ml-scanner

Model training + inference pipeline with a FastAPI gateway. Last analyzed 3 days ago.

Project detail

88/100

Project HealthNeeds Attention

Findings summary

0 Critical1 High2 Medium0 Low

Issues that need attention are listed below.

Score breakdown

Security

Based on detected findings

Reliability

Based on detected findings

Testing

Regression tests added

Code Quality

Based on detected findings

Dependencies

Based on detected findings

Issues

High

Unsafe Model Deserialization

Security

models/loader.py:24

Trusted pickle.load is used on models downloaded from a URL that can be influenced by an advisory banner, enabling code execution if the source is ever tampered with.

View Details Fix Issue

Medium

Sensitive Payload Logged at INFO

Security

api/predict.py:12

The inference gateway logs the full request body, silently capturing user-provided text that may include PII.

View Details Fix Issue

Medium

Model Cache Without TTL or Version Key

Reliability

api/cache.py:45

Cached predictions are never invalidated, so stale weights keep serving after a redeploy.

View Details Fix Issue

Uploaddone
Analyzedone
Findactive
Explainupcoming
Fixupcoming
Testupcoming
Verifyupcoming