Cubbyhole response wrapping

The term cubbyhole comes from an Americanism where you get a "locker" or "safe place" to store your belongings or valuables. In Vault, the cubbyhole is your "locker". All secrets are namespaced under your token. If that token expires or is revoked, all the secrets in its cubbyhole are revoked as well.

It is not possible to reach into another token's cubbyhole even as the root user. This is an important difference between the cubbyhole and the key/value secrets engine. The secrets in the key/value secrets engine are accessible to any token for as long as its policy allows it.

Personas

The end-to-end scenario described in this tutorial involves two personas:

• admin with privileged permissions to create tokens
• apps trusted entity retrieving secrets from Vault

Challenge

In order to tightly manage the secrets, you set the scope of who can do what using the Vault policy and attach that to tokens, roles, entities, etc.

Think of a case where you have a trusted entity (Chef, Jenkins, etc.) which reads secrets from Vault. This trusted entity must obtain a token. If the trusted entity or its host machine was rebooted, it must re-authenticate with Vault using a valid token.

How can you securely distribute the initial token to the trusted entity?

Solution

Use Vault's cubbyhole response wrapping where the initial token is stored in the cubbyhole secrets engine. The wrapped secret can be unwrapped using the single-use wrapping token. Even the user or the system created the initial token won't see the original value. The wrapping token is short-lived and can be revoked just like any other tokens so that the risk of unauthorized access can be minimized.

What is cubbyhole response wrapping?

• When response wrapping is requested, Vault creates a temporary single-use token (wrapping token) and insert the response into the token's cubbyhole with a short TTL
• Only the expecting client who has the wrapping token can unwrap this secret
• Any Vault response can be distributed using the response wrapping

Benefits of using the response wrapping:

• It provides cover by ensuring that the value being transmitted across the wire is not the actual secret. It's a reference to the secret.
• It provides malfeasance detection by ensuring that only a single party can ever unwrap the token and see what's inside
• It limits the lifetime of the secret exposure
  • The TTL of the response-wrapping token is separate from the wrapped secret's lease TTL

Prerequisites

To perform the tasks described in this tutorial, you need to have a Vault environment. Refer to the Getting Started tutorial to install Vault. Make sure that your Vault server has been initialized and unsealed.

Launch Terminal

This tutorial includes a free interactive command-line lab that lets you follow along on actual cloud infrastructure.

Start interactive lab

Policy requirements

To perform all tasks demonstrated in this tutorial, your policy must include the following permissions:

# Manage tokens
path "auth/token/*" {
  capabilities = [ "create", "read", "update", "delete", "sudo" ]
}

# Write ACL policies
path "sys/policies/acl/*" {
  capabilities = [ "create", "read", "update", "delete", "list" ]
}

# Manage secret/dev secrets engine - for Verification test
path "secret/dev" {
  capabilities = [ "create", "read", "update", "delete", "list" ]
}

If you are not familiar with policies, complete the policies tutorial.

Lab Setup

$ vault server -dev -dev-root-token-id root

$ export VAULT_ADDR=http://127.0.0.1:8200

$ export VAULT_TOKEN=root

Scenario Introduction

An app needs to retrieve secrets from Vault at the secret/kv path. However, the app does not have a valid client token to read secrets at secret/kv. Vault admin wraps the secret values using the cubbyhole response wrapping, and sends the wrapping token to the app. The app unwraps the secrets before the wrapping token expires.

Step 1: Create and wrap a token

(Persona: admin)

When a newly created token is wrapped, Vault inserts the generated token into the cubbyhole of a single-use token, returning that single-use wrapping token. Retrieving the secret requires an unwrap operation against this wrapping token.

Step 2: Unwrap the secret

(Persona: apps)

The apps persona receives a wrapping token from the admin. In order for the apps to acquire a valid token to read secrets from secret/dev path, it must run the unwrap operation using this token.

$ VAULT_TOKEN=$WRAPPING_TOKEN vault unwrap

Key         Value
---         -----
data        map[password:my-long-password username:webapp]
metadata    map[created_time:2022-04-06T04:24:24.391513Z custom_metadata:<nil> deletion_time: destroyed:false version:1]

$ VAULT_TOKEN=$WRAPPING_TOKEN vault unwrap

$ curl --header "X-Vault-Token: $WRAPPING_TOKEN" \
   --request POST \
   $VAULT_ADDR/v1/sys/wrapping/unwrap | jq ".data.data"

{
  "password": "my-long-password",
  "username": "webapp"
}

$ curl --header "X-Vault-Token: $WRAPPING_TOKEN" \
   --request POST \
   $VAULT_ADDR/v1/sys/wrapping/unwrap | jq

$ curl --header "X-Vault-Token: $WRAPPING_TOKEN" \
   --header "X-Vault-Namespace: $VAULT_NAMESPACE" \
   --request POST \
   $VAULT_ADDR/v1/sys/wrapping/unwrap | jq ".data.data"

{
  "password": "my-long-password",
  "username": "webapp"
}

$ curl --header "X-Vault-Token: $WRAPPING_TOKEN" \
   --header "X-Vault-Namespace: $VAULT_NAMESPACE" \
   --request POST \
   $VAULT_ADDR/v1/sys/wrapping/unwrap | jq

Default policy

The cubbyhole secrets engine is mounted at the cubbyhole/ prefix by default. The secrets you store in the cubbyhole/ path are tied to your token and all tokens are permitted to read and write to the cubbyhole secrets engine by the default policy. Every token has the default policy attached unless you set the -no-default-policy flag.

$ vault policy read default

The default policy grants permission to perform all operations against the cubbyhole/ path. Also notice that the default policy permits update operation on the sys/unwrapping/ paths.

...snip...

# Allow a token to manage its own cubbyhole
path "cubbyhole/*" {
    capabilities = ["create", "read", "update", "delete", "list"]
}

# Allow a token to wrap arbitrary values in a response-wrapping token
path "sys/wrapping/wrap" {
    capabilities = ["update"]
}

# Allow a token to look up the creation time and TTL of a given
# response-wrapping token
path "sys/wrapping/lookup" {
    capabilities = ["update"]
}

# Allow a token to unwrap a response-wrapping token. This is a convenience to
# avoid client token swapping since this is also part of the response wrapping
# policy.
path "sys/wrapping/unwrap" {
    capabilities = ["update"]
}

...snip...

Test the default policy

1. To better demonstrate the cubbyhole secrets engine, create a token with only default policy attached.

   $ vault token create -policy=default \
       -format=json | jq -r ".auth.client_token" > test_token.txt
   

2. Write some test data in the token's cubbyhole using the test token.

   $ VAULT_TOKEN=$(cat test_token.txt) vault write cubbyhole/private mobile="123-456-7890"
   Success! Data written to: cubbyhole/private
   

3. Read back the secret you just wrote. It should return the secret.

   $ VAULT_TOKEN=$(cat test_token.txt) vault read cubbyhole/private
   
   Key       Value
   ---       -----
   mobile    123-456-7890
   

4. Now, try to read the cubbyhole/private path.

   $ vault read cubbyhole/private
   

   Note

   The command uses the token stored in the VAULT_TOKEN environment variable instead of test_token.txt. When in doubt, you can use the vault token lookup command to see which token you are using.

   Output:

   No value found at cubbyhole/private
   

   Cubbyhole secret backend provide an isolated secrete storage area for an individual token where no other token can violate.

Help and Reference

• Cubbyhole
• Response Wrapping