trannguyenthaituan/vietnamese_sms_phishing_sample

Quality-Assured Vietnamese SMS Phishing Dataset (Sample Evaluation Corpus)

This repository provides an open-access 300-message evaluation sample of the Quality-Assured Vietnamese SMS Phishing Dataset, a leakage-controlled benchmark constructed using a closed-loop multi-annotator agreement framework and near-duplicate template deduplication (J ≥ 0.85).

Authors & Research Team

• Trần Nguyễn Thái Tuấn (Lead Researcher & Project Lead)
• Lê Hoàng Khang (Researcher)
• Nguyễn Minh Tài (Researcher)
• Nguyễn Văn Thắng (Researcher)
• Mai Hoàng Đỉnh (Project Advisor / Mentor)

Dataset Summary

• Language: Vietnamese (vi)
• Task: Binary SMS Spam / Phishing Detection (HAM vs. SPAM)
• Sanitization: 100% PII Anonymized (Phone numbers, bank accounts, names, and OTP tokens replaced with standardized tokens [PHONE], [BANK_ACC], [MONEY], [NUMBER], [TIME])
• Quality Audit: Manually audited on 200 random samples showing 0% residual PII leakage.
• Licensing: Creative Commons Attribution 4.0 International (CC BY 4.0)

Dataset Structure

The dataset file sample_300.csv contains the following columns:

Column Name	Data Type	Description
message_id	String	Unique tracking identifier (e.g., COM_001, COM_002)
text	String	Sanitized unaccented/accented Vietnamese SMS text

Citation & Paper Reference

If you use this dataset sample in your research, please cite our manuscript:

@article{tuan2026vietnamese_sms_phishing,
  title={Quality-Assured Vietnamese SMS Phishing Dataset: Closed-Loop Human Adjudication Framework and Leakage-Controlled Transformer Benchmarking},
  author={Tran, Nguyen Thai Tuan and Le, Hoang Khang and Nguyen, Minh Tai and Nguyen, Van Thang and Mai, Hoang Dinh},
  journal={IEEE Access / Computer Science Domain},
  year={2026}
}

Ethics Statement & Institutional Review

Collection protocol was granted formal institutional ethics exemption (IRB Protocol Ref: IRB-2026-NLP-0428). All user-contributed messages were gathered under voluntary informed consent.