How Amazon SageMaker works with IAM - Amazon SageMaker

How Amazon SageMaker works with IAM

Important

Custom IAM policies that allow Amazon SageMaker Studio or Amazon SageMaker Studio Classic to create Amazon SageMaker resources must also grant permissions to add tags to those resources. The permission to add tags to resources is required because Studio and Studio Classic automatically tag any resources they create. If an IAM policy allows Studio and Studio Classic to create resources but does not allow tagging, "AccessDenied" errors can occur when trying to create resources. For more information, see Provide permissions for tagging SageMaker resources.

AWS managed policies for Amazon SageMaker that give permissions to create SageMaker resources already include permissions to add tags while creating those resources.

Before you use IAM to manage access to SageMaker, you should understand what IAM features are available to use with SageMaker. To get a high-level view of how SageMaker and other AWS services work with IAM, see AWS Services That Work with IAM in the Service Authorization Reference.

Identity-based policies for Amazon SageMaker

With IAM identity-based policies, you can specify allowed or denied actions and resources as well as the conditions under which actions are allowed or denied. SageMaker supports specific actions, resources, and condition keys. To learn about all of the elements that you use in a JSON policy, see IAM JSON Policy Elements Reference in the Service Authorization Reference.

Resource-based policies within Amazon SageMaker

Supports resource-based policies: No

Resource-based policies are JSON policy documents that you attach to a resource. Examples of resource-based policies are IAM role trust policies and Amazon S3 bucket policies. In services that support resource-based policies, service administrators can use them to control access to a specific resource. For the resource where the policy is attached, the policy defines what actions a specified principal can perform on that resource and under what conditions. You must specify a principal in a resource-based policy. Principals can include accounts, users, roles, federated users, or AWS services.

To enable cross-account access, you can specify an entire account or IAM entities in another account as the principal in a resource-based policy. Adding a cross-account principal to a resource-based policy is only half of establishing the trust relationship. When the principal and the resource are in different AWS accounts, an IAM administrator in the trusted account must also grant the principal entity (user or role) permission to access the resource. They grant permission by attaching an identity-based policy to the entity. However, if a resource-based policy grants access to a principal in the same account, no additional identity-based policy is required. For more information, see Cross account resource access in IAM in the IAM User Guide.

Note

Use AWS Resource Access Manager for securely sharing supported SageMaker resources. To find the list of sharable resources, see shareable Amazon SageMaker resources.

Policy actions for Amazon SageMaker

Administrators can use AWS JSON policies to specify who has access to what. That is, which principal can perform actions on what resources, and under what conditions.

The Action element of a JSON policy describes the actions that you can use to allow or deny access in a policy. Policy actions usually have the same name as the associated AWS API operation. There are some exceptions, such as permission-only actions that don't have a matching API operation. There are also some operations that require multiple actions in a policy. These additional actions are called dependent actions.

Include actions in a policy to grant permissions to perform the associated operation.

Policy actions in SageMaker use the following prefix before the action: sagemaker:. For example, to grant someone permission to run a SageMaker training job with the SageMaker CreateTrainingJob API operation, you include the sagemaker:CreateTrainingJob action in their policy. Policy statements must include either an Action or NotAction element. SageMaker defines its own set of actions that describe tasks that you can perform with this service.

To specify multiple actions in a single statement, separate them with commas as follows:

"Action": [ "sagemaker:action1", "sagemaker:action2" ]

You can specify multiple actions using wildcards (*). For example, to specify all actions that begin with the word Describe, include the following action:

"Action": "sagemaker:Describe*"

To see a list of SageMaker actions, see Actions, resources, and condition keys for Amazon SageMaker in the Service Authorization Reference.

Policy resources for Amazon SageMaker

Supports policy resources: Yes

Administrators can use AWS JSON policies to specify who has access to what. That is, which principal can perform actions on what resources, and under what conditions.

The Resource JSON policy element specifies the object or objects to which the action applies. Statements must include either a Resource or a NotResource element. As a best practice, specify a resource using its Amazon Resource Name (ARN). You can do this for actions that support a specific resource type, known as resource-level permissions.

For actions that don't support resource-level permissions, such as listing operations, use a wildcard (*) to indicate that the statement applies to all resources.

"Resource": "*"

To see a list of Amazon SageMaker resource types and their ARNs, see the following references for actions, resource types, and condition keys defined by Amazon SageMaker in the Service Authorization Reference.

To learn with which actions you can specify the ARN of each resource, see Actions defined by Amazon SageMaker.

Policy condition keys for Amazon SageMaker

Administrators can use AWS JSON policies to specify who has access to what. That is, which principal can perform actions on what resources, and under what conditions.

The Condition element (or Condition block) lets you specify conditions in which a statement is in effect. The Condition element is optional. You can create conditional expressions that use condition operators, such as equals or less than, to match the condition in the policy with values in the request.

If you specify multiple Condition elements in a statement, or multiple keys in a single Condition element, AWS evaluates them using a logical AND operation. If you specify multiple values for a single condition key, AWS evaluates the condition using a logical OR operation. All of the conditions must be met before the statement's permissions are granted.

You can also use placeholder variables when you specify conditions. For example, you can grant an IAM user permission to access a resource only if it is tagged with their IAM user name. For more information, see IAM policy elements: variables and tags in the IAM User Guide.

AWS supports global condition keys and service-specific condition keys. To see all AWS global condition keys, see AWS global condition context keys in the IAM User Guide.

SageMaker defines its own set of condition keys and also supports using some global condition keys. To see all AWS global condition keys, see AWS Global Condition Context Keys in the Service Authorization Reference.

SageMaker supports a number of service-specific condition keys that you can use for fine-grained access control for the following operations:

To see a list of SageMaker condition keys, see Condition keys for Amazon SageMaker in the Service Authorization Reference. To learn with which actions and resources you can use a condition key, see Actions defined by Amazon SageMaker.

For examples of using SageMaker condition keys, see the following: Control creation of SageMaker resources with condition keys.

Examples

To view examples of SageMaker identity-based policies, see Amazon SageMaker identity-based policy examples.

Authorization based on SageMaker tags

You can attach tags to SageMaker resources or pass tags in a request to SageMaker. To control access based on tags, you provide tag information in the condition element of a policy using the sagemaker:ResourceTag/key-name, aws:RequestTag/key-name, or aws:TagKeys condition keys. For more information about tagging SageMaker resources, see Control access to SageMaker resources by using tags.

To view an example identity-based policy for limiting access to a resource based on the tags on that resource, see Control access to SageMaker resources by using tags.

SageMaker IAM roles

An IAM role is an entity within your AWS account that has specific permissions.

Using temporary credentials with SageMaker

You can use temporary credentials to sign in with federation, assume an IAM role, or to assume a cross-account role. You obtain temporary security credentials by calling AWS STS API operations such as AssumeRole or GetFederationToken.

SageMaker supports using temporary credentials.

Service-linked roles

SageMaker partially supports service-linked roles. Service-linked roles are currently available for SageMaker Studio Classic.

Service roles

This feature allows a service to assume a service role on your behalf. This role allows the service to access resources in other services to complete an action on your behalf. Service roles appear in your IAM account and are owned by the account. This means that an IAM administrator can change the permissions for this role. However, doing so might break the functionality of the service.

SageMaker supports service roles.

Choosing an IAM role in SageMaker

When you create a notebook instance, processing job, training job, hosted endpoint, or batch transform job resource in SageMaker, you must choose a role to allow SageMaker to access SageMaker on your behalf. If you have previously created a service role or service-linked role, then SageMaker provides you with a list of roles to choose from. It's important to choose a role that allows access to the AWS operations and resources you need. For more information, see How to use SageMaker execution roles.