Adversarial Prompt Detection on 10,000 Adversarial Prompts
83Detection RateOur Framework
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Our FrameworkProcessing Time=15ms2026.03 | 83 | 5 | |
| SecurityLinguaProcessing Time=12ms2026.03 | 80 | 7 | |
| PromptShieldProcessing Time=20ms2026.03 | 78 | 8 | |
| Commercial Moderation APIsProcessing Time=180-220ms2026.03 | 67 | 11 | |
| Open-Source DetoxifyProcessing Time=45ms2026.03 | 64 | 9 |