Locator/ID Separation Protocol (LISP) Control Plane
RFC 9301

This DISCUSS is somewhat arbitratrily on 6833bis, but many
of the same issues apply to 6830bis.

I concur with Ben's DISCUSS. I do not believe that these documents
have adequate security to advance to Proposed Standard.

I thought it might be helpful for me to lay out my starting
assumptions and threat model and what I think the appropriate standard
is here. That gives us an opportunity to discuss them prior to
getting into the specific security issues I raise below.


SYSTEM ARCHITECTURE
Per offline discussion, I understand that despite some of the
introductory material, LISP is not currently intended to be Internet
scale but rather to run in what seem to be fairly tightly controlled
environments. Thus, I am assuming the following facts about
the system:

- The Mapping Service itself is secure and trusted.  For the purposed
  of this discussion, I'm modelling all the entities in the services
  as one trusted element.

- The ETRs have a preconfigured relationship with the Mapping Service,
  which includes some sort of shared key and an ACL on the Mapping
  Service which tells it which EIDs anm ETR can advertise. How
  this gets established is out of scope of this discussion.

Note that neither of these assumptions would be reasonable in
an Internet scale system, but I'm assuming that the text about
that in these documents will be removed.

Because it's not in the document set before us, nor is it a normative
reference, I am disregarding LISP-SEC and only analyzing the system as
specified in these documents.


THREAT MODEL
I'm assuming the usual RFC 3552 threat model, I.e.,

- All non-Map Server elements in the system (specifically, endpoints
  and the xTRs are potentially malicious).
  
- Aside from the links between the Map Server elements, the network
  is controlled by the attacker.

Against this background, my expectation is that the attacker
should not be able to affect traffic in any fashion significantly
more effective than tampering with the data plane. For instance,
it's clearly the case that an on-path attacker between two xTRs
can drop all the packets or forward them to some third xTR, but
it should not be able to send a small number of packets which
would then affect the routing of a large number of packets.

I do not expect that the data plane should have better security
than native (non-IPsec) traffic. Given the nature of LISP and
the existence of a mapping system, it seems like it's kind of
a missed opportunity to deploy a credentials system that would
support IPsec-style data plane security, but given that this
isn't a generally safe assumption for IP traffic, and therefore
you need to provide some sort of transport or application security
anyway, I don't think it's the right standard to hold LISP to.


ATTACKS
LISP appears to be vulnerable to a number of routing attacks that
I claim above it should not be subject to. For example:

1. An on-path attacker can forge Map Replys to the ITR,
   thus redirecting traffic.

2. An ETR can perform an "overclaiming" attack in which it
   claims to be responsible for EIDs which it is not actually
   responsible for.

3. An off-path attacker can temporarily reroute traffic by exploiting
   the "gleaning" feature to cache poison an ITR. In addition, the
   "echo noncing" feature does not appear to have a sufficiently strong
   nonce to protect against forgery, and thus turning this into a
   long-term attack

4. An attacker may be able to perform a number of cache invalidation
   and contamination attacks by exploiting the Map-Version and
   Locator-Status bits. This may lead to DoS.

5. An attacker who was at time T responsible for an EID block
   can probably prolong its ability to respond for that block
   even after it is no longer responsible.
  
6. A number of the components appear to be subject to various replay
   attacks.

I note that many of these attacks are documented in the Security
Considerations for these documents. Also, I doubt this list is
exhaustive. As noted above, I have spent no time on the data plane
protocol.


DEFENSES
When looking at attacks, it's important to determine whether there are
plausible defenses. For most of these, I believe that the answer is
"yes", at varying levels of cost. As noted above, LISP-SEC appears to
be intended to address a number of these issues, so it's possible that
requiring LISP-SEC would go a fair ways towards addressing these
issues. A cursory look at LISP-SEC turns up some somewhat concerning
design choices, so I would have to examine it more closely to give a
real opinion.

I do not believe that LISP-SEC will address the attacks that do not
involve the Mapping Server. For instance, the gleaning
contamination/nonce attacks (3) would not appear to be fixed by
LISP-SEC. However, it's probably possible to fix them by lengthening
the nonce.

With that said, I tend to think that the overall authentication
architecture here would benefit from a rethink. At a high level, the
source of most of these problems is the "non-transferability" of the
mapping information from the Map Server. If the Map Server instead had
an asymmetric key pair which it used to sign mappings, then almost all
of these attacks would not work. Specifically:

- The map server could send signed Map Replys so forgery wouldn't work

- Map Replys from ETRs would be signed, so you couldn't overclaim

- Gleaning attacks would sort of work, but because the probe would
  elicit a Map Reply, you couldn't persist them

- Map Versions could be tied to signed objects, so you couldn't do
  cache invalidation by version. You'd probably need some other
  approach for Locator Status bits.

And so on.

Detailed review below, with some duplication....


Rich version of this review at:
https://mozphab-ietf.devsvcdev.mozaws.net/D4115


IMPORTANT
S 5.2.
>      s: This is the SMR-invoked bit.  This bit is set to 1 when an xTR is
>         sending a Map-Request in response to a received SMR-based Map-
>         Request.
>   
>      m: This is the LISP mobile-node m-bit.  This bit is set by xTRs that
>         operate as a mobile node as defined in [I-D.ietf-lisp-mn].

This would appear to create a normative reference to this document. To
avoid that, you need to specify how I behave if I receive it but I
don't implement lisp-mn.


S 5.2.
>      m: This is the LISP mobile-node m-bit.  This bit is set by xTRs that
>         operate as a mobile node as defined in [I-D.ietf-lisp-mn].
>   
>      I: This is the xTR-ID bit.  When this bit is set, what is appended to
>         the Map-Request is a 128-bit xTR router-ID.  See LISP PubSub usage
>         procedures in [I-D.ietf-lisp-pubsub] for details.

here too you seem to be creating a normative reference.


S 5.5.
>         is being mapped from a multicast destination EID.
>   
>   5.5.  EID-to-RLOC UDP Map-Reply Message
>   
>      A Map-Reply returns an EID-Prefix with a prefix length that is less
>      than or equal to the EID being requested.  The EID being requested is

How do I behave if I receive an EID-Prefix that is less than any of my
mappings. So, I might have mappings for 10.1.0.0/16 and 10.2.0.0/16
and someone asks me for 10.0.0.0/8? Also, when you talk about prefix
length, I assume you mean the length fo the mask?





S 5.6.
>      Authentication Data:  This is the message digest used from the output
>         of the MAC algorithm.  The entire Map-Register payload is
>         authenticated with this field preset to 0.  After the MAC is
>         computed, it is placed in this field.  Implementations of this
>         specification MUST include support for HMAC-SHA-1-96 [RFC2404],
>         and support for HMAC-SHA-256-128 [RFC4868] is RECOMMENDED.

What prevents replay attacks here? I'm guessing it's the Map-Version-
Number, but as I understand it, I can set this to 0.


S 6.1.
>      receives an SMR-based Map-Request and the source is not in the
>      Locator-Set for the stored Map-Cache entry, then the responding Map-
>      Request MUST be sent with an EID destination to the mapping database
>      system.  Since the mapping database system is a more secure way to
>      reach an authoritative ETR, it will deliver the Map-Request to the
>      authoritative source of the mapping data.

If I'm understanding this correctly, this allows an ETR to prevent an
ITR from learning that it is no longer the appropriate ETR for a
prefix. The way this attack works is that before the topology shift, I
send SMRs, thus causing Map-Requests, which, because my entry is
cached, refresh the cache on the ITR past the topology shift. I can
keep doing this indefinitely. Am I missing something


S 8.2.
>      authentication data, so prior to sending a Map-Register message, the
>      ETR and Map-Server SHOULD be configured with a shared secret or other
>      relevant authentication information.  A Map-Server's configuration
>      SHOULD also include a list of the EID-Prefixes for which each ETR is
>      authoritative.  Upon receipt of a Map-Register from an ETR, a Map-
>      Server accepts only EID-Prefixes that are configured for that ETR.

How does it know?

1) Versioning and backward compatibility

Section 5.2 says: "Support for requesting multiple EIDs in a single Map-Request
      message will be specified in a future version of the protocol."
However, there is no versioning mechanism for this protocol specified. How is versioning supposed to work?

Further given there is no new version, I wonder if the changes as outlined in section 10 are all backward-compatible? Especially for the introduction of the Message-Notify-Ack message, I guess there is no problem if a server sends it, however, as the sender of the Message-Notify message might not know if the other end supports sending of the Message-Notify-Ack it can't rely on it. This should be further discussed in the doc! Or is there another strategy to achieve backward compatibility?

2) Size and MTU

As outlined in the TSV-ART review (Thanks Colin!) this document does not discuss fragmentation or Path MTU discovery. RFC8085 recommends to either perform Path MTU discovery or limit the message to 576 bytes for IPv4 or 1280 bytes for IPv6 (minus any static header). As this seems to be an appropriate size for LISP messages, I would recommend this approach. Relying on IP fragmentation (as indicated in the reply to the TSV-ART review) is not recommended by RFC8085 as this would lead to IP packet without a UDP header, in the case of LISP, which can cause problem and loss when NATs are involved. In any case the chosen approach needs to be further discussed in the doc.

3) Rate-limiting and congestion control

Sec 5.3: "Map-Requests MUST be rate-limited.  It is RECOMMENDED that a Map-
   Request for the same EID-Prefix be sent no more than once per second."
As already noted by the TSV-ART review (Thanks Colin!), RFC8085 actually recommends to not send more the one packet per 3 seconds, and that is a restriction for all traffic not on a per-receiver base, or implement congestion control. This limit is meant to not only protect the receiver but also the network from overloading. Why do you use a smaller interval here? Also if (appropriate) rate limiting is used, this should either be a MUST or more explanation when it is okay to use a smaller rate limit should be provided.
However, after all, I don't think you those the right approach here for rate limiting. A Map-Request is always expected to be followed by some reply. For these kind of communication pattern, RFC8085 recommends to limit the number of outstanding requests to 1 (see sec 3.1.1 of RFC8085 recommending one packet per RTT), also for all traffic and not only per receiver. However, this would also require to implement some simple mechanism to detect a message as lost (see also further below in point 4).

Similarly I'm not sure about the intent of this requirement in section 5.5:
"Map-Replies SHOULD be sent for an EID-Prefix no more often than once
   per second to the same requesting router. "
My understanding is that Replies are only sent when a request is received. Why is this additional rate limit needed? Again if used it should be 3 seconds for all traffic to be inline with RFC8085. Also again, why is that not a MUST? Further recommendation are needed here.

Further section 6.1 say "Both the SMR sender and the Map-Request responder MUST rate-limit
   these messages.  Rate-limiting can be implemented as a global rate-
   limiter or one rate-limiter per SMR destination."
This seems to be the same rate limit as mention above, or not...? It would probably make sense to rate limit the SMR even further. Please clarify and provide more guidance, e.g. what should the value of a potential additional rate limit for SMR be?

Respectively the following sentence in section 6.1 is also unclear:
"The remote ITR MUST rate-limit the Map-Request until it gets a Map-Reply"
Why is the rate-limit as currently proposed depend on the fact if a Map-Reply is received? Is the ITR supposed to retransmit the Map-Request...? 

And finally the Map-Register, Map-Notify and Map-Notify-Ack messages does not seem to have any rate-limits. Recommendations inline with RFC8085 should be provided for the total traffic and not only for a few message types. Again, Map-Notify and Map-Notify-Ack messages should be send only once per RTT as there is a feedback mechanism.

For Map-Register sec 8.2 say: "Map-Register messages are sent periodically from an ETR to a Map-
   Server with a suggested interval between messages of one minute."
However, this a rather a low bound than an upper bound. A required (MUST) rate limit is still needed.

4) Loss detection and retransmission

As also mention by the TSV-ART review (Once more thanks to Colin!), this spec has an ACK mechanism for Map-Requests and now also for Map-Notify, however, it does not specify what to do if the ACK is not received (loss detection and retransmission scheduling). This makes the spec incomplete and needs to be further specified in the doc (and also has a relation to the point 3 above of course).