity security data.

Main Methods:

• Leveraging distributed computing and stream processing for efficient handling of diverse cybersecurity datasets.
• Integrating advanced Machine Learning (ML) models for phishing detection.
• Incorporating Explainable Artificial Intelligence (XAI) methods to generate human-understandable model explanations.

Main Results:

• Experimental evaluation on four public cybersecurity datasets demonstrated improved phishing detection performance.
• The framework enhanced the interpretability of ML model decisions in threat detection.
• Actionable insights into malicious URL behavior and patterns were generated.

Conclusions:

• The proposed approach advances interpretable and scalable cybersecurity analytics.
• It bridges the gap between predictive accuracy and decision transparency in threat detection.
• The framework provides a trustworthy solution for real-time threat detection, supporting informed decision-making and regulatory compliance.