pickle.load() used on untrusted model file
The application loads model files using pickle.load() which can execute arbitrary code during deserialization.
CWE
CWE-502
CVE
N/A
OWASP
N/A
CVSS Score
9.8
Source
AISecurity
Category
Insecure Model Loading
Exploitability
Exploitable
Detected
2026-05-31 19:02
Code Location
src/ml_pipeline/model_loader.py:23model = pickle.load(open(model_path, 'rb'))
AI Explanation
This finding was detected by automated scanning. Use the Security Copilot for a detailed AI-powered explanation of this vulnerability, including attack vectors and business impact analysis.
Remediation Suggestion
Use safetensors format for model serialization, or torch.load() with weights_only=True parameter.
Related Findings
| Severity | Title | CWE | Status |
|---|---|---|---|
High |
Unsafe JSON deserialization with TypeNameHandling | CWE-502 | Open |
Change Status