TorchServe is a flexible and easy-to-use tool for serving and scaling PyTorch models in production. In affected versions the two gRPC ports 7070 and 7071, are not bound to localhost by default, so when TorchServe is launched, these two interfaces are bound to all interfaces. Customers using PyTorch inference Deep Learning Containers (DLC) through Amazon SageMaker and EKS are not affected. This issue in TorchServe has been fixed in PR #3083. TorchServe release 0.11.0 includes the fix to address this vulnerability. Users are advised to upgrade. There are no known workarounds for this vulnerability.
The product exposes a resource to the wrong control sphere, providing unintended actors with inappropriate access to the resource.
Resources such as files and directories may be inadvertently exposed through mechanisms such as insecure permissions, or when a program accidentally operates on the wrong object. For example, a program may intend that private files can only be provided to a specific user. This effectively defines a control sphere that is intended to prevent attackers from accessing these private files. If the file permissions are insecure, then parties other than the user will be able to access those files. A separate control sphere might effectively require that the user can only access the private files, but not any other files on the system. If the program does not ensure that the user is only requesting private files, then the user might be able to access other files on the system. In either case, the end result is that a resource has been exposed to the wrong party.