Policies

Violations

Warnings

AWSSEC-0001: Encrypt S3 Buckets

Severity: Violation

Resources: Any Resource

S3 Buckets must have server-side encryption enabled. See https://www.terraform.io/docs/backends/types/s3.html#encrypt.

While the security benefits of server-side bucket encryption are nebulous given practical threat scenarios, those wishing to apply such a control may do so with this policy. You may also be required to enforce this as a compliance checkbox.

Rego

package terraform_encrypt_s3_buckets

import data.util_functions

policyID := "AWSSEC-0001"

violation[{"policyId": policyID, "msg": msg}] {
  resource := input.resource.aws_s3_bucket
  a_resource := resource[name]
  not util_functions.has_key(a_resource, "server_side_encryption_configuration")

  msg := sprintf("Missing S3 encryption for `%s`. Required flag: `server_side_encryption_configuration`", [name])
}

source: https://github.com/RallyHealth/conftest-policy-packs/policies/terraform/encrypt_s3_buckets/src.rego

AWSSEC-0002: EC2 Instances Must Use Instance Metadata Service Version 2

Severity: Violation

Resources: Any Resource

EC2 instances must use instance metadata service version 2 (IMDSv2) to prevent server-side request forgery (SSRF) attacks.

Set http_tokens to required in the metadata-options.

AWS released v2 of the instance metadata service as a response to the 2019 Capital One breach. IMDSv2 helps prevent SSRF from being executed against instance metadata, preventing attackers from stealing instance credentials via a vulnerability in a web server application.

IMDSv2 adds a session token in the X-aws-ec2-metadata-token header that must be present to retrieve any information from instance metadata. This occurs automatically for systems using the AWS CLI. Systems making direct curl requests to instance metadata must modify their requests to the following format:

# Get a token with a 60-second lifetime
TOKEN=`curl -X PUT "http://196.254.196.254/latest/api/token" -H "X-aws-ec2-metadata-token-ttl-seconds: 60"`
# Make instance metadata request
curl http://169.254.169.254/latest/meta-data/profile -H "X-aws-ec2-metadata-token: $TOKEN"

Rego

package terraform_ec2_imdsv2_required

import data.util_functions

policyID := "AWSSEC-0002"

violation[{"policyId": policyID, "msg": msg}] {
  resource := input.resource.aws_instance
  aws_resource := resource[resource_name]

  # Check for metadata options
  util_functions.has_key(aws_resource, "metadata_options")
  metadata := aws_resource.metadata_options

  # Check for http_tokens and correct value
  util_functions.has_key(metadata, "http_tokens")
  metadata.http_tokens != "required"

  msg := sprintf("Instance metadata version 2 not enabled for resource `aws_instance.%s`. Add a `metadata_options` block with `http_tokens` set to `required`.", [resource_name])
}

violation[{"policyId": policyID, "msg": msg}] {
  resource := input.resource.aws_instance
  aws_resource := resource[resource_name]

  # Check for metadata options
  util_functions.has_key(aws_resource, "metadata_options")
  metadata := aws_resource.metadata_options

  # If no http_tokens field, flag it
  not util_functions.has_key(metadata, "http_tokens")

  msg := sprintf("Instance metadata version 2 not enabled for resource `aws_instance.%s`. Add a `metadata_options` block with `http_tokens` set to `required`.", [resource_name])
}

violation[{"policyId": policyID, "msg": msg}] {
  resource := input.resource.aws_instance
  aws_resource := resource[resource_name]

  # Check for metadata_options
  not util_functions.has_key(aws_resource, "metadata_options")

  msg := sprintf("Instance metadata version 2 not enabled for resource `aws_instance.%s`. Add a `metadata_options` block with `http_tokens` set to `required`.", [resource_name])
}

source: https://github.com/RallyHealth/conftest-policy-packs/policies/terraform/imdsv2_required/src.rego

AWSSEC-0003: RDS Instances May Not Be Public

Severity: Violation

Resources: Any Resource

RDS instances must block public access. The publicly_accessible attribute, if defined, must be set to false. The attribute is false by default if not specified.

See https://registry.terraform.io/providers/hashicorp/aws/latest/docs/resources/db_instance#publicly_accessible.

Rego

package terraform_no_public_rds

import data.util_functions

policyID := "AWSSEC-0003"

has_public_attribute(resource) {
  util_functions.has_key(resource, "publicly_accessible")
}

violation[{"policyId": policyID, "msg": msg}] {
  resource := input.resource.aws_db_instance
  a_resource := resource[name]
  has_public_attribute(a_resource)
  a_resource.publicly_accessible != false

  msg := sprintf("RDS instances must not be publicly exposed. Set `publicly_accessible` to `false` on aws_db_instance.`%s`", [name])
}

source: https://github.com/RallyHealth/conftest-policy-packs/policies/terraform/no_public_rds/src.rego

AWSSEC-0004: Block Public Access of S3 Buckets

Severity: Violation

Resources: Any Resource

S3 Block Public Access ensures that objects in a bucket never have public access, now and in the future. S3 Block Public Access settings override S3 permissions that allow public access. If an object is written to an S3 bucket with S3 Block Public Access enabled, and that object specifies any type of public permissions via ACL or policy, those public permissions are blocked.

Unintentionally exposed S3 Buckets are a frequent source of data breaches and restricting public access helps prevent unintended data exposure.

See https://registry.terraform.io/providers/hashicorp/aws/latest/docs/resources/s3_bucket_public_access_block.

Rego

package terraform_block_public_acls_s3

import data.util_functions

policyID := "AWSSEC-0004"

violation[{"policyId": policyID, "msg": msg}] {
  input.resource.aws_s3_bucket[bucket_name]

  check_is_bucket_missing_public_access_block(bucket_name)
  msg := sprintf("Public access is not explicitly disabled on the following S3 Bucket: `%s`. You must set an `[s3_bucket_public_access_block](https://registry.terraform.io/providers/hashicorp/aws/latest/docs/resources/s3_bucket_public_access_block)`.", [bucket_name])
}

violation[{"policyId": policyID, "msg": msg}] {
  input.resource.aws_s3_bucket[bucket_name]
  acl_details := get_public_access_block_for_bucket(bucket_name)

  check_if_block_public_acls_is_missing(acl_details)
  msg := sprintf("Missing s3 block public access for the following resource(s): `%s`. Required flag: `%s`", [bucket_name, "block_public_acls"])
}

violation[{"policyId": policyID, "msg": msg}] {
  input.resource.aws_s3_bucket[bucket_name]
  acl_details := bucket_to_publicAccess(bucket_name)

  check_public_acls_not_blocked(acl_details)
  msg := sprintf("Missing s3 block public access for the following resource(s): `%s`. Required flag: `%s`", [bucket_name, "block_public_acls"])
}

bucket_to_publicAccess(bucket_name) = public_access_block {
  input.resource.aws_s3_bucket[bucket_name]
  public_access_block := get_public_access_block_for_bucket(bucket_name)
}

check_is_bucket_missing_public_access_block(bucket_name) {
  not get_public_access_block_for_bucket(bucket_name)
}

check_if_block_public_acls_is_missing(acl_details) {
  not util_functions.has_key(acl_details, "block_public_acls")
}

check_public_acls_not_blocked(acl_details) {
  acl_details.block_public_acls == false
}

bucket_names_match(first_bucket_name, second_bucket_name) {
  first_bucket_name == second_bucket_name
}

get_public_access_block_for_bucket(bucket_name) = acl_details {
  acl_details := input.resource.aws_s3_bucket_public_access_block[_]

  # Since terraform source code is scanned (not `terraform plan`), it is reasonable to expect
  # the s3 bucket resource to exist in the same file as the s3 public access block resource.
  # "${aws_s3_bucket.example.id}" -> ["${aws_s3_bucket", "example", "id}"]
  s3_bucket_id := split(acl_details.bucket, ".")

  bucket_names_match(bucket_name, s3_bucket_id[1])
}

source: https://github.com/RallyHealth/conftest-policy-packs/policies/terraform/block_public_acls_s3/src.rego

CTNRSEC-0001: Dockerfiles Must Pull From An Approved Private Registry

Severity: Violation

Resources: Any Resource

Dockerfiles must pull images from an approved private registry and not from public repositories. The FROM statement must have a private registry prepended, e.g. "our.private.registry/..." as the value.

Rego

package docker_pull_from_registry

import data.approved_private_registries
import data.docker_utils
import data.util_functions

policyID := "CTNRSEC-0001"

cmds := ["env", "arg"]

violation[{"policyId": policyID, "msg": msg}] {
  input[i].Cmd == "from"
  val := input[i].Value
  not docker_utils.is_a_variable(val)
  not docker_utils.is_a_multistage_build(input, val[0])
  not docker_utils.from_scratch(val[0])

  not util_functions.item_startswith_in_list(val[0], approved_private_registries)
  msg := sprintf("Dockerfiles must pull images from an approved private registry (`FROM my.private.registry/...`). The image `%s` does not pull from an approved private registry. The following are approved registries: `%v`.", [val[0], approved_private_registries])
}

violation[{"policyId": policyID, "msg": msg}] {
  input[i].Cmd == "from"
  val := input[i].Value
  not docker_utils.is_a_multistage_build(input, val[0])
  docker_utils.is_a_variable(val)

  # Get variable name without the $
  variableNameWithVersion := substring(val[0], 1, -1)

  # Drop the version, if present, to make it easier to find the right variable
  variableName := split(variableNameWithVersion, ":")[0]

  input[j].Cmd == cmds[_]
  argCmd := input[j].Value[0]

  # ARG or ENV is a match for the variable we're looking for
  startswith(argCmd, variableName)

  # Grab the value of the ARGument or ENV var
  # e.g. ARG MYIMAGE=ubuntu:latest => ubuntu:latest
  argNameAndValue := split(argCmd, "=")
  imageInArg := trim(argNameAndValue[1], "\"")

  not util_functions.item_startswith_in_list(imageInArg, approved_private_registries)
  msg := sprintf("Dockerfiles must pull images from an approved private registry (`FROM my.private.registry/...`). The image `%s` in variable `%s` does not pull from an approved private registry. The following are approved registries: `%v`.", [imageInArg, argNameAndValue[0], approved_private_registries])
}

source: https://github.com/RallyHealth/conftest-policy-packs/policies/docker/deny_image_unless_from_registry/src.rego

CTNRSEC-0002: Dockerfiles Should Not Use Environment Variables For Sensitive Values

Severity: Violation

Resources: Any Resource

Docker images should not pass sensitive values through ENV or ARG variables. This binds the secret value into a layer of the Docker image and makes the secret recoverable by anyone with access to the final built image through the docker history command.

Instead, users should use the Buildkit --secret flag or a multi-stage build.

If you use a multi-stage build, an ARG or ENV with a sensitive value must not exist in the final built image.

Rego

package sensitive_keys_in_env_args

import data.docker_utils
import data.excepted_env_keys

policyID := "CTNRSEC-0002"

sensitive_env_keys = [
  "secret",
  "apikey",
  "token",
  "passwd",
  "password",
  "pwd",
  "api_key",
  "credential",
]

cmds := ["env", "arg"]

violation[{"policyId": policyID, "msg": msg}] {
  # Get all indices where cmd is 'from'
  from_stmt_indices := [index | input[i].Cmd == "from"; index := i]
  from_index := from_stmt_indices[x]

  # from_val is an array like ["my.private.registry/ubuntu:20.04", "AS", "builder"]
  from_val := input[from_index].Value
  not docker_utils.is_a_multistage_build(input, from_val[0])

  # We only care about evaluating 'env' statements that correspond to the final 'from' statement
  start := from_stmt_indices[count(from_stmt_indices) - 1]
  end := count(input)
  final_from_slice := array.slice(input, start, end)

  cmd := cmds[_]
  final_from_slice[j].Cmd == cmd
  val := final_from_slice[j].Value
  sensitive_key := sensitive_env_keys[_]
  excepted_key := excepted_env_keys[_]
  contains(lower(val[0]), sensitive_key)
  not contains(lower(val[0]), excepted_key)

  msg := sprintf("A %s key [`%s`] was found in this Dockerfile that suggets you are storing a sensitive value in a layer of your Docker image. Dockerfiles should instead use the [Buildkit --secret](https://docs.docker.com/develop/develop-images/build_enhancements/#new-docker-build-secret-information) flag or place the sensitive value in an earlier stage of a multi-stage build.", [upper(cmd), val[0]])
}

source: https://github.com/RallyHealth/conftest-policy-packs/policies/docker/sensitive_keys_in_env_args/src.rego

PKGSEC-0001: NPM Packages Must Be Published Under An Organization Scope

Severity: Violation

Resources: Any Resource

NodeJS packages are subject to typosquatting, in which a malicious package is published with a slight misspelling. The aim is to infect end users who misspell your package. While this relies on end user misconfiguration, package owners can still take steps to reduce the viability of such a mistake.

This can be avoided by scoping an organizational package beneath an organization scope. Organizations publishing packages to a private registry can use a private organization scope.

Scopes are a way of grouping related packages together, and also affect a few things about the way npm treats the package. Each npm user/organization has their own scope, and only you can add packages in your scope. This means you don't have to worry about someone taking your package name ahead of you. Thus it is also a good way to signal official packages for organizations.

Rego

package nodejs_package_must_have_org_scope

import data.approved_org_scopes
import data.packages_functions
import data.util_functions

policyID := "PKGSEC-0001"

has_org_scope(name) {
  startswith(name, "@")
}

violation[{"policyId": policyID, "msg": msg}] {
  packages_functions.is_package_json(input)
  package_name := input.name
  not has_org_scope(package_name)
  msg := sprintf("NPM packages must be wrapped beneath an organization scope (e.g. `@orgscope/mypackage`). `%s` does not use any organization scope. Approved scopes are: `%v`.", [package_name, approved_org_scopes])
}

violation[{"policyId": policyID, "msg": msg}] {
  packages_functions.is_package_json(input)
  package_name := input.name
  has_org_scope(package_name)
  org_name := substring(package_name, 1, -1)
  not util_functions.item_startswith_in_list(org_name, approved_org_scopes)
  msg := sprintf("NodeJS packages must be wrapped beneath an organization scope (e.g. `@orgscope/mypackage`). `%s` does not use an approved organization scope. Approved scopes are: `%v`.", [package_name, approved_org_scopes])
}

source: https://github.com/RallyHealth/conftest-policy-packs/policies/packages/nodejs_package_must_use_org_scope/src.rego

PKGSEC-0002: NodeJS Projects Must Use A Recent NodeJS Version

Severity: Violation

Resources: Any Resource

NodeJS projects must use a recent NodeJS release. Only 1 LTS version is active at one time, however we allow 1 previous LTS version to be used to accomodate upgrade migration periods.

You may use a non-LTS "current" NodeJS release as long as that version is more current than the most recently deprecated LTS release. For example, if the LTS version is 16 (meaning version 14 was the most recently deprecated LTS version), you may use NodeJS 14, 15, 16, or 17. Once NodeJS 18 is released as an LTS version, you may use versions 16, 17, 18, or 19.

See https://nodejs.org/en/about/releases/ for more information about Node's release schedule.

See https://docs.npmjs.com/cli/v7/configuring-npm/package-json#engines for more information about the engines field in package.json files.

:memo: This policy is "online" in that it makes an HTTP request to raw.githubusercontent.com and requires connectivity to receive a response.

Rego

package nodejs_must_use_approved_version

import data.packages_functions
import data.util_functions

policyID := "PKGSEC-0002"

nodejs_release_schedule_json := "https://raw.githubusercontent.com/nodejs/Release/main/schedule.json"

get_nodejs_releases = output {
  # This is a separate function (with no input params) to allow us to mock the call in the tests
  output := http.send({
    "url": nodejs_release_schedule_json,
    "method": "GET",
    "force_json_decode": true,
    "cache": true,
  })
}

get_latest_lts_version = latest_lts_release {
  output := get_nodejs_releases

  releases := filter_lts_releases(output)
  num_releases := count(releases)

  # This may be an LTS in the future, not the currently released "latest" LTS version
  # This would be an even-numbered current release with a start date in the future, when it becomes the LTS release
  latest_lts := releases[num_releases - 1]

  # e.g. { "codename": "", "end": "2025-04-30", "lts": "2022-10-25", "maintenance": "2023-10-18", "start": "2022-04-19" }
  release_metadata := output.body[sprintf("v%d", [latest_lts])]

  # Output is a float representing the difference in nanoseconds between the two dates
  # All we care about is whether it is positive or negative
  time_diff := determine_time_difference_between_today_and_latest_lts(release_metadata)

  # This will either return the latest LTS release or the second-latest, depending on that time difference outcome
  latest_lts_release := determine_current_lts_release(releases, time_diff)
}

filter_lts_releases(output) = sorted_releases {
  # We want the latest LTS release
  # This comprehension will filter out versions that do not contain an 'lts' field
  # Leaving us with only the LTS releases
  # We prune the 'v' in the versions and convert them to numbers
  # e.g. "v16" -> 16
  # and sort so the highest number (latest release) is at the end of the list
  releases := [to_number(substring(version, 1, -1)) | record := output.body[version]; record.lts]
  sorted_releases := sort(releases)
}

determine_time_difference_between_today_and_latest_lts(release_metadata) = time_diff {
  today := time.now_ns()

  # Layout comes from requirements in Golang time.Parse
  # https://golang.org/pkg/time/#Parse
  release_time := time.parse_ns("2006-01-02", release_metadata.start)
  time_diff := today - release_time
}

determine_current_lts_release(sorted_releases, time_diff) = sorted_releases[count(sorted_releases) - 1] {
  # If release time is in the future, use the second-latest LTS, which would be the current LTS version
  # If time diff is positive, then LTS release comes out in the future
  # If time diff is negative, then the LTS release came out before this moment
  time_diff >= 0
} else = sorted_releases[count(sorted_releases) - 2] {
  true
}

has_node_engine(resource) {
  util_functions.has_key(resource, "engines")
  util_functions.has_key(resource.engines, "node")
}

is_unapproved_node_version(engine_string) {
  # This is going to strip symbols from the string then try to convert it into a number.
  # It is possible to have multiple version constraints in the node engine string.
  # This function attempts to require every version in the string to comply with the LTS policy.
  # e.g. ">=10 <15" => ["10", "15"] (possible_multiple_versions variable)
  # It will trigger a violation if any version number in the string is outside the acceptable range.

  # List any possible symbols or other characters we don't care about that are valid in the engine string
  engine_string_no_symbols := strings.replace_n(
    {
      "<": "",
      ">": "",
      "=": "",
      "~": "",
    },
    engine_string,
  )

  possible_multiple_versions := split(engine_string_no_symbols, " ")
  numbers_outside_acceptable_range(possible_multiple_versions)
}

numbers_outside_acceptable_range(number_string_list) {
  some i
  version := to_number(number_string_list[i])
  version < latest_lts_version - 2
}

missing_minimum_version_constraint(engine_string) {
  index_of_minimum_constraint := indexof(engine_string, ">")
  index_of_minimum_constraint == -1
}

latest_lts_version := get_latest_lts_version

violation[{"policyId": policyID, "msg": msg}] {
  packages_functions.is_package_json(input)
  not has_node_engine(input)
  msg := sprintf("NodeJS projects must enforce a Node engine version within the last 2 LTS releases. This project does not enforce any Node engine version. See the [NodeJS documentation](https://docs.npmjs.com/cli/v7/configuring-npm/package-json#engines) on how to require a Node version. You must use a version of Node >= %d.", [latest_lts_version - 2])
}

violation[{"policyId": policyID, "msg": msg}] {
  packages_functions.is_package_json(input)
  has_node_engine(input)
  is_unapproved_node_version(input.engines.node)
  msg := sprintf("NodeJS projects must enforce a Node engine version within the last 2 LTS releases. This project uses an older NodeJS version in its engine constraint: [`%s`]. You must use a version of Node >= %d.", [input.engines.node, latest_lts_version - 2])
}

violation[{"policyId": policyID, "msg": msg}] {
  packages_functions.is_package_json(input)
  has_node_engine(input)
  missing_minimum_version_constraint(input.engines.node)
  msg := sprintf("NodeJS projects must enforce a Node engine version within the last 2 LTS releases. This project does not enforce a minimum Node engine version. You must use a version of Node >= %d.", [latest_lts_version - 2])
}

source: https://github.com/RallyHealth/conftest-policy-packs/policies/packages/nodejs_must_use_recent_version/src.rego

PKGSEC-0003: NPM Packages Must Be Published To Approved Registry

Severity: Violation

Resources: Any Resource

NodeJS packages with a publishConfig object must have the registry field set to an approved organizational registry.

For more information about the registry field of the publishConfig object, see https://docs.npmjs.com/cli/v7/using-npm/registry#how-can-i-prevent-my-package-from-being-published-in-the-official-registry.

Rego

package nodejs_package_must_use_org_publish_config

import data.approved_publishConfig_registries
import data.packages_functions
import data.util_functions

policyID := "PKGSEC-0003"

approved_registries := {registry_name | registry_name := approved_publishConfig_registries[i]}

violation[{"policyId": policyID, "msg": msg}] {
  packages_functions.is_package_json(input)
  util_functions.has_key(input, "publishConfig")

  publish_config := input.publishConfig
  util_functions.has_key(publish_config, "registry")

  not approved_registries[publish_config.registry]

  msg := sprintf("NPM packages must have a `publishConfig` field set to an approved registry. An unapproved registry is listed. Approved registries are: `%v`.", [approved_publishConfig_registries])
}

violation[{"policyId": policyID, "msg": msg}] {
  packages_functions.is_package_json(input)
  util_functions.has_key(input, "publishConfig")

  publish_config := input.publishConfig
  not util_functions.has_key(publish_config, "registry")

  msg := sprintf("NPM packages must have a `publishConfig` field set to an approved registry. No `registry` is set. Approved registries are: `%v`.", [approved_publishConfig_registries])
}

violation[{"policyId": policyID, "msg": msg}] {
  packages_functions.is_package_json(input)
  not util_functions.has_key(input, "publishConfig")

  msg := sprintf("NPM packages must have a `publishConfig` field set to an approved registry. No `publishConfig` is set. Approved registries are: `%v`.", [approved_publishConfig_registries])
}

source: https://github.com/RallyHealth/conftest-policy-packs/policies/packages/nodejs_use_publishConfig/src.rego

AWSSEC-0005: Resources Must Use Required Tags

Severity: Warning

Resources: Any Resource

AWS resources should be tagged with a minimum set of organization tags for logistical purposes.

As this policy is executing on Terraform source code, this policy is reported as a "warn" instead of a "violation." This policy is best evaluated against a Terraform plan, in which case the policy may need to be adapted to correctly parse a resource in a plan file.

Rego

package terraform_required_tags

import data.minimum_required_tags

policyID := "AWSSEC-0005"

tags_contain_proper_keys(tags) {
  keys := {key | tags[key]}
  minimum_tags_set := {x | x := minimum_required_tags[i]}
  leftover := minimum_tags_set - keys

  # If all minimum_tags exist in keys, the leftover set should be empty - equal to a new set()
  leftover == set()
}

warn[msg] {
  resource := input.resource[resource_type]
  tags := resource[name].tags

  # Create an array of resources, only if they are missing the minimum tags
  resources := [sprintf("%v.%v", [resource_type, name]) | not tags_contain_proper_keys(tags)]

  resources != []
  msg := sprintf("%s: Invalid tags (missing minimum required tags) for the following resource(s): `%v`. Required tags: `%v`", [policyID, resources, minimum_required_tags])
}

source: https://github.com/RallyHealth/conftest-policy-packs/policies/terraform/required_tags/src.rego