Security documentation update

README.md

@@ -1,5 +1,8 @@
 # TorchServe
 
+## Security Changes
+TorchServe now enforces token authorization and model API control by default. Check the following documentation for more information: [Token Authorization](https://github.com/pytorch/serve/blob/master/docs/token_authorization_api.md), [Model API control](https://github.com/pytorch/serve/blob/master/docs/model_api_control.md)
+
 ![Nightly build](https://github.com/pytorch/serve/actions/workflows/torchserve-nightly-build.yml/badge.svg)
 ![Docker Nightly build](https://github.com/pytorch/serve/actions/workflows/docker-nightly-build.yml/badge.svg)
 ![Benchmark Nightly](https://github.com/pytorch/serve/actions/workflows/benchmark_nightly.yml/badge.svg)

docker/README.md

@@ -1,3 +1,6 @@
+## Security Changes
+TorchServe now enforces token authorization and model API control by default. Check the following documentation for more information: [Token Authorization](https://github.com/pytorch/serve/blob/master/docs/token_authorization_api.md), [Model API control](https://github.com/pytorch/serve/blob/master/docs/model_api_control.md)
+
 ### Deprecation notice:
 [Dockerfile.neuron.dev](https://github.com/pytorch/serve/blob/master/docker/Dockerfile.neuron.dev) has been deprecated. Please refer to [deep learning containers](https://github.com/aws/deep-learning-containers/blob/master/available_images.md) repository for neuron torchserve containers.
 

docs/README.md

@@ -1,5 +1,9 @@
 # TorchServe
 
+## Security Changes
+TorchServe now enforces token authorization and model API control by default. Check the following documentation for more information: [Token Authorization](https://github.com/pytorch/serve/blob/master/docs/token_authorization_api.md), [Model API control](https://github.com/pytorch/serve/blob/master/docs/model_api_control.md)
+
+
 TorchServe is a performant, flexible and easy to use tool for serving PyTorch eager mode and torchscripted models.
 
 ## Basic Features

docs/inference_api.md

@@ -2,6 +2,8 @@
 
 Inference API is listening on port 8080 and only accessible from localhost by default. To change the default setting, see [TorchServe Configuration](configuration.md).
 
+For all Inference API requests, TorchServe requires the correct Inference token to be included. For more details see [token authorization documentation](./token_authorization_api.md)
+
 The TorchServe server supports the following APIs:
 
 * [API Description](#api-description) - Gets a list of available APIs and options

docs/management_api.md

@@ -8,9 +8,12 @@ TorchServe provides the following APIs that allows you to manage models at runti
 4. [Unregister a model](#unregister-a-model)
 5. [List registered models](#list-models)
 6. [Set default version of a model](#set-default-version)
+7. [Refresh tokens for token authorization](#token-authorization-api)
 
 The Management API listens on port 8081 and is only accessible from localhost by default. To change the default setting, see [TorchServe Configuration](./configuration.md).
 
+For all Management API requests, TorchServe requires the correct Management token to be included. For more details see [token authorization documentation](./token_authorization_api.md)
+
 Similar to the [Inference API](inference_api.md), the Management API provides a [API description](#api-description) to describe management APIs with the OpenAPI 3.0 specification.
 
 Alternatively, if you want to use KServe, TorchServe supports both v1 and v2 API. For more details please look into this [kserve documentation](https://github.com/pytorch/serve/tree/master/kubernetes/kserve)
@@ -19,6 +22,8 @@ Alternatively, if you want to use KServe, TorchServe supports both v1 and v2 API
 
 This API follows the [ManagementAPIsService.RegisterModel](https://github.com/pytorch/serve/blob/master/frontend/server/src/main/resources/proto/management.proto) gRPC API.
 
+To use this API after TorchServe starts, model API control has to be enabled. For more details see [model API control](./model_api_control.md)
+
 `POST /models`
 
 * `url` - Model archive download url. Supports the following locations:
@@ -441,6 +446,8 @@ print(customizedMetadata)
 
 This API follows the [ManagementAPIsService.UnregisterModel](https://github.com/pytorch/serve/blob/master/frontend/server/src/main/resources/proto/management.proto) gRPC API. It returns the status of a model in the ModelServer.
 
+To use this API after TorchServe starts, model API control has to be enabled. For more details see [model API control](./model_api_control.md)
+
 `DELETE /models/{model_name}/{version}`
 
 Use the Unregister Model API to free up system resources by unregistering specific version of a model from TorchServe:
@@ -522,3 +529,21 @@ curl -v -X PUT http://localhost:8081/models/noop/2.0/set-default
 ```
 
 The out is OpenAPI 3.0.1 json format. You use it to generate client code, see [swagger codegen](https://swagger.io/swagger-codegen/) for detail.
+
+## Token Authorization API
+
+TorchServe now enforces token authorization by default. Check the following documentation for more information: [Token Authorization](https://github.com/pytorch/serve/blob/master/docs/token_authorization_api.md).
+
+This API is used in order to generate a new key to replace either the management or inference key.
+
+Management Example:
+```
+curl localhost:8081/token?type=management -H "Authorization: Bearer {API Token}"
+```
+will replace the current management key in the key_file with a new one and will update the expiration time.
+
+Inference example:
+```
+curl localhost:8081/token?type=inference -H "Authorization: Bearer {API Token}"
+```
+will replace the current inference key in the key_file with a new one and will update the expiration time.

docs/model_control_mode.md → docs/model_api_control.md

@@ -1,9 +1,11 @@
-# Model Control Mode
+# Model API Control
 
-TorchServe now supports model control mode with two settings "none"(default) and "enabled"
+TorchServe now supports model API control with two settings enabled and disabled(default).
+
+TorchServe introduces a new feature called model-api-control which allow users to prevent registering and deleting models once the servers are running. This is a security feature which addresses the concern of unwanted registering and deleting of models once the TorchServe servers have started. This is applicable in the scenario where a user may upload malicious code to the model server in the form of a model or where a user may delete a model that is being used. The default behavior prevents users from registering or deleting models once TorchServe is running, and then you can enable the model APIs to allow users to register and delete models whenever using the TorchServe model load APIs.
 
 ## Two ways to set Model Control
-1. Add `--model-api-enabled` to command line when running TorchServe to switch from none to enabled mode. Command line cannot be used to set mode to none, can only be used to set to enabled
+1. Add `--model-api-enabled` to command line when running TorchServe to switch from disabled to enabled. Command line cannot be used to disabled, can only be used to enabled
 2. Add `model_api_enabled=false` or `model_api_enabled=true` to config.properties file
     * `model_api_enabled=false` is default and prevents users from registering or deleting models once TorchServe is running
     * `model_api_enabled=true` is not default and allows users to register and delete models using the TorchServe model load APIs
@@ -22,7 +24,7 @@ Priority between cmd and config file follows the following [TorchServer standard
 
     Result: Mode is enabled (no way to disable api mode through cmd)
 
-## Model Control Mode Default
+## Model API Control Default
 At startup TorchServe loads only those models specified explicitly with the `--models` command-line option. After startup users will be unable to register or delete models in this mode.
 
 ### Example default
@@ -40,7 +42,7 @@ ubuntu@ip-172-31-11-32:~/serve$ curl -X POST  "http://localhost:8081/models?url=
 ```
 
 ## Model Control API Enabled
-Setting model control to `enabled` allows users to load and unload models using the model load APIs.
+Setting model API to `enabled` allows users to load and unload models using the model load APIs.
 
 ### Example using cmd line to set mode to enabled
 ```

docs/token_authorization_api.md

@@ -2,6 +2,7 @@
 
 TorchServe now enforces token authorization by default
 
+TorchServe introduces a new feature called token-authorization which requires the correct token to be provided when calling an API. This is a security feature which addresses the concern of unwanted API calls. This is applicable in the scenario where an unauthorized user may try to access a running TorchServe server. The default behavior is to enable this feature which creates a key file with the appropriate tokens to be used for API calls. Users can disable this feature to prevent token authorization from being required for API calls.
 
 ## How to set and disable Token Authorization
 * Global environment variable: use `TS_DISABLE_TOKEN_AUTHORIZATION` and set to `true` to disable and `false` to enable token authorization. Note that `enable_envvars_config=true` must be set in config.properties for global environment variables to be used

examples/README.md

@@ -1,5 +1,8 @@
 # [Examples showcasing TorchServe Features and Integrations](#torchserve-internals)
 
+## Security Changes
+TorchServe now enforces token authorization and model API control by default. This change will impact the current examples so please check the following documentation for more information: [Token Authorization](https://github.com/pytorch/serve/blob/master/docs/token_authorization_api.md), [Model API control](https://github.com/pytorch/serve/blob/master/docs/model_api_control.md)
+
 ## TorchServe Internals
 
 * [Creating mar file for an eager mode model](#creating-mar-file-for-eager-mode-model)

kubernetes/README.md

@@ -1,5 +1,8 @@
 # Torchserve on Kubernetes
 
+## Security Changes
+TorchServe now enforces token authorization and model API control by default. Check the following documentation for more information: [Token Authorization](https://github.com/pytorch/serve/blob/master/docs/token_authorization_api.md), [Model API control](https://github.com/pytorch/serve/blob/master/docs/model_api_control.md)
+
 ## Overview
 
 This page demonstrates a Torchserve deployment in Kubernetes using Helm Charts. It uses the DockerHub Torchserve Image for the pods and a PersistentVolume for storing config / model files.