Phishing validation emails dataset

Description:A balanced, labelled collection of 2,000 short English-language email-body texts for validating machine-learning phishing detectors. Each record contains the plain text of one email and a binary label (Safe Email or Phishing Email), with 1,000 records per class and all 2,000 texts distinct. Ninety-five percent of the messages are synthetically generated from 26 phishing and 26 safe templates; the remaining 100 are real communications collected by the authors from personal email inboxes and SMS message histories, with headers and identifiers removed. Version 3 extends version 2 by adding four phone- and reply-based phishing templates and six URL-bearing safe templates, reducing the URL-rule accuracy from 96.8% (v2) to 77.6%, and by supplying a companion metadata file (phishing_v3_metadata.csv) with a template identifier for every record to support group-aware cross-validation. The dataset is intended as a held-out test set: train a classifier on a separate corpus and evaluate it on these 2,000 records. Dataset Structure: Total Emails: 2,000 Email Types: Safe Emails Phishing Emails Attributes: Full text of the email Label indicating whether the email is safe or phishing Example Entries: Email Text: "Dear Jordan, your subscription has been succes..." Email Type: Safe Email Email Text: "Congratulations! You've won a $3000 gift card...." Email Type: Phishing Email Acknowledgments Special gratitude is extended to the University of Twente, the Technologies for Criminal Investigations (TCI) research group, part of Saxion University of Applied Sciences and the Police Academy in the Netherlands, the Technical University of Sofia and the AI & CAD Systems Lab at Sofia Tech Park. Additionally, the work is supported by project No. BG16RFPR002-1.014-0014-C01, "Development Program with a Business Plan for the Laboratory Complex of Sofia Tech Park", implemented under the "Research, Innovation and Digitalization for Smart Transformation" Program, co-financed by the European Union through the European Regional Development Fund and part of the project "Regiodeal Stedendriehoek" funded by the Dutch National Government.