draft-l3dl-signing/draft-ymbk-lsvr-l3dl-signing.xml
2020-05-07 14:16:11 -07:00

530 lines
17 KiB
XML

<?xml version="1.0" encoding="utf-8"?>
<!DOCTYPE rfc SYSTEM "rfc2629.dtd">
<?rfc comments="yes"?>
<?rfc compact="yes"?>
<?rfc subcompact="no"?>
<?rfc inline="yes"?>
<?rfc sortrefs="yes"?>
<?rfc symrefs="yes"?>
<?rfc toc="yes"?>
<?rfc tocdepth="6"?>
<?rfc tocindent="yes"?>
<?rfc tocompact="yes"?>
<rfc category="std" docName="draft-ymbk-lsvr-l3dl-signing-01" ipr="trust200902">
<front>
<title>Layer 3 Discovery and Liveness Signing</title>
<author fullname="Randy Bush" initials="R." surname="Bush">
<organization>Arrcus &amp; IIJ</organization>
<address>
<postal>
<street>5147 Crystal Springs</street>
<city>Bainbridge Island</city>
<region>WA</region>
<code>98110</code>
<country>United States of America</country>
</postal>
<email>randy@psg.com</email>
</address>
</author>
<!--
<author fullname="Russ Housley" initials="R" surname="Housley">
<organization abbrev="Vigil Security">Vigil Security, LLC</organization>
<address>
<postal>
<street>516 Dranesville Road</street>
<city>Herndon</city>
<region>VA</region>
<code>20170</code>
<country>USA</country>
</postal>
<email>housley@vigilsec.com</email>
</address>
</author>
-->
<author initials="R." surname="Austein" fullname="Rob Austein">
<organization abbrev="Arrcus">Arrcus, Inc.</organization>
<address>
<email>sra@hactrn.net</email>
</address>
</author>
<date />
<abstract>
<t>The Layer 3 Discovery and Liveness protocol OPEN PDU may contain
a key and a certificate, which can be used to verify signatures on
subsequent PDUs. This document describes two mechanisms based on
digital signatures, one that is Trust On First Use (TOFU), and one
that uses certificates to provide authentication as well as session
integrity.</t>
</abstract>
<note title="Requirements Language">
<t>The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and
"OPTIONAL" in this document are to be interpreted as described in
BCP&nbsp;14 <xref target="RFC2119"/> <xref target="RFC8174"/> when,
and only when, they appear in all capitals, as shown here.</t>
</note>
</front>
<middle>
<section anchor="intro" title="Introduction">
<t>This draft is being published without incorporating changes from
an excellent security review. This is being done so a couple of
other drafts can reference it. While all comments will, of course,
be appreciated, readers may want to wait for the -01 version.</t>
<t>The Layer 3 Discovery and Liveness protocol <xref
target="I-D.ietf-lsvr-l3dl"/> OPEN PDU contains an algorithm
specifier, a key, and a certificate, which can be used to verify
signatures on subsequent PDUs. This document describes two methods
of key generation and signing for use by L3DL, Trust On First Use
(TOFU) and a PKI-based mechanism to provide authentication as well
as session integrity.</t>
<t>The Key in the OPEN PDU SHOULD be the public key of an asymmetric
key pair. The sender signs with the private key, of course. The
device sending the OPEN may use one key for all links, a different
key for each link, or some aggregation(s) thereof.</t>
<t>In the TOFU method the OPEN key is generated on the sending
device, believed without question by the receiver, and used to
verify all subsequent PDUs from the same sender with the same Key
Type.</t>
<t>With the PKI-mechanism, an enrollment step is performed. The
public key is put into a certificate <xref target="RFC5280"/>, which
is signed by the the operational environment's trust anchor. In
this way, the relying party can be confident that the public key is
under control of the identified L3DL protocol entity.</t>
<t>To the receiver verifying signatures on PDUs, the two methods are
indistinguishable; the key provided in the OPEN PDU is used to
verify the signatures of subsequent PDUs. The difference that
PKI-based keys may be verified against the trust anchor when the
OPEN PDU is received.</t>
<t>In the PKI method the OPEN key MUST be verified against the trust
anchor for the operational domain. It is then used to verify all
subsequent PDUs in the session.</t>
</section>
<section anchor="tofu" title="Trust On First Use Method">
<t>There are three parts to using a key: signing PDUs, verifying
the OPEN PDU, and verifying subsequent PDUs.</t>
<section anchor="tofu-pdu-signing" title="Signing a PDU">
<t>All signed PDUs are generated in the same way:</t>
<t>
<list style="symbols">
<t>
Compose the PDU, with all fields including "Sig Type" and
"Signature Length" set, but omitting the trailing
"Signature" field itself. The Certificate Length should
be zero and the Certificate field should be empty. This
is the "message to be signed" for purposes of the
signature algorithm.
</t>
<t>
Generate the signature as specified for the chosen signature
suite, using the private member of the asymmetric key pair.
In general this will involve first hashing the "message to
be signed" then signing the hash, but the precise details
may vary with the specific algorithm. The result will be a
sequence of octets, the length of which MUST be equal to the
setting of the "Signature Length" field.
</t>
<t>
Construct the complete message by appending the signature
octets to the otherwise complete message composed above.
</t>
</list>
</t>
<t>
In the case of the OPEN PDU, the message to be signed will
include the public member of the asymmetric keypair, but as
far as the signature algorithm is concerned that's just
payload, no different from any other PDU content.
</t>
</section>
<section anchor="tofu-open-verifying" title="Verifying the OPEN PDU">
<t>
The process for verifying an OPEN PDU is slightly different
from the process for verifying other PDU types, because the
OPEN PDU also establishes the session key.
</t>
<t>
<list style="symbols">
<t>
Verify that the PDU is syntactically correct, and extract
the Auth Type, Key, Sig Type, and Signature fields.
</t>
<t>
Verify that Auth Type and Sig Type refer to the same
algorithm suite, and that said algorithm suite is one that
the implementation understands.
</t>
<t>
Construct the "message to be verified" by truncating the PDU
to remove the Signature field (in practice this should not
require copying any data, just subtract the signature length
from the PDU length).
</t>
<t>
Verify the message constructed above against the public key
using the rules for the specific signature suite.
</t>
<t>
Record Auth Type and Key as this sessions's authentication
type and session key, for use in verifying subseuqent PDUs.
</t>
</list>
</t>
<t>
If any of the above verification steps fail, generate an error
using error code 2 ("Authorization failure in OPEN").
</t>
<!--
Why are we using a different error code for failures in OPEN
PDUs than we do in other PDUs? We don't want to provide an
oracle, so we want to return the same error code for any
verification failure for a particular PDU, so the only effect
would be to have all failures in OPEN PDUs return a different
single error code than all failures in any other PDU would
use, which doens't seem useful.
-->
</section>
<section anchor="tofu-other-verifying" title="Verifying Other PDUs">
<t>
The process for verifying non-OPEN PDUs is slightly simpler,
but follows the same basic pattern as for OPEN PDUs.
</t>
<t>
<list style="symbols">
<t>
Verify that the PDU is syntactically correct, and extract
the Sig Type and Signature fields.
</t>
<t>
Verify that Sig Type refers to the same algorithm suite as
the Auth Type recorded during verification of the OPEN PDU.
</t>
<t>
Construct the "message to be verified" by truncating the PDU
to remove the Signature field.
</t>
<t>
Verify the message constructed above against the recorded
session key using the rules for the specific signature
suite.
</t>
</list>
</t>
<t>
If any of the above verification steps fail, generate an error
using error code 3 ("Signature failure in PDU").
</t>
<!--
See note in previous section regarding error codes.
-->
</section>
</section>
<section anchor="pki" title="Public Key Infrastructure Method">
<t>
Using a PKI is almost the same as using TOFU, but with one
additional step: during verification of an OPEN PDU, after
extracting the Key field from the PDU but before attempting to
use it to verify the PDU's signature, the receiver MUST verify
the received key against the PKI to confirm that it's an
authorized key.
</t>
<t>
Generating an OPEN PDU using the PKI method requires a
certificate, which must be supplied via out of band
configuration. The certificate is a signature of the public key
to be sent in the Key field of the OPEN PDU, signed by the trust
anchor private key.
</t>
<t>
Verifying an OPEN PDU using the PKI method requires the public
key of the trust anchor, which the receiver uses to verify the
certificate, thereby demonstrating that the supplied is
represents an authorized L3DL speaker in this administrative
domain.
</t>
<t>
We use the term "certificate" here in the generic sense. These
are not X.509 certificates: X.509 is much more complicated than
we need for L3DL. The certificates used here are just
signatures of one key (the session key supplied in the Key field
of the OPEN PDU) by another key (the trust anchor).
</t>
<section anchor="pki-open-signing" title="Signing OPEN PDU with PKI">
<t>
Generating and signing the OPEN PDU with the PKI method is
almost the same as in <xref target="tofu-pdu-signing"/>. The
only difference is that the PKI method MUST supply the
appropriate certificate in the Certificate field.
</t>
<t>
Note that the Auth Type field applies to both the Key and
Certificate fields. That is: the certificate uses the same
certificate suite as the session keys, L3DL does not support
cross-algorithm-suite certification.
</t>
</section>
<section anchor="verify-pki-open" title="Verifying OPEN PDU with PKI">
<t>
Verifying the OPEN PDU with PKI is similar to verifying with
TOFU as described in <xref target="tofu-open-verifying"/>, but
includes one critical extra step:
</t>
<t>
After extracting the Key field from the PDU but before
verifying the Signature, extract the Certificate field and
verfiy that the Certificate is a valid signature of the Key
field, according to the rules for the signature suite
specified by Auth Type. If this step fails, handle as in
<xref target="tofu-open-verifying"/>.
</t>
</section>
</section>
<section anchor="local-policy" title="Local Policy">
<t>
Whether to use TOFU, PKI, or no signatures at all is a matter of
local policy, to be decided by the operator. The useful
policy combinations for Key and Certificate are probably:
</t>
<t>
<list style="symbols">
<t>
Not signing: sender need not sign, receiver does not check.
</t>
<t>
Require TOFU: sender MUST supply key and receiver MUST
check, certificate not needed and ignored if sent.
</t>
<t>
Allow TOFU: sender must supply key and receiver MUST check,
receiver SHOULD check certificate if supplyed by sender.
</t>
<t>
Require PKI: sender must supply key and certificate,
receiver must check both.
</t>
</list>
</t>
</section>
<section anchor="roll" title="NEWKEY, Key Roll">
<t>Modern key management allows for agility in 'rolling' to a new
key or even algorithm in case of key expiry, key compromise, or
merely prudence. Declaring a new key with an L3DL OPEN PDU would
cause serious churn in topology as a new OPEN may cause a withdraw
of previously announced encapsulations. Therefore, a gentler
rekeying is needed.</t>
<!--
protocol "Type = 8:8,Payload Length:16,New Key Type:8,New Key Length:16,New Key ...:32,New Cert Length:16,New Certificate ...:32,Old Sig Type:8,Old Signature Length:16,Old Signature ...:40"
-->
<figure>
<artwork>
0 1 2 3
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Type = 8 | Payload Length | New Key Type |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| New Key Length | New Key ... |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| | New Cert Length |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| New Certificate ... |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
| Old Sig Type | Old Signature Length | |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ +
| Old Signature ... |
+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
</artwork>
</figure>
<t>The New Key Type, New Key Length, New Key, New Cert Length, and
New Certificate field declare the replacement algorithm suite,
key, and certificate.</t>
<t>The NEWKEY PDU is signed using the current (soon to be old)
algorithm suite and key.</t>
<t>The sender and the receiver should be cautious of algorithm suite
downgrade attacks.</t>
<t>To avoid possible race conditions, the receiver SHOULD accept
signatures using either the new or old key for a configurable time
(default 30 seconds). This is intended to accommodate situations
such as senders with high peer out-degree and a single per-device
asymmetric key.</t>
<t>If the sender does not receive an ACK in the normal window,
including retransmission, then the sender MAY choose to allow a
session reset by either issuing a new OPEN or by letting the
receiver eventually have a signature failure (error code 3) on a
PDU.</t>
<t>
The rekeying operation changes the session key and algorithm
suite described in <xref target="tofu-other-verifying"/>. The
NEWKEY PDU itself is verified using the old algorithm and
session key, subsequent PDUs are verified with the new algorithm
and session key recorded after the NEWKEY PDU has been accepted.
</t>
</section>
<section anchor="security" title="Security Considerations">
<t>The TOFU method requires a leap of faith to accept the key in the
OPEN PDU, as it can not be verified against any authority. Hence it
is jokingly referred to as Married On First Date. The assurance it
does provide is that subsequent signed PDUs are from the same peer.
And data integrity is a positive side effect of the signature
covering the payload.</t>
<t>The PKI-based method offers assurance that the certificate, and
hence the keying material, provided in the OPEN PDU are authorized
by a central authority, e.g. the network's network security team. The
onward assurance of talking to the same peer and data integrity are
the same as in the TOFU method.</t>
<t>With the PKI-based method, automated device provisioning could
restrict which certificates are allowed from which peers
on a per interface basis. This would complicate key rolls. Where
one draws the line between rigidity, flexibility, and security
varies.</t>
<t>The REKEY PDU is open to abuse to create an algorithm suite
downgrade attack.</t>
</section>
<section anchor="iana" title="IANA Considerations">
<t>This document requests the IANA create a new entry in the L3DL PDU
Type registry as follows:</t>
<figure>
<artwork>
PDU
Code PDU Name
---- -------------------
8 NEWKEY
</artwork>
</figure>
<t>This document requests the IANA add a registry entry for "TOFU -
Trust On First Use" to the L3DL-Signature-Type registry as follows:</t>
<figure>
<artwork>
Number Name
------ -------------------
1 TOFU - Trust On First Use
2 PKI
</artwork>
</figure>
</section>
<section anchor="acks" title="Acknowledgments">
<t>The authors than Russ Housley for advice and review.</t>
</section>
</middle>
<back>
<references title="Normative References">
<?rfc include="reference.RFC.2119"?>
<?rfc include="reference.RFC.5280"?>
<?rfc include="reference.RFC.8174"?>
<?rfc include="reference.I-D.ietf-lsvr-l3dl"?>
</references>
<!--
<references title="Informative References">
</references>
-->
</back>
</rfc>