Skip to main content

TasksSozu Multiple-Binary Backdoor Detection

Binary AnalysisSozu21% pass rateView TaskView Prompt

Detect backdoor in Sozu load balancer (multiple binaries): X-Browser-ID header triggers shell command execution, returning base64-encoded output.

Performance

ModelPass RateRunsAvg CostAvg Time
Anthropicclaude-opus-4.6100%
$1.8730m
Googlegemini-3-pro-preview67%
$0.436m
Googlegemini-3.1-pro-preview67%
$1.1017m
OpenAIgpt-5.3-codex-xhigh67%
$1.5665m
OpenAIgpt-5.2-codex-xhigh67%
$3.5471m
Grokgrok-433%
$0.6018m
OpenAIgpt-5.3-codex-high33%
$0.6813m
OpenAIgpt-5.2-codex33%
$0.8017m
Anthropicclaude-opus-4.533%
$1.5820m
OpenAIgpt-5.2-codex-high33%
$2.3934m
Grokgrok-4.1-fast0%
$0.0520m
DeepSeekdeepseek-v3.20%
$0.0615m
Googlegemini-3-flash-preview0%
$0.126m
Z.aiglm-4.70%
$0.2214m
Anthropicclaude-haiku-4.50%
$0.237m
Kimikimi-k2.50%
$0.2621m
Googlegemini-2.5-pro0%
$0.299m
Anthropicclaude-sonnet-40%
$0.423m
Z.aiglm-50%
$0.4928m
OpenAIgpt-5.3-codex0%
$0.5617m
OpenAIgpt-50%
$0.7830m
OpenAIgpt-5.20%
$0.9224m
Anthropicclaude-sonnet-4.50%
$1.0525m
OpenAIgpt-5.4-xhigh0%
$1.94120m
OpenAIgpt-5.4-high0%
$2.1052m
OpenAIgpt-5.40%
$8.7881m