Schema in a response does not define the type of JSON values
Issue ID: schema-response-notype
Average severity: Medium
Description
The API defines a schema for a JSON payload, but the schema does not actually limit what is accepted. This can happen for two reasons:
- The schema is for an object (a set of properties) but it does not have
type
set toobject
. The properties are matched to the ones defined in the schema, but no limit to their types is enforced. - Some of the properties themselves do not have
type
defined so any type is accepted for them.
For more details, see the OpenAPI Specification.
Example
The following is an example of how this type of risk could look in your API definition:
{
"post": {
"description": "Creates a new pet in the store",
// ...
"responses": {
"200": {
"description": "success",
"schema": {
"type": "object",
"$ref": "#/definitions/NewPet"
}
}
}
},
// ...
"definitions": {
"NewPet": {
"required": [
"name"
],
"properties": {
"name": {
"type": "string",
"description": "Pet name"
},
"tag": {
"description": "Pet tag"
}
}
}
}
}
Possible exploit scenario
Attackers strive to make your APIs behave in an unexpected way to learn more about your system or to cause a data breach. Good data definition quality in the schemas used in API responses allows reliably validating that the contents of outgoing API responses are as expected.
While filtering API responses does not block a specific kind of attack, it is there as a damage control mechanism in the unfortunate event that a successful attack has been conducted: it allows blocking the response and prevent attackers from retrieving data they should not access.
In the vast majority of cases (with the notable exception of Denial of Service (DoS and DDoS) attacks) attacks are conducted because attackers want to access data or resources they should not have access to. Often, this means that the structure or the size of the API response changes as a result of a successful attack, compared to a normal API response.
Validating that API responses are as expected can be achieved through proper schema validation of the API responses. The accuracy of this depends on the quality of the response schemas: the better defined your schemas are, the easier it is to detect when something is not right.
Remediation
Make sure you define the types of all properties in the JSON payload and correctly set the type for objects to enforce limitations to what the schema accepts.
{
"post": {
"description": "Creates a new pet in the store",
// ...
"responses": {
"200": {
"description": "success",
"schema": {
"type": "object",
"$ref": "#/definitions/NewPet"
}
}
}
},
// ...
"definitions": {
"NewPet": {
"type": "object",
"required": [
"name"
],
"properties": {
"name": {
"type": "string",
"description": "Pet name"
},
"tag": {
"type": "string",
"description": "Pet tag"
}
}
}
}
}