You cannot select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
ansible/hacking/aws_config/build_iam_policy_framework.py

328 lines
12 KiB
Python

# Requires pandas, bs4, html5lib, and lxml
#
# Call script with the output from aws_resource_actions callback, e.g.
# python build_iam_policy_framework.py ['ec2:AuthorizeSecurityGroupEgress', 'ec2:AuthorizeSecurityGroupIngress', 'sts:GetCallerIdentity']
#
# The sample output:
# {
# "Version": "2012-10-17",
# "Statement": [
# {
# "Sid": "AnsibleEditor0",
# "Effect": "Allow",
# "Action": [
# "ec2:AuthorizeSecurityGroupEgress",
# "ec2:AuthorizeSecurityGroupIngress"
# ],
# "Resource": "arn:aws:ec2:${Region}:${Account}:security-group/${SecurityGroupId}"
# },
# {
# "Sid": "AnsibleEditor1",
# "Effect": "Allow",
# "Action": [
# "sts:GetCallerIdentity"
# ],
# "Resource": "*"
# }
# ]
# }
#
# Policy troubleshooting:
# - If there are more actions in the policy than you provided, AWS has documented dependencies for some of your actions and
# those have been added to the policy.
# - If there are fewer actions in the policy than you provided, some of your actions are not in the IAM table of actions for
# that service. For example, the API call s3:DeleteObjects does not actually correlate to the permission needed in a policy.
# In this case s3:DeleteObject is the permission required to allow both the s3:DeleteObjects action and the s3:DeleteObject action.
# - The policies output are only as accurate as the AWS documentation. If the policy does not permit the
# necessary actions, look for undocumented dependencies. For example, redshift:CreateCluster requires ec2:DescribeVpcs,
# ec2:DescribeSubnets, ec2:DescribeSecurityGroups, and ec2:DescribeInternetGateways, but AWS does not document this.
#
import json
import requests
import sys
missing_dependencies = []
try:
import pandas as pd
except ImportError:
missing_dependencies.append('pandas')
try:
import bs4
except ImportError:
missing_dependencies.append('bs4')
try:
import html5lib
except ImportError:
missing_dependencies.append('html5lib')
try:
import lxml
except ImportError:
missing_dependencies.append('lxml')
irregular_service_names = {
'a4b': 'alexaforbusiness',
'appstream': 'appstream2.0',
'acm': 'certificatemanager',
'acm-pca': 'certificatemanagerprivatecertificateauthority',
'aws-marketplace-management': 'marketplacemanagementportal',
'ce': 'costexplorerservice',
'cognito-identity': 'cognitoidentity',
'cognito-sync': 'cognitosync',
'cognito-idp': 'cognitouserpools',
'cur': 'costandusagereport',
'dax': 'dynamodbacceleratordax',
'dlm': 'datalifecyclemanager',
'dms': 'databasemigrationservice',
'ds': 'directoryservice',
'ec2messages': 'messagedeliveryservice',
'ecr': 'ec2containerregistry',
'ecs': 'elasticcontainerservice',
'eks': 'elasticcontainerserviceforkubernetes',
'efs': 'elasticfilesystem',
'es': 'elasticsearchservice',
'events': 'cloudwatchevents',
'firehose': 'kinesisfirehose',
'fms': 'firewallmanager',
'health': 'healthapisandnotifications',
'importexport': 'importexportdiskservice',
'iot1click': 'iot1-click',
'kafka': 'managedstreamingforkafka',
'kinesisvideo': 'kinesisvideostreams',
'kms': 'keymanagementservice',
'license-manager': 'licensemanager',
'logs': 'cloudwatchlogs',
'opsworks-cm': 'opsworksconfigurationmanagement',
'mediaconnect': 'elementalmediaconnect',
'mediaconvert': 'elementalmediaconvert',
'medialive': 'elementalmedialive',
'mediapackage': 'elementalmediapackage',
'mediastore': 'elementalmediastore',
'mgh': 'migrationhub',
'mobiletargeting': 'pinpoint',
'pi': 'performanceinsights',
'pricing': 'pricelist',
'ram': 'resourceaccessmanager',
'resource-groups': 'resourcegroups',
'sdb': 'simpledb',
'servicediscovery': 'cloudmap',
'serverlessrepo': 'serverlessapplicationrepository',
'sms': 'servermigrationservice',
'sms-voice': 'pinpointsmsandvoiceservice',
'sso-directory': 'ssodirectory',
'ssm': 'systemsmanager',
'ssmmessages': 'sessionmanagermessagegatewayservice',
'states': 'stepfunctions',
'sts': 'securitytokenservice',
'swf': 'simpleworkflowservice',
'tag': 'resourcegrouptaggingapi',
'transfer': 'transferforsftp',
'waf-regional': 'wafregional',
'wam': 'workspacesapplicationmanager',
'xray': 'x-ray'
}
irregular_service_links = {
'apigateway': [
'https://docs.aws.amazon.com/IAM/latest/UserGuide/list_manageamazonapigateway.html'
],
'aws-marketplace': [
'https://docs.aws.amazon.com/IAM/latest/UserGuide/list_awsmarketplace.html',
'https://docs.aws.amazon.com/IAM/latest/UserGuide/list_awsmarketplacemeteringservice.html',
'https://docs.aws.amazon.com/IAM/latest/UserGuide/list_awsprivatemarketplace.html'
],
'discovery': [
'https://docs.aws.amazon.com/IAM/latest/UserGuide/list_applicationdiscovery.html'
],
'elasticloadbalancing': [
'https://docs.aws.amazon.com/IAM/latest/UserGuide/list_elasticloadbalancing.html',
'https://docs.aws.amazon.com/IAM/latest/UserGuide/list_elasticloadbalancingv2.html'
],
'globalaccelerator': [
'https://docs.aws.amazon.com/IAM/latest/UserGuide/list_globalaccelerator.html'
]
}
def get_docs_by_prefix(prefix):
amazon_link_form = 'https://docs.aws.amazon.com/IAM/latest/UserGuide/list_amazon{0}.html'
aws_link_form = 'https://docs.aws.amazon.com/IAM/latest/UserGuide/list_aws{0}.html'
if prefix in irregular_service_links:
links = irregular_service_links[prefix]
else:
if prefix in irregular_service_names:
prefix = irregular_service_names[prefix]
links = [amazon_link_form.format(prefix), aws_link_form.format(prefix)]
return links
def get_html(links):
html_list = []
for link in links:
html = requests.get(link).content
try:
parsed_html = pd.read_html(html)
html_list.append(parsed_html)
except ValueError as e:
if 'No tables found' in str(e):
pass
else:
raise e
return html_list
def get_tables(service):
links = get_docs_by_prefix(service)
html_list = get_html(links)
action_tables = []
arn_tables = []
for df_list in html_list:
for df in df_list:
table = json.loads(df.to_json(orient='split'))
table_data = table['data'][0]
if 'Actions' in table_data and 'Resource Types (*required)' in table_data:
action_tables.append(table['data'][1::])
elif 'Resource Types' in table_data and 'ARN' in table_data:
arn_tables.append(table['data'][1::])
# Action table indices:
# 0: Action, 1: Description, 2: Access level, 3: Resource type, 4: Condition keys, 5: Dependent actions
# ARN tables indices:
# 0: Resource type, 1: ARN template, 2: Condition keys
return action_tables, arn_tables
def add_dependent_action(resources, dependency):
resource, action = dependency.split(':')
if resource in resources:
resources[resource].append(action)
else:
resources[resource] = [action]
return resources
def get_dependent_actions(resources):
for service in dict(resources):
action_tables, arn_tables = get_tables(service)
for found_action_table in action_tables:
for action_stuff in found_action_table:
if action_stuff is None:
continue
if action_stuff[0] in resources[service] and action_stuff[5]:
dependencies = action_stuff[5].split()
if isinstance(dependencies, list):
for dependency in dependencies:
resources = add_dependent_action(resources, dependency)
else:
resources = add_dependent_action(resources, dependencies)
return resources
def get_actions_by_service(resources):
service_action_dict = {}
dependencies = {}
for service in resources:
action_tables, arn_tables = get_tables(service)
# Create dict of the resource type to the corresponding ARN
arn_dict = {}
for found_arn_table in arn_tables:
for arn_stuff in found_arn_table:
arn_dict["{0}*".format(arn_stuff[0])] = arn_stuff[1]
# Create dict of the action to the corresponding ARN
action_dict = {}
for found_action_table in action_tables:
for action_stuff in found_action_table:
if action_stuff[0] is None:
continue
if arn_dict.get(action_stuff[3]):
action_dict[action_stuff[0]] = arn_dict[action_stuff[3]]
else:
action_dict[action_stuff[0]] = None
service_action_dict[service] = action_dict
return service_action_dict
def get_resource_arns(aws_actions, action_dict):
resource_arns = {}
for resource_action in aws_actions:
resource, action = resource_action.split(':')
if action not in action_dict:
continue
if action_dict[action] is None:
resource = "*"
else:
resource = action_dict[action].replace("${Partition}", "aws")
if resource not in resource_arns:
resource_arns[resource] = []
resource_arns[resource].append(resource_action)
return resource_arns
def get_resources(actions):
resources = {}
for action in actions:
resource, action = action.split(':')
if resource not in resources:
resources[resource] = []
resources[resource].append(action)
return resources
def combine_arn_actions(resources, service_action_arn_dict):
arn_actions = {}
for service in service_action_arn_dict:
service_arn_actions = get_resource_arns(aws_actions, service_action_arn_dict[service])
for resource in service_arn_actions:
if resource in arn_actions:
arn_actions[resource].extend(service_arn_actions[resource])
else:
arn_actions[resource] = service_arn_actions[resource]
return arn_actions
def combine_actions_and_dependent_actions(resources):
aws_actions = []
for resource in resources:
for action in resources[resource]:
aws_actions.append('{0}:{1}'.format(resource, action))
return set(aws_actions)
def get_actions_restricted_by_arn(aws_actions):
resources = get_resources(aws_actions)
resources = get_dependent_actions(resources)
service_action_arn_dict = get_actions_by_service(resources)
aws_actions = combine_actions_and_dependent_actions(resources)
return combine_arn_actions(aws_actions, service_action_arn_dict)
def main(aws_actions):
arn_actions = get_actions_restricted_by_arn(aws_actions)
statement = []
for resource_restriction in arn_actions:
statement.append({
"Sid": "AnsibleEditor{0}".format(len(statement)),
"Effect": "Allow",
"Action": arn_actions[resource_restriction],
"Resource": resource_restriction
})
policy = {"Version": "2012-10-17", "Statement": statement}
print(json.dumps(policy, indent=4))
if __name__ == '__main__':
if missing_dependencies:
sys.exit('Missing Python libraries: {0}'.format(', '.join(missing_dependencies)))
actions = sys.argv[1:]
if len(actions) == 1:
actions = sys.argv[1].split(',')
aws_actions = [action.strip('[], "\'') for action in actions]
main(aws_actions)