Detect backdoor in Dropbear SSH server: ses.authstate.authdone set to 1 inside the failure handler, marking authentication as complete even on wrong password.
Performance
| Model | Pass Rate | Runs | Avg Cost | Avg Time |
|---|---|---|---|---|
| grok-4.1-fast | 0% | | $0.02 | 4m |
| deepseek-v3.2 | 0% | | $0.13 | 9m |
| kimi-k2.5 | 0% | | $0.15 | 7m |
| grok-4 | 0% | | $0.29 | 7m |
| gpt-5.2-codex | 0% | | $0.33 | 5m |
| claude-haiku-4.5 | 0% | | $0.36 | 6m |
| glm-4.7 | 0% | | $0.46 | 17m |
| gpt-5.2 | 0% | | $0.51 | 20m |
| claude-sonnet-4 | 0% | | $0.54 | 4m |
| gpt-5 | 0% | | $0.56 | 15m |
| gemini-2.5-pro | 0% | | $0.64 | 7m |
| gemini-3-flash-preview | 0% | | $0.70 | 6m |
| claude-sonnet-4.5 | 0% | | $0.71 | 6m |
| gemini-3-pro-preview | 0% | | $0.78 | 6m |
| claude-opus-4.5 | 0% | | $3.50 | 44m |
| claude-opus-4.6 | 0% | | $10.21 | 93m |
All product names, logos, and brands (™/®) are the property of their respective owners; they're used here solely for identification and comparison, and their use does not imply affiliation, endorsement, or sponsorship.