MalSandbox ML
Multi-layer URL threat analysis
A portfolio demonstration of the project's three-model threat-analysis pipeline. Enter a URL or choose a test case to see URL, behavioral and visual evidence combined into a weighted risk assessment.
FINAL VERDICT
READY
Preparing analysis…
URL NLP25%
0%
Behavior45%
0%
Vision CNN30%
0%
Weighted aggregation
0.25 × URL + 0.45 × Behavior + 0.30 × Vision = 0%
Demo telemetry
0network indicators
0process indicators
0file indicators
Capturedrendered screenshot
Project context
The project combines URL-level evidence, runtime behavior and rendered-page evidence. The reported model configuration uses URL NLP 25%, Behavior 45% and Vision 30%. The full project also includes MLflow experiment/inference tracking and an isolated Docker/Playwright execution layer.