# YAML Deserialization via yaml.load on Untrusted Input

Language: Python
Severity: Critical
CWE: CWE-502

## Source
8

## Flow
8-9

## Sink
9

## Vulnerable Code
```python
import yaml
from flask import Flask, request

app = Flask(__name__)

@app.route('/iot/device/provision', methods=['POST'])
def provision_iot_device():
    device_manifest = request.data.decode('utf-8')
    device_config = yaml.load(device_manifest, Loader=yaml.Loader)
    device_id = device_config.get('device_id')
    firmware_url = device_config.get('firmware_endpoint')
    print(f"Provisioning device {device_id} with firmware from {firmware_url}")
    return {"status": "provisioned", "device": device_id}, 200
```

## Explanation

The application uses yaml.load() with yaml.Loader on untrusted input from request.data without validation. This allows attackers to execute arbitrary Python code through malicious YAML payloads containing Python object serialization tags, leading to Remote Code Execution during deserialization.

## Remediation

The fix replaces yaml.load() with yaml.safe_load(), which only deserializes standard YAML tags (strings, numbers, lists, dicts) and refuses to construct arbitrary Python objects. Additionally, input validation was added to ensure the parsed YAML is a dictionary and contains the required fields before processing.

## Secure Code
```python
import yaml
from flask import Flask, request

app = Flask(__name__)

@app.route('/iot/device/provision', methods=['POST'])
def provision_iot_device():
    device_manifest = request.data.decode('utf-8')
    device_config = yaml.safe_load(device_manifest)
    if not isinstance(device_config, dict):
        return {"status": "error", "message": "Invalid manifest format"}, 400
    device_id = device_config.get('device_id')
    firmware_url = device_config.get('firmware_endpoint')
    if not device_id or not firmware_url:
        return {"status": "error", "message": "Missing required fields"}, 400
    print(f"Provisioning device {device_id} with firmware from {firmware_url}")
    return {"status": "provisioned", "device": device_id}, 200
```
