aws_iam_policy resource

[edit on GitHub]

Use the aws_iam_policy InSpec audit resource to test properties of a single managed AWS IAM Policy.

Syntax

An aws_iam_policy resource block identifies a policy by policy name or arn

# Find a policy by name
describe aws_iam_policy('AWSSupportAccess') do
  it { should exist }
end

# Hash syntax for policy name
describe aws_iam_policy(policy_name: 'AWSSupportAccess') do
  it { should exist }
end

Parameters

This resource requires either the policy_name or the policy_arn to be provided.

policy_name (required if policy_arn not provided)

The Policy Name which uniquely identifies the Policy. This must be passed as a policy_name: 'value' key-value entry in a hash.

policy_arn (required if policy_name not provided)

The Policy ARN which uniquely identifies the Policy. This must be passed as a policy_arn: 'value' key-value entry in a hash.

See also the AWS documentation on IAM Policy.

Properties

Examples

Test that a policy does exist

describe aws_iam_policy('AWSSupportAccess') do
  it { should exist }
end

Test that a policy is attached to at least one entity

describe aws_iam_policy('AWSSupportAccess') do
  it { should be_attached }
end

Examine the policy statements

describe aws_iam_policy('my-policy') do
  # Verify that there is at least one statement allowing access to S3
  it { should have_statement(Action: 's3:PutObject', Effect: 'allow') }

  # have_statement does not expand wildcards. If you want to verify
  # they are absent, an explicit check is required.
  it { should_not have_statement(Action: 's3:*') }

  # You can also check NotAction
  it { should_not have_statement(NotAction: 'iam:*') }
end

Matchers

This InSpec audit resource has the following special matchers. For a full list of available matchers, please visit our Universal Matchers page.

exist

The control will pass if the describe returns at least one result.

Use should_not to test the entity should not exist.

  it { should exist }

  it { should_not exist }

be_attached

The test will pass if the identified policy is attached to at least one IAM user, group, or role.

describe aws_iam_policy('AWSSupportAccess') do
  it { should be_attached }
end

be_attached_to_group(GROUPNAME)

The test will pass if the identified policy attached the specified group.

describe aws_iam_policy('AWSSupportAccess') do
  it { should be_attached_to_group(GROUPNAME) }
end

be_attached_to_user(USERNAME)

The test will pass if the identified policy attached the specified user.

describe aws_iam_policy('AWSSupportAccess') do
  it { should be_attached_to_user(USERNAME) }
end

be_attached_to_role(ROLENAME)

The test will pass if the identified policy attached the specified role.

describe aws_iam_policy('AWSSupportAccess') do
  it { should be_attached_to_role(ROLENAME) }
end

have_statement

Examines the list of statements contained in the policy and passes if at least one of the statements matches. This matcher does not interpret the policy in a request authorization context, as AWS does when a request processed. Rather, have_statement examines the literal contents of the IAM policy, and reports on what is present (or absent, when used with should_not).

have_statement accepts the following criteria to search for matching statements. If any statement matches all the criteria, the test is successful. All criteria may be used as Titlecase (as in the AWS examples) or lowercase, string or symbol.

• Action - Expresses the requested operation. Acceptable literal values are any AWS operation name, including the ‘*’ wildcard character. Action may also use a list of AWS operation names.
• Effect - Expresses if the operation is permitted. Acceptable values are ‘Deny’ and ‘Allow’.
• Sid - A user-provided string identifier for the statement.
• Resource - Expresses the operation’s target. Acceptable values are ARNs, including the ‘*’ wildcard. Resource may also use a list of ARN values.

Please note the following about the behavior of have_statement:

• Action, Sid, and Resource allow using a regular expression as the search critera instead of a string literal.
• It does not support wildcard expansion; to check for a wildcard value, check for it explicitly. For example, if the policy includes a statement with "Action": "s3:*" and the test checks for Action: "s3:PutObject", the test will not match. You must write an additional test checking for the wildcard case.
• It supports searching list values. For example, if a statement contains a list of 3 resources, and a have_statement test specifes one of those resources, it will match.
• Action and Resource allow using a list of string literals or regular expressions in a test, in which case all must match on the same statement for the test to match. Order is ignored.
• It does not support the Principal, NotPrincipal or Condition.

Examples:

# Verify there is no full-admin statement
describe aws_iam_policy('kryptonite') do
  it { should_not have_statement('Effect' => 'Allow', 'Resource' => '*', 'Action' => '*')}
end

# Symbols and lowercase also allowed as criteria
describe aws_iam_policy('kryptonite') do
  # All 4 the same
  it { should_not have_statement('Effect' => 'Allow', 'Resource' => '*', 'Action' => '*')}
  it { should_not have_statement('effect' => 'Allow', 'resource' => '*', 'action' => '*')}
  it { should_not have_statement(Effect: 'Allow', Resource: '*', Action: '*')}
  it { should_not have_statement(effect: 'Allow', resource: '*', action: '*')}
end

# Verify bob is allowed to manage things on S3 buckets that start with bobs-stuff
describe aws_iam_policy('bob-is-a-packrat') do
  it { should have_statement(Effect: 'Allow',
                             # Using the AWS wildcard - this must match exactly
                             Resource: 'arn:aws:s3:::bobs-stuff*',
                             # Specify a list of actions - all must match, no others, order isn't important
                             Action: ['s3:PutObject', 's3:GetObject', 's3:DeleteObject'])}

  # Bob would make new buckets constantly if we let him.
  it { should_not have_statement(Effect: 'Allow', Action: 's3:CreateBucket')}
  it { should_not have_statement(Effect: 'Allow', Action: 's3:*')}
  it { should_not have_statement(Effect: 'Allow', Action: '*')}

  # An alternative to checking for wildcards is to specify the
  # statements you expect, then restrict statement count
  its('statement_count') { should cmp 1 }
end

# Use regular expressions to examine the policy
describe aws_iam_policy('regex-demo') do
  # Check to see if anything mentions RDS at all.
  # This catches `rds:CreateDBinstance` and `rds:*`, but would not catch '*'.
  it { should_not have_statement(Action: /^rds:.+$/)}

  # This policy should refer to both sally and kim's s3 buckets.
  # This will only match if there is a statement that refers to both resources.
  it { should have_statement(Resource: [/arn:aws:s3.+:sally/, /arn:aws:s3.+:kim/]) }
  # The following also matches on a statement mentioning only one of them
  it { should have_statement(Resource: /arn:aws:s3.+:(sally|kim)/) }
end

AWS Permissions

Your Principal will need the iam:GetPolicy, iam:ListPolicy, and iam:ListEntitiesForPolicy actions set to allow.

You can find detailed documentation at Actions, Resources, and Condition Keys for Identity And Access Management.