- Introduction
- How does Regula work?
- Running Regula locally
- Regula rules
- Compliance controls vs. rules
- Interpreting the results
- Running Regula in CI
- Development
Regula is a tool that evaluates Terraform infrastructure-as-code for potential security misconfigurations and compliance violations prior to deployment.
Regula includes a library of rules written in Rego, the policy language used by the Open Policy Agent (opa) project. Regula works with your favorite CI/CD tools such as Jenkins, Circle CI, and AWS CodePipeline; we’ve included a GitHub Actions example so you can get started quickly. Where relevant, we’ve mapped Regula policies to the CIS AWS Foundations Benchmark so you can assess your compliance posture. We'll be adding more rules in the coming weeks, sourced from Fugue.
There are two parts to Regula. The first is a shell script that generates a terraform plan in JSON format, ready for consumption by opa.
The second part is a Rego framework that:
- Merges resource info from
planned_valuesandconfigurationin the terraform plan into a more conveniently accessible format. - Walks through the imported terraform modules and merges them into a flat format.
- Looks for rules and executes them.
- Creates a report with the results of all rules and a control mapping in the output.
./bin/regula [TERRAFORM_PATH] [REGO_PATHS...]
TERRAFORM_PATH is the directory where your terraform configuration files are
located.
REGO_PATHS are the directories that need to be searched for Rego code. This
should at least include lib/.
Some examples:
./bin/regula ../my-tf-infra .: conveniently check../my-tf-infraagainst all rules in this main repository../bin/regula ../my-tf-infra lib examples/aws/t2_only.rego: run Regula using only the specified rule../bin/regula ../my-tf-infra lib ../custom-rules: run Regula using a directory of custom rules.
It is also possible to set the name of the terraform executable, which is
useful if you have several versions installed:
env TERRAFORM=terraform-v0.12.18 ./bin/regula ../regula-action-example/ lib
Regula rules are written in standard Rego and use a similar format to Fugue Custom Rules. This means there are (currently) two kinds of rules: simple rules and advanced rules.
Simple rules are useful when the policy applies to a single resource type only, and you want to make simple yes/no decision.
# Rules must always be located right below the `rules` package.
package rules.my_simple_rule
# Simple rules must specify the resource type they will police.
resource_type = "aws_ebs_volume"
# Simple rules must specify `allow` or `deny`. For this example, we use
# an `allow` rule to check that the EBS volume is encrypted.
default allow = false
allow {
input.encrypted == true
}Advanced rules are harder to write, but more powerful. They allow you to observe different kinds of resource types and decide which specific resources are valid or invalid.
# Rules still must be located in the `rules` package.
package rules.user_attached_policy
# Advanced rules typically use functions from the `fugue` library.
import data.fugue
# We mark an advanced rule by setting `resource_type` to `MULTIPLE`.
resource_type = "MULTIPLE"
# `fugue.resources` is a function that allows querying for resources of a
# specific type. In our case, we are just going to ask for the EBS volumes
# again.
ebs_volumes = fugue.resources("aws_ebs_volume")
# Auxiliary function.
is_encrypted(resource) {
resource.encrypted == true
}
# Regula expects advanced rules to contain a `policy` rule that holds a set
# of _judgements_.
policy[p] {
resource = ebs_volumes[_]
is_encrypted(resource)
p = fugue.allow_resource(resource)
} {
resource = ebs_volumes[_]
not is_encrypted(resource)
p = fugue.deny_resource(resource)
}The fugue API consists of four functions:
fugue.resources(resource_type)returns an object with all resources of the requested type.fugue.allow_resource(resource)marks a resource as valid.fugue.deny_resource(resource)marks a resource as invalid.fugue.missing_resource(resource_type)marks a resource as missing. This is useful if you for example require a log group to be present.
See rules directory. Fugue is currently working on open sourcing more rules from our product to Regula.
| Provider | Service | Rule Name | Rule Summary |
|---|---|---|---|
| AWS | CloudTrail | cloudtrail_log_file_validation | CloudTrail log file validation should be enabled |
| AWS | EBS | ebs_volume_encrypted | EBS volume encryption should be enabled |
| AWS | IAM | iam_admin_policy | IAM policies should not have full ":" administrative privileges |
| AWS | IAM | iam_user_attached_policy | IAM policies should not be attached directly to users |
| AWS | KMS | kms_rotate | KMS CMK rotation should be enabled |
| AWS | VPC | security_group_ingress_anywhere | VPC security group rules should not permit ingress from '0.0.0.0/0' except to ports 80 and 443 |
| AWS | VPC | security_group_ingress_anywhere_rdp | VPC security group rules should not permit ingress from '0.0.0.0/0' to port 3389 (Remote Desktop Protocol) |
| AWS | VPC | security_group_ingress_anywhere_ssh | VPC security group rules should not permit ingress from '0.0.0.0/0' to port 22 (SSH) |
| AWS | VPC | vpc_flow_log | VPC flow logging should be enabled |
Whereas the rules included in the Regula rules library are generally applicable, we've built rule examples that look at tags, region restrictions, and EC2 instance usage that should be modified to fit user/organization policies.
| Provider | Service | Rule Name | Rule Description |
|---|---|---|---|
| AWS | EC2 | ec2_t2_only | Restricts instances to a whitelist of instance types |
| AWS | Tags | tag_all_resources | Checks whether resources that are taggable have at least one tag with a minimum of 6 characters |
| AWS | Regions | useast1_only | Restricts resources to a given AWS region |
What's the difference between controls and rules? A control represents an individual recommendation within a compliance standard, such as "IAM policies should not have full "*:*" administrative privileges" (CIS AWS Foundations Benchmark 1-22).
In Regula, a rule is a Rego policy that validates whether a cloud resource violates a control (or multiple controls). One example of a rule is iam_admin_policy, which checks whether an IAM policy in a Terraform file has "*:*" privileges. If it does not, the resource fails validation.
Controls map to sets of rules, and rules can map to multiple controls. For example, control CIS_1-22 and REGULA_R00002 both map to the rule iam_admin_policy.
Controls can
be specified within the rules: just add controls set.
# Rules must always be located right below the `rules` package.
package rules.my_simple_rule
# Simple rules must specify the resource type they will police.
resource_type = "aws_ebs_volume"
# Controls.
controls = {"CIS_1-16"}
# Rule logic
...Here's a snippet of test results from a Regula report. The output is from an example GitHub Action:
{
"result": [
{
"expressions": [
{
"value": {
"controls": {
"CIS_1-22": {
"rules": [
"iam_admin_policy"
],
"valid": false
},
},
"rules": {
"iam_admin_policy": {
"resources": {
"aws_iam_policy.basically_allow_all": {
"id": "aws_iam_policy.basically_allow_all",
"message": "invalid",
"type": "aws_iam_policy",
"valid": false
},
"aws_iam_policy.basically_deny_all": {
"id": "aws_iam_policy.basically_deny_all",
"message": "",
"type": "aws_iam_policy",
"valid": true
}
},
"valid": false
},
"summary": {
"controls_failed": 2,
"controls_passed": 12,
"rules_failed": 2,
"rules_passed": 8,
"valid": false
}
},
"text": "data.fugue.regula.report",
"location": {
"row": 1,
"col": 1
}
}
]
}
]
}
8 rules passed, 2 rules failed
12 controls passed, 2 controls failed
##[error] 2 rules failed
##[error]Docker run failed with exit code 1
These are the important bits:
- Summary
- Controls
- Rules
The summary block contains a breakdown of the compliance state of your Terraform files. In the output above, the Terraform violated 2 rules and 2 controls, so the CI/CD test as a whole failed. You'll also see this information in the last few lines of the output.
Regula shows you compliance results for both controls and rules, in addition to which specific resources failed. Above, in the controls block, you can see that the Terraform in the example is noncompliant with CIS_1-22, and the mapped rules that failed are listed underneath (in this case, iam_admin_policy).
In the rules block further down from controls, each rule lists the resources that failed. Above, you'll see that the resource aws_iam_policy.basically_allow_all was the one that failed the mapped rule -- as noted by "valid": false. In contrast, aws_iam_policy.basically_deny_all passed.
You can see the full example report in this GitHub Action log. For a detailed explanation of the report, see the regula-ci-example README.
Regula is designed to be easy to run in CI. We provide a GitHub Action that can be easily added to your repository:
https://github.com/fugue/regula-action
Setting up Regula with different CI/CD solutions such as Jenkins, CodePipeline, CircleCI, TravisCI, and others would follow a similar pattern. This repository contains an example:
https://github.com/fugue/regula-ci-example
bin/: the main Regula script that callsterraform&opa.lib/: the OPA library code to evaluate rules and mangle input.rules/: a collection of rules. We may split this up further as the number of rules increases.examples/: a collection of example rules that you can use as inspiration for your own rules.scripts/: scripts for development; currently only a script to generate test input.tests/:tests/lib: internal tests for the library.tests/rules/: tests for the various rules.tests/rules/inputs: terraform files that can be used to generate Rego files.tests/examples/: tests for the example rules.tests/examples/inputs: input files for the example rules.
If you would like to add a rule, we recommend starting with a test.
Put your terraform code in a file in tests/rules/inputs; for example
kms_rotate_infra.tf.
From this, you can generate a mock input by running:
bash scripts/generate-test-inputs.sh
The mock input will then be placed in a .rego file with the same name,
in our case kms_rotate_infra.tf.
It is then customary to add the actual tests in a name with the same file,
but outside of the inputs/ subdirectory. In this case, that would be
here.
Once you have generated the mock input, it is easy to debug a rule with
fregot. Fire up fregot with the right directories and set a breakpoint on
the rule you are trying to debug:
$ fregot repl lib rules tests
F u g u e R E G O T o o l k i t
fregot v0.7.2 repl - use :help for usage info
repl% :break data.rules.t2_only.allow
Now, we can just evaluate the entire report with the mock input. If your rule is triggered, that will drop you into a debug prompt:
repl% data.fugue.regula.report with input as data.tests.rules.t2_only.mock_input
19| valid_instance_types[input.instance_type]
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
From here, you can evaluate anything in context; such as input to look at the
resource, or any other auxiliary rules such as valid_instance_types in this
example.
In some cases, you may want to produce the steps that Regula performs manually. If that is something you want to step through, this section is for you.
We first need to obtain a JSON-formatted terraform plan. In order to do get that, you can use:
terraform init
terraform plan -refresh=false -out=plan.tfplan
terraform show -json plan.tfplan >input.json
This gives you input.json. Now you can test this input against the rules by
evaluating data.fugue.regula.report with OPA. In order to do that, point OPA
to the input file, and the regula project directory.
opa eval -d /path/to/regula --input input.json 'data.fugue.regula.report'
Or using fregot:
fregot eval --input input.json 'data.fugue.regula.report' . | jq
If all goes well, you should now see the results for each rule.