---
dataset_info:
  features:
  - name: sender
    dtype: large_string
  - name: sender_domain
    dtype: large_string
  - name: receiver
    dtype: large_string
  - name: receiver_domain
    dtype: large_string
  - name: date
    dtype: large_string
  - name: subject
    dtype: large_string
  - name: content_types
    dtype: large_string
  - name: body
    dtype: large_string
  - name: urls
    dtype: large_string
  - name: url_count
    dtype: float64
  - name: url_length_max
    dtype: float64
  - name: url_length_avg
    dtype: float64
  - name: url_subdom_max
    dtype: float64
  - name: url_subdom_avg
    dtype: float64
  - name: attachment_count
    dtype: float64
  - name: has_attachments
    dtype: bool
  - name: attachment_types
    dtype: large_string
  - name: language
    dtype: large_string
  - name: source
    dtype: large_string
  - name: label
    dtype: float64
  splits:
  - name: train
    num_bytes: 201554227
    num_examples: 108685
  download_size: 82741534
  dataset_size: 201554227
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
license: cc-by-4.0
tags:
- email
- phishing
pretty_name: 'MeAJOR: Merged email Assets from Joint Open-source Repositories'
---
# Dataset Card for MeAJOR: Merged email Assets from Joint Open-source Repositories

The MeAJOR dataset was created with multi-source legitimate emails and malicious emails from different attack campaigns to enable a more reliable phishing detection. MeAJOR provides a total of 108685 data samples, in preprocessed format with engineered features suitable for the training, validation, and testing of ML and DL models.

## Dataset Details

The dataset was created at GECAD, ISEP, Polytechnic of Porto. It contains data collected from publicly available spam email and phishing email datasets:

TREC-05
TREC-06
TREC-07
Nazario Phishing Corpus
Nigerian Fraud
The dataset contains the email messages anonymized, source dataset and label (0 = benign; 1 = phishing). It also contains several extracted features suitable for email phishing attack detection. The email messages are anonymized using Entity Removal, with tokens replacing them, enclosed between square brackets. The tokens are as follow: [PGP], [EMOJI], [SYMBOL], [NAME], [USERNAME], [INITIALS], [EMAIL_ADDRESS], [PHONE_NUMBER], [ADDRESS], [ORGANIZATION], [URL], [IP_ADDRESS], [FILE_PATH], [FILE_NAME], [FILE], [DATE], [TIME], [FINANCIAL_INFO], [PRODUCT], [REFERENCE_NUMBER].

- **License:** CC-BY-4.0

### Dataset Sources [optional]


- **Repository:** https://zenodo.org/records/18471483
- **Paper:** https://arxiv.org/abs/2507.17978v2