Skip to content
This repository has been archived by the owner on Jun 8, 2022. It is now read-only.

Commit

Permalink
Merge pull request #53 from axelsimon/secure_payloads
Browse files Browse the repository at this point in the history
Secure payload documentation
  • Loading branch information
lukehinds committed Oct 13, 2020
2 parents 32bd1da + dbc6c8a commit 62571ee
Show file tree
Hide file tree
Showing 2 changed files with 235 additions and 6 deletions.
6 changes: 0 additions & 6 deletions docs/user_guide/encrypted_payload.rst

This file was deleted.

235 changes: 235 additions & 0 deletions docs/user_guide/secure_payloads.rst
@@ -0,0 +1,235 @@
Secure Payloads
================

.. warning::
This page is still under development and not complete. It will be so until
this warning is removed.

Secure payloads offer the ability to provision encrypted data to an enrolled node.
This encrypted data can be used to deliver secrets needed by the node such as
keys, passwords, certificate roots of trust, etc.

Secure payloads are for anything which requires strong confidentiality and
integrity to bootstrap your system.

The payload itself is encrypted and sent via the Keylime Tenant CLI (or rest API)
to the Keylime Agent. The Agent also sends part of the key needed to decrypt the
payload, a key share, called the `u_key` or user key. Only when the Agent has
passed its enrolment criteria (including any `tpm_policy` or IMA whitelist),
will the other key share of the decryption key, called the `v_key` or verification
key, be passed to the Agent by the Keylime Verifier to decrypt the payload.

.. note:: An alternative to secure payloads is to deliver the encrypted data to
the node through some other mechanism like `cloud-init` or pre-embedded in a
disk image. The Keylime protocol described above will still run to derive
the decryption key for this data, but the data itself will never been seen
or transported by Keylime. This guide does not discuss this method.

Keylime offers two modes for sending secure payloads: single file encryption
and certificate package mode. In the following sections we describe each. If
you're interested in using the more advanced certificate package mode, we
recommend you also read the Single File Encryption section as it contains
configuration options and other information that both modes share.

Single File Encryption
----------------------

In this mode, a file you specify to the `keylime_tenant` application with the
`-f` option will be encrypted by the Tenant using the bootstrap key and securely
delivered to the Agent. Once the Keylime protocol with the Tenant and Verifier
has completed, the Keylime Agent will decrypt this file and place it in
`/var/lib/keylime/secure/decrypted_payload` This is the default file name, but
you can adjust the name of this file using the `dec_payload_file` option in
`keylime.conf`. You can also optionally specify a zip file as the file to be
securely delivered. If the `extract_payload_zip` option in `keylime.conf` is
set (which it is by default), then Keylime will automatically extract the zip
file to `/var/lib/keylime/secure/unzipped`. Finally, Keylime can also execute a
script contained in the zip file once it has been unzipped. You can think of
this as a very simple form of `cloud-init <https://cloudinit.readthedocs.io/>`_.
By default this script is called `autorun.sh`. You can override this default
with a different script name by adjusting the `payload_script` option in
`keylime.conf`. Note also that this script must be contained in the encrypted
zip file, from which it will be extraced and then placed in
`/var/lib/keylime/secure/unzipped`.

Because the keys that Keylime uses to decrypt the data and the decrypted data
itself are very sensitive, Keylime will only write those files to the
memory-backed (and therefore non-persistent) `/var/lib/keylime/secure`
directory. This is a bind-mounted tmpfs partition. As such, depending on how
large your payload is, you may need to increase the size of this mounted
partition by adjusting the `secure_size` option in `keylime.conf`.

This simple mode of operation is suitable for many situations where the secrets
and other bootstrapping information are basic. However, there are several
features that Keylime supports like revocation and certificate management that
do not work in this mode. For those, you'll need the next mode: Certificate
Package Mode.


Certificate Package Mode
------------------------

This mode of Keylime automates many common actions that tenants will want to
take when provisioning their Agents. First, Keylime can create an X509
certificate authority (CA) using `keylime_ca -d listen` and then issue
certificates and the corresponding private keys to each provisioned node. This
CA lives on the same host where the tenant issues the `keylime_ca` command and
can be used to bootstrap many other security solutions like mutual TLS or SSH.
To use this mode, pass the `--cert` option and a directory where the CA is
located as the parameter to this option. Keylime will then create a certificate
for the node (with the common name set to the Agent's UUID) and then create a
zip file containing the newly generated X509 certificates, trust roots, and
private keys. It then uses the same process for single file encryption as
described above to securely deliver all the keys to the Agent. Optionally, the
user can specify with the `--include` option a directory of additional files to
be put into the certification package zip and securely delivered to the Agent.

This mode of operation also natively supports certificate revocation. If the Keylime
Verifier detects an Agent that no longer satisfies its integrity policy (e.g., it booted
an authorized kernel or ran an unauthorized binary not on the IMA whitelist), it
will create a signed revocation notification. These revocation notifications are
signed by a special certificate/private key called the RevocationNotifier. Keylime
will automatically create this certificate and pass it to the verifier when you add
a new Agent to the verifier. Keylime will also include the public certificate for
this key in the zip it sends to the Agent. This way Agents can validate the
revocation notifications they receive from the verifier.

By default all Keylime Agents listen for these revocation notifications (see
the `listen_notifications` option in `keylime.conf`). Using the keys in the
unzipped certificate package, Agents can check that the revocations are valid.
Keylime Agents can also take actions in response to a valid revocation.
You can configure these actions by putting additional files into the delivered zip
file using `--include`.

Revocation actions are small Python scripts that will run on an Agent when a valid
revocation is received. They should contain an `execute` function that takes
one argument. This argument is a Python dictionary of metadata that can be used
to tailor what the revocation action does. In the cert package mode, Keylime
will specify the certificate serial number and common name (aka UUID) of the node
that has failed its integrity check inside this metadata passed to the revocation
action. For example, you can use this info to revoke the the offending X509
certificate.

One subtlety to revocation actions is that they are not intended for the Agent
that has been revoked. If an Agent has failed its integrity check, then we
really can't trust that it won't ignore the revocations and do arbitrarily
malicious things. So, revocation actions are for other well-behaving Agents in
the system to take action against the revoked Agent. For example, by revoking
its certificate as described above or blacklisting it from the network, etc.

There are several revocation actions implemented in Keylime that you can look at
to get an idea of how to write your own. See the files starting with `local_action`
in https://github.com/keylime/keylime/tree/master/auto-ipsec/libreswan/src

There are some conventions to specifying revocation actions. As described above,
their names must start with `local_action` to be executed. They also must be
listed (without `.py` extensions) in a comma separated list in a file called
`action_list` in the zip file. For example to run `local_action_a.py` and
`local_action_b.py` the `action_list` file should contain `local_action_a,local_action_b`.

So far we've described all the details of this in fine detail, but much of this
automation will happen by default.

Certificate Package Example
---------------------------

Let's put all of the above together with an example.

For the following example, we will provision some SSH keys onto the Agent.

1. Create a directory to host the files and `autorun.sh` script. For this example, we will use the directory `payload`
2. Create an `autorun.sh` script in the `payload` directory:

```
#!/bin/bash

# this will make it easier for us to find our own cert
ln -s `ls *-cert.crt | grep -v Revocation` mycert.crt

mkdir -p /root/.ssh/
cp payload_id_rsa* /root/.ssh/
chmod 600 /root/.ssh/payload_id_rsa*
```

3. Copy the files you wish to provision into the `payload` directory.

```
$ ls payload/
autorun.sh
payload_id_rsa.pub
payload_id_rsa
```

Send the files using the Keylime Tenant tool:

```
keylime_tenant -t <agent-ip> --cert myca --include payload
```

Recall that the `--cert` option tells Keylime to create a certificate authority
at the default location `/var/lib/keylime/ca` and give this machine an X509
identity with its UUID. Keylime will also create a revocation notifier
certificate for this CA and make it available to the verifier. Finally, the
`--include` option tells Keylime to securely deliver the files in the specified
directory (`payload` in our case) along with the X509 certs to the targeted
Agent machine.

If the enrolment was been successful, you will be able to see the contents of
the `payload` directory in `/var/lib/keylime/secure/unzipped` along with the
certs and included files. You should also see the SSH keys we included made in
`/root/.ssh` directory from where the autorun.sh script was ran.

Now, let's extend this example with revocation. In this example, we're going to
execute a simple revocation action on the node that was revoked.

It is also possible to configure scripts for execution should a node fail any
given criteria (IMA measurements, for example).

To configure this, we will use our `payload` directory again.

First create a Python script with the preface of `local_action`

For example `local_action_rm_ssh.py`

Within this script create an `execute` function:

```
import os
from M2Crypto import X509
import keylime.secure_mount as secure_mount

async def execute(event):
if event['type']!='revocation':
return

serial = revocation.get("metadata",{}).get("cert_serial",None)
if revocation.get('type',None) != 'revocation' or serial is None:
logger.error("Unsupported revocation message: %s"%revocation)

# load up my own cert
secdir = secure_mount.mount()
ca = X509.load_cert('%s/unzipped/mycert.crt'%secdir)

# is this revocation meant for me?
if serial === mycert.get_serial_number():
os.remove("/root/.ssh/payload_id_rsa")
os.remove("/root/.ssh/payload_id_rsa.pub")
```

Next create the `payload` directory the `action_list` file containing
`local_action_rm_ssh` (remember not to put the `.py` extension).

.. warning::
In the above example, the node that fails its integrity check is the same one
that we're expecting to run the revocation action to delete the key. Since
the node is potentially compromised, we really can't expect that it will
actually do this and not just ignore the revocation. A more realistic
scenario for SSH keys is to provision one node with the SSH key generated
as above, then provision a second server and add `payload_id_rsa.pub` to `.ssh/authorized_keys`
using an autorun script. At this point, you can SSH from the first server to
the second one. Should the first machine fail its integrity, then an
revocation action on the second server can remove the compromised first
machine from its list of Secure machines in `.ssh/authorized_keys`

Many actions can be executed based on CA revocation. For more details
and examples, please refer to the `Agent Revocation`_ page.

0 comments on commit 62571ee

Please sign in to comment.