Datasets:
|
Download README.md from Sahildhonde-9/INJEXIS-Prompt-Injection-Dataset: direct link, hf CLI and curl.
- Browser
- Download file 7.5 kB
-
https://huggingface.co/datasets/Sahildhonde-9/INJEXIS-Prompt-Injection-Dataset/resolve/main/README.md
- Command line
-
hf download hf://datasets/Sahildhonde-9/INJEXIS-Prompt-Injection-Dataset/README.md
-
curl -L -o README.md https://huggingface.co/datasets/Sahildhonde-9/INJEXIS-Prompt-Injection-Dataset/resolve/main/README.md
7.5 kB
| license: apache-2.0 | |
| task_categories: | |
| - text-classification | |
| language: | |
| - en | |
| tags: | |
| - security | |
| - LLM-safety | |
| - prompt-injection | |
| - jailbreak | |
| - adversarial-ai | |
| - ai-security | |
| - llm-security | |
| - red-team | |
| - prompt-defense | |
| - ai-firewall | |
| - cybersecurity | |
| - instruction-override | |
| - system-prompt-protection | |
| - dataset | |
| - ai-defense | |
| size_categories: | |
| - 100K<n<1M | |
| <div align="center"> | |
| <img src="https://readme-typing-svg.demolab.com?font=Orbitron&weight=700&size=26&pause=1000&color=00BFFF¢er=true&vCenter=true&width=850&lines=Prompt+Injection+Dataset;LLM+Security+Benchmark;Adversarial+AI+Research;Jailbreak+Attack+Samples;AI+Firewall+Training+Data" /> | |
| <img src="https://capsule-render.vercel.app/api?type=waving&color=0:001F3F,50:00BFFF,100:001F3F&height=220§ion=header&text=RedLockX%20Dataset&fontSize=50&fontColor=ffffff&animation=fadeIn&fontAlignY=35&desc=AI%20Security%20%7C%20Prompt%20Injection%20Research&descAlignY=58&descSize=18"/> | |
| </div> | |
| <style> | |
| /* ================================ | |
| π GLOBAL BACKGROUND | |
| ================================ */ | |
| body { | |
| margin: 0; | |
| font-family: 'Segoe UI', sans-serif; | |
| background: | |
| radial-gradient(circle at top left, rgba(59,130,246,0.15), transparent 25%), | |
| radial-gradient(circle at bottom right, rgba(96,165,250,0.12), transparent 25%), | |
| linear-gradient(135deg, #020617, #081120, #0b1220, #0f172a); | |
| background-size: cover; | |
| color: #e5e7eb; | |
| overflow-x: hidden; | |
| } | |
| /* ================================ | |
| π§ WATER BLUE LIGHT EFFECT | |
| ================================ */ | |
| body::before { | |
| content: ""; | |
| position: fixed; | |
| inset: 0; | |
| background: | |
| repeating-radial-gradient( | |
| circle at 50% 50%, | |
| rgba(59,130,246,0.08) 0px, | |
| rgba(59,130,246,0.03) 40px, | |
| transparent 80px | |
| ); | |
| animation: waterFlow 18s linear infinite; | |
| pointer-events: none; | |
| z-index: -1; | |
| } | |
| @keyframes waterFlow { | |
| 0% { | |
| transform: translateY(0px) scale(1); | |
| } | |
| 50% { | |
| transform: translateY(-20px) scale(1.03); | |
| } | |
| 100% { | |
| transform: translateY(0px) scale(1); | |
| } | |
| } | |
| /* ================================ | |
| β¨ BLUE GLOW PARTICLES | |
| ================================ */ | |
| body::after { | |
| content: ""; | |
| position: fixed; | |
| inset: 0; | |
| background-image: | |
| radial-gradient(rgba(96,165,250,0.25) 1px, transparent 1px); | |
| background-size: 40px 40px; | |
| opacity: 0.35; | |
| animation: sparkleMove 30s linear infinite; | |
| pointer-events: none; | |
| z-index: -1; | |
| } | |
| @keyframes sparkleMove { | |
| from { | |
| transform: translateY(0px); | |
| } | |
| to { | |
| transform: translateY(-100px); | |
| } | |
| } | |
| /* ================================ | |
| π¦ MAIN CONTAINER | |
| ================================ */ | |
| .container { | |
| display: flex; | |
| flex-direction: column; | |
| align-items: center; | |
| justify-content: center; | |
| text-align: center; | |
| padding: 60px 20px; | |
| animation: fadeIn 1.5s ease; | |
| } | |
| /* ================================ | |
| πΌ FLOATING HERO IMAGE | |
| ================================ */ | |
| .hero-images { | |
| display: flex; | |
| gap: 25px; | |
| margin-bottom: 30px; | |
| } | |
| .hero-images img { | |
| width: 220px; | |
| border-radius: 16px; | |
| border: 1px solid rgba(96,165,250,0.25); | |
| box-shadow: | |
| 0 0 25px rgba(59,130,246,0.35), | |
| 0 0 60px rgba(37,99,235,0.18); | |
| animation: float 5s ease-in-out infinite; | |
| transition: transform 0.3s ease, box-shadow 0.3s ease; | |
| } | |
| .hero-images img:hover { | |
| transform: scale(1.05); | |
| box-shadow: | |
| 0 0 35px rgba(96,165,250,0.55), | |
| 0 0 90px rgba(37,99,235,0.25); | |
| } | |
| @keyframes float { | |
| 0% { transform: translateY(0px); } | |
| 50% { transform: translateY(-15px); } | |
| 100% { transform: translateY(0px); } | |
| } | |
| /* ================================ | |
| π· TITLE STYLE | |
| ================================ */ | |
| h1 { | |
| font-size: 2.9rem; | |
| margin: 10px 0; | |
| background: linear-gradient( | |
| 90deg, | |
| #60a5fa, | |
| #93c5fd, | |
| #38bdf8, | |
| #60a5fa | |
| ); | |
| background-size: 300% auto; | |
| -webkit-background-clip: text; | |
| -webkit-text-fill-color: transparent; | |
| animation: | |
| shimmer 6s linear infinite, | |
| fadeIn 1.5s ease forwards; | |
| text-shadow: 0 0 30px rgba(96,165,250,0.35); | |
| } | |
| @keyframes shimmer { | |
| to { | |
| background-position: 300% center; | |
| } | |
| } | |
| /* ================================ | |
| βοΈ PARAGRAPH STYLE | |
| ================================= */ | |
| p { | |
| max-width: 850px; | |
| line-height: 1.7; | |
| font-size: 1.05rem; | |
| color: #cbd5e1; | |
| animation: fadeIn 2s ease forwards; | |
| } | |
| /* ================================ | |
| π¬ FADE IN ANIMATION | |
| ================================ */ | |
| @keyframes fadeIn { | |
| from { | |
| opacity: 0; | |
| transform: translateY(18px); | |
| } | |
| to { | |
| opacity: 1; | |
| transform: translateY(0); | |
| } | |
| } | |
| /* ================================ | |
| π₯ SECTION HEADINGS | |
| ================================ */ | |
| h2 { | |
| margin-top: 40px; | |
| font-size: 1.8rem; | |
| color: #93c5fd; | |
| text-shadow: 0 0 18px rgba(96,165,250,0.25); | |
| } | |
| /* ================================ | |
| π TABLE DESIGN | |
| ================================ */ | |
| table { | |
| margin-top: 30px; | |
| width: 95%; | |
| max-width: 900px; | |
| border-collapse: collapse; | |
| background: rgba(15,23,42,0.45); | |
| backdrop-filter: blur(12px); | |
| border: 1px solid rgba(96,165,250,0.15); | |
| border-radius: 14px; | |
| overflow: hidden; | |
| box-shadow: | |
| 0 10px 35px rgba(0,0,0,0.35), | |
| 0 0 20px rgba(59,130,246,0.08); | |
| } | |
| th, td { | |
| padding: 14px; | |
| border-bottom: 1px solid rgba(255,255,255,0.06); | |
| text-align: left; | |
| } | |
| th { | |
| background: rgba(96,165,250,0.08); | |
| color: #93c5fd; | |
| } | |
| td { | |
| color: #e5e7eb; | |
| } | |
| /* ================================ | |
| π LIST STYLE | |
| ================================ */ | |
| ul { | |
| max-width: 850px; | |
| text-align: left; | |
| color: #cbd5e1; | |
| line-height: 1.7; | |
| } | |
| /* ================================ | |
| π GLASS PANELS | |
| ================================ */ | |
| /* ================================ | |
| π± RESPONSIVE | |
| ================================ */ | |
| @media (max-width: 768px) { | |
| .hero-images { | |
| flex-direction: column; | |
| align-items: center; | |
| } | |
| h1 { | |
| font-size: 2rem; | |
| } | |
| p, | |
| ul { | |
| font-size: 0.98rem; | |
| } | |
| } | |
| </style> | |
| <div class="container"> | |
| <p> | |
| The <b>RedLockX Dataset</b> is a large-scale curated security dataset designed for | |
| evaluating and training AI systems against adversarial threats such as prompt injection, | |
| jailbreak attempts, system prompt leakage, and LLM manipulation attacks. | |
| <br><br> | |
| It contains structured real-world and synthetic attack patterns used in modern AI red-teaming. | |
| </p> | |
| <!-- OVERVIEW --> | |
| <h2>π Dataset Overview</h2> | |
| <ul> | |
| β 109,000+ labeled adversarial & safe samples <br> | |
| β Multi-category threat classification system <br> | |
| β OWASP LLM Top 10 mapping included <br> | |
| β Severity scoring (0β10 scale) <br> | |
| β Risk scoring (0β100 business impact model) <br> | |
| β Suitable for fine-tuning, evaluation, and benchmarking | |
| </ul> | |
| <!-- USAGE --> | |
| <h2>π How to Use</h2> | |
| <p> | |
| You can load this dataset using the Hugging Face <code>datasets</code> library. | |
| </p> | |
| <ul> | |
| <li>Use for prompt injection detection models</li> | |
| <li>Train LLM guardrails and safety classifiers</li> | |
| <li>Benchmark adversarial robustness</li> | |
| <li>Red-team AI systems before deployment</li> | |
| </ul> | |
| <!-- LICENSE --> | |
| <h2>βοΈ License</h2> | |
| <p> | |
| This dataset is released under the <b>Apache 2.0 License</b>.<br> | |
| It is intended strictly for research in AI safety, adversarial robustness, and security evaluation. | |
| </p> | |
| </div> | |
| <div align="center"> | |
| <img src="https://capsule-render.vercel.app/api?type=waving&color=0:00BFFF,100:001F3F&height=140§ion=footer"/> | |
| </div> |