Three ComfyUI nodes for security scanning:
- Scan Model: Scans model files before loading, blocks execution
if threats found (pickle exploits, polyglot attacks, CVE bypasses)
- Audit Installed Nodes: Static analysis of all custom node packages
for eval/exec, subprocess, network access, obfuscation patterns
- Analyze Workflow: Graph analysis of node connections for dangerous
data flows, known CVE node types, URL injection, and suspicious inputs
Includes:
- Node auditor with regex + AST-based detection
- Workflow graph analyzer tracing data flows through node connections
- Known dangerous node database (8+ CVEs cataloged)
- Detection of code injection, command execution, network access,
file access, dynamic imports, obfuscation, and supply chain risks
Built on TensorTrap security research documenting 8 CVEs, 4 major
real-world attacks, and 700+ confirmed server compromises in the
ComfyUI ecosystem.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
