Security Attacks Analysis on Travel Planning, Insurance, and Real Estate Security
84.68ASR (%)GPT-5
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GPT-5Firewall protection=w/o Firewall2025.02 | 84.68 | 7.99 | 96.55 | |
| Claude Sonnet 4Firewall protection=w/o Firewall2025.02 | 72.89 | 8.33 | 96.28 | |
| GPT-5Firewall protection=w/o Firewall2025.02 | 60.39 | 7.71 | 96.46 | |
| Gemini 2.5 FlashFirewall protection=w/o Firewall2025.02 | 37.91 | 7.23 | 82.81 | |
| Gemini 2.5 FlashFirewall protection=w/o Firewall2025.02 | 23.87 | 7.38 | 80.86 | |
| Claude Sonnet 4Firewall protection=w/ Firewall2025.02 | 16.77 | 8.55 | 99.21 | |
| Claude Sonnet 4Firewall protection=w/o Firewall2025.02 | 10.66 | 8.02 | 95.96 | |
| GPT-5Firewall protection=w/ Firewall2025.02 | 10.2 | 8.35 | 91.49 | |
| Gemini 2.5 FlashFirewall protection=w/ Firewall2025.02 | 10.18 | 7.78 | 92.93 | |
| Claude Sonnet 4Firewall protection=w/ Firewall2025.02 | 4.19 | 8.41 | 99 | |
| GPT-5Firewall protection=w/ Firewall2025.02 | 3.42 | 8.07 | 95.66 | |
| Gemini 2.5 FlashFirewall protection=w/ Firewall2025.02 | 1.02 | 7.54 | 89.33 |