fix: the /predict endpoint accepts file uploads with... in app.py

[orbisai0security]

Summary

Fix high severity security issue in official/projects/waste_identification_ml/docker_solution/prediction_api/app.py.

Vulnerability

Field	Value
ID	V-005
Severity	HIGH
Scanner	multi_agent_ai
Rule	V-005
File	official/projects/waste_identification_ml/docker_solution/prediction_api/app.py:47

Description: The /predict endpoint accepts file uploads without any rate limiting, file size restrictions, or input dimension validation. No MAX_CONTENT_LENGTH or equivalent framework configuration is present. ML inference is computationally expensive — processing large images consumes significant CPU and memory. An attacker (who requires no authentication per V-004) can submit a continuous stream of large image files or a single extremely large file (e.g., a gigapixel image or decompression bomb) to exhaust all available CPU, memory, and disk resources on the prediction server.

Changes

• official/projects/waste_identification_ml/docker_solution/prediction_api/app.py

Verification

• Build passes
• Scanner re-scan confirms fix
• LLM code review passed

---

Automated security fix by OrbisAI Security