MeAJOR: Merged email assets from joint open-source repositories

Area of Science:

• Cybersecurity
• Machine Learning
• Data Science

Background:

• Phishing remains a significant cybercrime threat, exploiting human vulnerabilities.
• Existing datasets limit the effectiveness of machine learning models for phishing detection.
• Data quality and diversity are crucial for robust phishing detection systems.

Purpose of the Study:

• To introduce the MeAJOR Corpus, a novel, multi-source phishing email dataset.
• To address limitations in existing resources for training effective phishing detection models.
• To provide a comprehensive dataset for advancing cybersecurity research.

Main Methods:

• Integration of 108,685 phishing and legitimate email samples from joint open-source repositories.
• Anonymization of email bodies and subjects using standard tokens to preserve context.
• Inclusion of diverse phishing tactics and statistical features for comprehensive analysis.

Main Results:

• The MeAJOR Corpus offers a large-scale, varied dataset for phishing email analysis.
• Anonymization techniques ensure data privacy while maintaining contextual integrity.
• The dataset supports the development of more accurate and resilient phishing detection models.

Conclusions:

• The MeAJOR Corpus is a valuable resource for improving machine learning-based phishing detection.
• Enhanced datasets are critical for combating evolving cybercrime threats.
• This work contributes to strengthening cybersecurity defenses against phishing attacks.