Safeguards

Safeguards is a policy-as-code framework for Serverless Framework which enables you to inspect your serverless.yml file, and the generated Cloud Formation templates, for compliance with security, operational, and organizational, best practices. Safeguards are made available as a stand-alone Serverless Framework plugin with no external dependencies.

Highlights

• Stand-alone - it has no external dependencies on any services.
• Extensible - While the plugin comes with over a dozen policies out of the box, you can define new Safeguards and add them to your repo.
• Configurable - Safeguards are implemented to accept configuration as input so you can customize the policies for each safeguard.
• Proactive - Safeguards are evaluated before any infrastructure is provisioned by evaluating the generated cloud formation template and serverless.yml.
• Environment-specific - Policies can be associated with stages you can enforce different policies on development environments and production environments.
• Independent - While policies will get run when you deploy, you can run and validate the policies as a standalone without deploying.
• Enforceable - You can run the safeguards validate command while specifying additional policy files in addition to policies specified in serverless.yaml.

Docs

• Installation
• Defining policies
• Usage
• Example
• Migrating from Serverless Framework Pro
• Available Safeguards
• Custom policies

Installation

To install automatically run this single command:

serverless plugin install --name @serverless/safeguards-plugin

To install manually, run this command,

npm i @serverless/safeguards-plugin --save-dev

and add this to your serverless.yml:

plugins:
  - '@serverless/safeguards-plugin'

Defining Policies

Safeguard policies are defined as an array in the serverless.yml in the custom.safeguards field. You can add multiple policies, and you can use the same safeguard multiple times.

custom:
  safeguards:
    - safeguard:
      title:
      description:
      enforcementLevel:
      config:
      stage:

Fields

title (optional)

This is a user-readable name for the Safeguard policy. When the policy check is run in the CLI, the Safeguard policy name is used in the output.

description (optional)

The description should explain the intent of the policy. When the Safeguard policy check runs in the CLI this description will be displayed if the policy check fails. It is recommended that the description provides instructions on how to resolve an issue if the service is not compliant with the policy.

safeguard (required)

The Safeguard ID. There are over a dozen safeguards made availabe with the plugin. Each plugin has an ID (e.g. allowed-runtimes) which is used to reference in the policy.

enforcementLevel (optional, default: error)

The enforcement level can be set to either warning or error. When the Safeguard policy check runs in the CLI and the policy check passes, then enforcement level will have no impact on the deployment. However, if the policy check fails, then the enforcement level will control if the deployment can continue. If the enforcement level is set to warning, then the CLI will return a warning message but the deployment will continue. If the enforcement level is set to error, then the CLI will return an error message and the deployment will be blocked from continuing.

config

Some safeguards may allow or require configurations. For example, the Allowed Runtimes (allowed-runtimes) Safeguard requires a list of allowed AWS Lambda Runtimes for functions. This field allows you to customize the settings for the Safeguard policy.

path (optional)

If using a custom policy, this references the relative path to the safeguard base directory.

stage (optional)

By default a policy will run on all deployments, regardless of stage. However, if you want to scope the policy to only certain stages (e.g. prod), you can enforce the policy only on the selected stages. The stage field accepts string, or an array of strings, and if the current stage matches any of those, then the policy will be enforced.

Usage

The policy checks are performed as a part of the serverless deploy command. This will load the safeguard settings from the serverless.yml file to determine which policies to evaluate.

In addition, you simply can validate the configuration without doing a deploy. When validating standalone, you can specify additional policy files - zero, one or several - whose policies will be validated in addition to any policies in the serverless.yaml.

Example deploy

$ sls deploy
...
Serverless: Safeguards Results:

   Summary --------------------------------------------------

   passed - require-dlq
   passed - allowed-runtimes
   passed - no-secret-env-vars
   passed - allowed-stages
   failed - require-cfn-role
   passed - allowed-regions
   passed - framework-version
   failed - no-wild-iam-role-statements

   Details --------------------------------------------------

   1) Failed - no cfnRole set
      details: https://git.io/fhpFZ
      Require the cfnRole option, which specifies a particular role for CloudFormation to assume while deploying.


   2) Failed - iamRoleStatement granting Resource='*'. Wildcard resources in iamRoleStatements are not permitted.
      details: https://git.io/fjfk7
      Prevent "*" permissions being used in AWS IAM Roles by checking for wildcards on Actions and Resources in grant statements.


Serverless: Safeguards Summary: 6 passed, 0 warnings, 2 errors
...

Example standalone validate

$ sls safeguards validate
...
Serverless: Safeguards Results:

   Summary --------------------------------------------------

   passed - require-dlq
   passed - allowed-runtimes
   passed - no-secret-env-vars
   passed - allowed-stages
   failed - require-cfn-role
   passed - allowed-regions
   passed - framework-version
   failed - no-wild-iam-role-statements

   Details --------------------------------------------------

   1) Failed - no cfnRole set
      details: https://git.io/fhpFZ
      Require the cfnRole option, which specifies a particular role for CloudFormation to assume while deploying.


   2) Failed - iamRoleStatement granting Resource='*'. Wildcard resources in iamRoleStatements are not permitted.
      details: https://git.io/fjfk7
      Prevent "*" permissions being used in AWS IAM Roles by checking for wildcards on Actions and Resources in grant statements.


Serverless: Safeguards Summary: 6 passed, 0 warnings, 2 errors

Example standalone validate with additional policy files

$ sls safeguards validate --policy-file /home/policies/centralpolicyone.yaml --policy-file /home/policies/complianceglobal.yaml
...
Serverless: Safeguards Results:

   Summary --------------------------------------------------

   passed - require-dlq
   passed - allowed-runtimes
   passed - no-secret-env-vars
   passed - allowed-stages
   failed - require-cfn-role
   passed - allowed-regions
   passed - framework-version
   failed - no-wild-iam-role-statements

   Details --------------------------------------------------

   1) Failed - no cfnRole set
      details: https://git.io/fhpFZ
      Require the cfnRole option, which specifies a particular role for CloudFormation to assume while deploying.


   2) Failed - iamRoleStatement granting Resource='*'. Wildcard resources in iamRoleStatements are not permitted.
      details: https://git.io/fjfk7
      Prevent "*" permissions being used in AWS IAM Roles by checking for wildcards on Actions and Resources in grant statements.


Serverless: Safeguards Summary: 6 passed, 0 warnings, 2 errors

When providing additional policy files via --policy-file, these files must contain only policies, per the description in this file. The contents match precisely what would be put in the custom.safeguards section of a normal serverless.yaml file. For example:

- title: Restrict runtime
  safeguard: allowed-runtimes
  description: Only nodejs 8.10 allowed
  enforcementLevel: error # if this policy fails, then BLOCK the deployment
  config: # this configures the allowed-runtimes safeguard
    - nodejs8.10
  stage: # this policy will only be enforced if you deploy to prod or qa
    - prod
    - qa

The above is the entire policy file.

Policy check results

When a policy check is performed, the policy can respond with a pass, fail or warning. A fail will block and prevent the deploy from occurring. A warning will display a message but the deploy will continue.

If one or more of the policy checks fail the command will return a 1 exit code so it can be detected from a script or CI/CD service.

Example

service: aws-node-rest-api

provider:
  name: aws
  runtime: nodejs12.x
  region: us-east-1

functions:
  hello:
    handler: handler.hello

plugins:
  - @serverless/safeguards-plugin

custom:
  safeguards:
    - title: No secrets in lambda ENV VARs
      safeguard: no-secret-env-vars

    - title: Restrict regions
      safeguard: allowed-regions
      description: Only deployments in US regions are allowed
      enforcementLevel: error # if this policy fails, then BLOCK the deployment
      config: # this configures the allowed-regions safeguard
        - us-east-1
        - us-east-2
        - us-west-1
        - us-west-2
      stage: # this policy will only be enforced if you deploy to prod or qa
        - prod
        - qa

Migrating from Serverless Framework Pro

Serverless Framework Pro safeguards have been open source and repackaged in this plugin. Here is how you can migrate your existing configuration in SF Pro to use the @serverless/safeguards-plugin instead.

How safeguards work in Serverless Framework Pro

• Safeguard policies are added to deployment profiles.
• Each deployment profile then can be associated with an individual stage in an app.
• A deployment profile can also be assocaited with the default stage in the app.

How safeguard work in @serverless/safeguards-plugin

• Safeguard policies are added to each serverless.yml file under custom.safeguards.
• Safeguard policies are associated with stages by setting the stage field of each policy.

Breaking change

The breaking change is best described with the example below. Suppose you have an app with two stages configured, dev and prod. You also have a default stage which has some safeguard policies.

• default
• dev
• prod

In Serverless Framework Pro, if you deploy to one of the two stages (dev or prod) then the safegurad policies from the default stage willl NOT run. However, with @serverless/safeguards-plugin, you specify default policies by not setting the stage field, in which case, the default policies will also run.

Automatic Migration

1. Install the @serverless/safeguard-plugin to every serverless.yml which uses the Serverless Framework Pro safeguards.
2. Run serverless safeguards export from the working directory of the project.
3. Copy the generated YAML into the serverless.yml.
4. Delete the safeguards from the deployment profiles in Serverless Framework Pro.

Manually Migration

1. Install the @serverless/safeguard-plugin to every serverless.yml which uses the Serverless Framework Pro safeguards.
2. For each policy defined in the SF Pro dashboard, copy the configuration (name, description, enforcement level, config), into the custom.safeguards of your serverless.yml. The fields from the Safeguard Policies in the SF Pro dashboard match 1-1 with the fields in the @serverless/safeguards-plugin, so it should be as easy as copy-pasting.
3. Set the stage field of each policy in serverless.yml to match the stage names used in the app. For example, if you had a policy allowed-regions in the deployment profile and it was associated with the prod stage, then add the field stage: prod to the policy in the serverless.yml.
4. In SF Pro you have the ability to define stages (e.g. prod, qa) or use the default stage. The default stage is used to enforce safeguard policies from the deployment profile on any stages that don't match the other defined stages. For example, if you have prod and qa defined, but you deploy to feature-x, then the policies associated with the default stage will be used. For these policies, do not set the stage field, which will cause those policies to be enforced on all stages. At the moment, there isn't a way to define a blacklist for the stages.

Safeguards available with plugin

The following policies are included and configurable in the Serverless Framework Dashboard.

No "*" in IAM Role statements

ID: no-wild-iam-role-statements

This policy performs a simple check to prevent "*" permissions being used in AWS IAM Roles by checking for wildcards on Actions and Resources in grant statements.

Resolution

Update the custom IAM Roles in the serverless.yml to remove IAM Role Statements which grant access to "*" on Actions and Resources. If a plugin generates IAM Role Statements, follow the instructions provided by the plugin developer to mitigate the issue.

No clear-text credentials in environment variables

ID: no-secret-env-vars

Ensures that the environment variables configured on the AWS Lambda functions do not contain environment variables values which follow patterns of common credential formats.

Resolution

Resolving this issue requires that the AWS Lambda function environment variables do not contain any plain-text credentials; however, your functions may still require those credentials to be passed in by other means.

There are two recommended alternatives of passing in credentials to your AWS Lambda functions:

• SSM Parameter Store: The article "You should use SSM Parameter Store over Lambda env variables" by Yan Cui provides a detailed explanation for using the SSM Parameters in your Serverless Framework service to save and retrieve credentials.
• KMS Encryption: Encrypt the environment variables using KMS Keys.

Ensure Dead Letter Queues are attached to functions

ID: require-dlq

Ensures all functions with any of the events listed below, or functions with zero events, have an attached Dead Letter Queue.

Events:

• s3
• sns
• alexaSkill
• iot
• cloudwachEvent
• cloudwatchLog
• cognitoUserPool
• alexaHomeSkill

Resolution

Configure the Dead Letter Queue with SNS or SQS for all the functions which require the DLQ to be configured.

Allowed Runtimes

ID: allowed-runtimes

This limits the runtimes that can be used in services. It is configurable with a list of allowed runtimes or a regular expression.

- nodejs8.10
- python3.7
# or:
node.*

Resolution

Ensure you are using a runtime that is in the list of allowed runtimes or matches the regex of allowed runtimes.

Allowed stages

ID: allowed-stages

This limits the stages that can be used in services. It is configurable with a list of allowed stages or a regular expression.

- prod
- dev
# or:
'(prod|qa|dev-.*)'

Resolution

Ensure you are using a runtime that is in the list of allowed stages or matches the regex of allowed stages.

Framework Version

ID: framework-version

This policy limits which versions of the Serverless Framework can be used. It is configured with a semver expression.

>=1.44.0 <2.0.0

Resolution

Install an allowed version of the framework: npm i -g serverless@$ALLOWED_VERSION

Require Cloudformation Deployment Role

ID: require-cfn-role

This rule requires you to specify the cfnRole option in your serverless.yml. It has no configuration options.

Resolution

Add cfnRole to your serverless.yml.

Required stack tags

ID: required-stack-tags

This rule requires you to specify certain tags in the stackTags option in your serverless.yml. It is configured with a mapping of keys to regex's. All the keys must be present and value must match the regex.

someTagName: '.*'

Require Global VPC

ID: require-global-vpc

This rule requires all your functions to be configured with a VPC. By default they are required to have at least two subnet IDs to allow for AZ failover. It is configurable with a minNumSubnets option:

minNumSubnets: 1 # if you don't want to require 2 and AZ support

Resolution

Add a global VPC configuration to your config: https://serverless.com/framework/docs/providers/aws/guide/functions/#vpc-configuration

Allowed function names

ID: allowed-function-names

This rule allows you enforce naming conventions functions deployed to AWS lambda. It is configured with a regular expression. It features one extra addition: variables for stage, service and function(the key in the serverless yaml) names. See below for some examples.

Require using Serverless's standard naming scheme:

${SERVICE}-${STAGE}-${FUNCTION}

Or, if you want custom names with stage first and underscores instead of dashes:

${STAGE}_${SERVICE}_${FUNCTION}

Resolution

Use the name: config option on the function object to customize the deployed function name to match the regex: https://serverless.com/framework/docs/providers/aws/guide/functions/#configuration

Require Description

ID: require-description

This rule requires that all functions have a description of minimum or maximum length. By default it requires a minimum length of 30 and the lambda maximum of 256. Both these values are configurable however. Here is a config that requires a slightly longer config but doesn't allow as long a maximum:

minLength: 50
maxLength: 100

Resolution

Add a function description to all your lambdas that is with in the minimum and maximum required lengths.

Allowed Regions

ID: allowed-regions

This rule allows you to restrict the regions to which a service may be deployed. It is configured with a list of regions:

# eg, us-east-1 and us-west-2 only
- us-east-1
- us-west-2

Restricted deploy times

ID: restricted-deploy-times

This policy blocks deploys at certain times. It is configured with a list of objects containing a time, duration and optional interval.

# no deploy specific holidays, eg Rosh Hashanah 2019
- time: 2019-09-29T18:20 # ISO8601 date or datetime
  duration: P2D30M # IS8601 duration
# no deploy a specific day but repeating, eg all future Christmases
- time: 2019-12-25
  duration: P1D
  interval: P1Y
# no deploy fri noon - monday 6AM
- time: 2019-03-08T12:00:00
  duration: P2D18H
  interval: P1W

If you only need to specify one interval you can also directly use that object, eg:

# no deployments on friday, saturday, sunday
time: 2019-03-08
duration: P3D
interval: P1W

Resolution

Wait! You're not supposed to be deploying!

Forbid S3 HTTP Access

ID: forbid-s3-http-access

This policy requires that you have a BucketPolicy forbidding access over HTTP for each bucket. There are no configuration options.

Resolution

For a bucket without a name such as the ServerlessDeploymentBucket ensure that the resources section of your serverless yaml contains a policy like the following using Refs. If using a different bucket, update the logical name in the Ref.

resources:
  Resources:
    ServerlessDeploymentBucketPolicy:
      Type: 'AWS::S3::BucketPolicy'
      Properties:
        Bucket: { Ref: ServerlessDeploymentBucket }
        PolicyDocument:
          Statement:
            - Action: 's3:*'
              Effect: 'Deny'
              Principal: '*'
              Resource:
                Fn::Join:
                  - ''
                  - - 'arn:aws:s3:::'
                    - Ref: ServerlessDeploymentBucket
                    - '/*'
              Condition:
                Bool:
                  aws:SecureTransport: false

If using a bucket with a name, say configured in the custom section of your config, use a policy like this:

resources:
  Resources:
    NamedBucketPolicy:
      Type: 'AWS::S3::BucketPolicy'
      Properties:
        Bucket: ${self:custom.bucketName}
        PolicyDocument:
          Statement:
            - Action: 's3:*'
              Effect: 'Deny'
              Principal: '*'
              Resource: 'arn:aws:s3:::${self:custom.bucketName}/*'
              Condition:
                Bool:
                  aws:SecureTransport: false

Custom Policies

In addition to built-in policies, you can add custom policies to your application.

Creating a custom service policy

A service policy is simply a Javascript packaged in a module export which you can use in the Serverless Framework project for your service. To start with a custom policy first create a directory in your working directory (e.g. ./policies) to store the policy files.

Create a single JS file to define your policy (e.g. my-custom-policy.js) in the policies directory.

./policies/my-custom-policy.js

module.exports = function myCustomPolicy(policy, service) {
  // policy.fail(“Configuration is not compliant with policy”)
  policy.approve();
};

There are two primary methods you can use to control the behavior of the policy checks when running the deploy command.

• approve() - Passes the policy to allow the deploy to continue.
• fail(message) - Fails the policy check and returns an failure message.

To define the policy method you’ll need to inspect the configuration. The entire configuration is made available in the service object. Use the default policies and example policies as reference to the content of the service object.

Enabling a custom policy

Once the policy is implemented and saved in the directory, add the safeguards block to the serverless.yml file and set the location property to reference the relative path of the policies directory. To enable the policy you must also add it to the list of policies.

serverless.yml

custom:
  safeguards:
    - title: Require stage name in table name
      safeguard: stage-in-table-name
      path: ./policies

Adding settings to your policy

Custom policies may also include configuration parameters. The policy function accepts a third parameter (options in the example below) which contains the settings defined in the serverless.yml file.

./policies/my-custom-policy.js

module.exports = function myCustomPolicy(policy, service, options) {
  // options.max = 2
  policy.approve();
};

serverless.yml

custom:
  safeguards:
    - title: my custom policy
      safeguard: my-custom-policy
      path: ./policies
      config:
        max: 2