license: cc-by-4.0
language: [en, tr]
task_categories: [text-classification]
tags: [prompt-injection, llm-security, ai-security, unicode, invisible-text, red-teaming, security, agent-security]
pretty_name: "Invisible Unicode Prompt Injection"
size_categories: [n<1K]
configs: [{config_name: default, data_files: iui.jsonl}]
Invisible Unicode Prompt Injection
from datasets import load_dataset
ds = load_dataset("fevziegeyurtsevenler/invisible-unicode-injection")
42 examples where innocent visible text hides an instruction in invisible Unicode (Tags block U+E0000–E007F, zero-width). The model reads the payload; a human reviewer sees nothing. Decode + detect them.
Each row: visible_text (innocent), hidden_payload (decoded), full_text (with the real invisible chars), technique, detected_by (uncloak rule), owasp, severity, defense. Scanned by uncloak.
Schema
| column | meaning |
|---|---|
| visible_text | what a human sees |
| hidden_payload | decoded hidden instruction |
| full_text | text with the invisible characters |
| technique, detected_by, owasp, severity, defense | labels |
Related AltaySec resources
- 🕵️ uncloak — hidden prompt-injection scanner: GitHub · live demo
- 📚 Datasets: detection-rules · glossary · skills-in-the-wild · agent-skill-vulnerabilities
- 🌐 AltaySec · Açık Kaynak Lab
Citation
@misc{yurtsevenler2026invisibleunicodeinjection,
title = {Invisible Unicode Prompt Injection Dataset},
author = {Yurtsevenler, Fevzi Ege},
year = {2026},
publisher = {AltaySec / Hugging Face},
howpublished = {\url{https://huggingface.co/datasets/fevziegeyurtsevenler/invisible-unicode-injection}}
}
⚠️ Defensive / authorized-use only. Multilingual-first LLM/agent security research by AltaySec.