when you evaluate the security of your model using another one of your models