Detect backdoor in Sozu load balancer (multiple binaries): X-Browser-ID header triggers shell command execution, returning base64-encoded output.
Performance
| Model | Pass Rate | Runs | Avg Cost | Avg Time |
|---|---|---|---|---|
| claude-opus-4.6 | 100% | | $1.87 | 30m |
| gemini-3-pro-preview | 67% | | $0.43 | 6m |
| grok-4 | 33% | | $0.60 | 18m |
| gpt-5.2-codex | 33% | | $0.80 | 17m |
| claude-opus-4.5 | 33% | | $1.58 | 20m |
| grok-4.1-fast | 0% | | $0.05 | 20m |
| deepseek-v3.2 | 0% | | $0.06 | 15m |
| gemini-3-flash-preview | 0% | | $0.12 | 6m |
| glm-4.7 | 0% | | $0.22 | 14m |
| claude-haiku-4.5 | 0% | | $0.23 | 7m |
| kimi-k2.5 | 0% | | $0.26 | 21m |
| gemini-2.5-pro | 0% | | $0.29 | 9m |
| claude-sonnet-4 | 0% | | $0.42 | 3m |
| gpt-5 | 0% | | $0.78 | 30m |
| gpt-5.2 | 0% | | $0.92 | 24m |
| claude-sonnet-4.5 | 0% | | $1.05 | 25m |
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.