Re: [I2nsf] RAW (i.e., too long) notes from the I2NSF WG meeting

[Linda Dunbar <[e-mail]>]

John,

Thank you very much for taking the notes,

Also thanks to Sue Hares and Frank for the extensive notes on the Etherpad. I will combine them together.

Linda

From: John Strassner [mailto:[e-mail]]
Sent: Friday, April 08, 2016 3:42 PM
To: [e-mail]
Cc: Linda Dunbar; Adrian Farrel; Xialiang (Frank); John Strassner
Subject: RAW (i.e., too long) notes from the I2NSF WG meeting

Hopefully these could be used as fodder for producing final minutes.

--
regards,
John


Introduction from the Chairs

Provided a summary of where the WG standards with respect to our
immediate milestones. Concerned about being late, and keeping up
(or increasing) our momentum. We must concentrate on getting the
framework and terminology statement ready for adoption by the WG;
this is overdue. Worried about four upcoming milestones (extensions
to protocols, existing secure communication mechanisms, and info
model in June, and data models in July).

We want to do milestones IN ORDER (with the exception of the info
model). Must contribute to previous milestones BEFORE we work
on new ones.

--- Working Group Drafts ---

Problem Statement and Use Cases
Sue Hares,   draft-ietf-i2nsf-problem-and-use-cases-00

Sue recommends shipping it, because no comments have been received
Wants people to say ON THE MAILING LIST they have read and
understood it.

Adrian asked about the Management Considerations section.
Dan Romascanu said that he thought that this is OK.
John Strassner offered to do a thorough review and send to the mailing list.

Gap Analysis
Sue Hares,   draft-ietf-i2nsf-gap-analysis-00

Thanks to Myo Zarny's comments. Helped get the terminology document
going. The framework draft is now aligned to the terminology document,
but wants WG to agree with this.

Dan Romascanu would like data modeling and data models to be added
to this draft. Sue said OK. However, she is concerned about the current
turmoil in opstate.

John Messenger asked why 802.1ax and 802.1ar were not covered; this
will be added in. Bob, John, and Dan will review for Sue.

--- Drafts for Immediate Milestones ---

Terminology
John Strassner,  draft-hares-i2nsf-terminology-00

Purpose is to enable precise and consistent terminology to be used,
and that I2NSF terminology does not clash with accepted applicable
terminology from relevant domains. The document was created by
scanning existing WG documents looking for key terms.

Status: security and policy terminology are reasonably complete for
I2NSF purposes. Included terms to facilitate implementation; Sue
met with SACM terminology editor (Henk) and they came up with a
preliminary list of terms that need alignment. John gave some
examples, but the conclusion is that both SACM and I2NSF could
benefit from such co-working and alignment.

While related, two separate documents (one from SACM and one
from I2NSF) will continue to exist. We will work together to ensure
that they do not conflict

Framework
Diego Lopez,  draft-merged-i2nsf-framework-05

Reviewed high-level architecture. Diego prefers "Developer Facing
Interface" to "Vendor Facing Interface".

Major changes:
   - Clarified packet- and flow-based processing
   - Changes Subject-Object-Action-Function to ECA
     - defined policy rule, event, condition, action, and how each of
       these are used in I2NSF
     - Removed references to PCIM (doubts about flaws associated
       with PCIM)
   - clarified that we want to standardize the form and function of
     profiles and signature files while concurrently enabling
     vendor-specific elaborations of each
   - added capability layer interface details
   - added vendor-facing interface details
   - Updated to use terms in Terminology Draft
   - added security requirements

Additional discussion:
     - There are three types of interfaces: configuration, signaling, and
       rule provisioning
     - Diego showed a nice mapping of how different functions (e.g.,
       how context and constraints) are mapped to ECA
     - Brief discussion about the "v" in vNSF. Starting to lean towards
       NOT explicitly separating vNSF from pNSF (physical NSF).
       However, it was noted that some problems that are unique to
       vNSFs can arise (e.g., restart, having policies follow an NSF,
        having multiple vNSFs collectively enforce policies)
     - We have covered controller-NSF; should we consider client-
       controller as well? Should we elaborate on differences between
       rules for the capability vs. service layers? What about capability
       negotiation?
    - Question about security controller. Diego clarified that this was
      NOT the typical "reactive" controller. It needs to be smarter, to
      translate capabilities and constraints to NSFs. Question from
      someone about "is this a controller" or not. John replied that
      it MAY be more instructive to think of this as a continuum - it is
      called "controller" because of the real-time nature of the decision,
      but it could be construed as management.

Preliminary poll for adoption for these two drafts
- Chairs
   draft-hares-i2nsf-terminology-00
   draft-merged-i2nsf-framework-05

No one disagrees; therefore, the poll will go to the list

--- Interfaces and Information Models ---

Capability Interface Information Model
Frank Xia,    draft-xia-i2nsf-capability-interface-im-04

This draft is about how to monitor what is going on in the I2NSF
architecture, and is more about designing the information model for
the capability interface for NSF. It will realize the security policy
provisioning rules that govern how packets are treated by the I2NSF
framework. This decouples the network security controller from
vendor-specific NSFs, and avoids unnecessary constraints on
using the functionality of NSFs.

There are three categories of security functions:
    - Network security control (inspecting and processing packets and
      flows using ECA)
    - Content security control (e.g., detect and remediate against
      malicious contents)
    - attack mitigation control (detect and remediate against different
      types of network attacks); needs a stadnardized interface

Showed a functional logic diagram of how ECA works. When an Event
fires, this triggers the evaluation of the Condition. If the Condition is
then true, then the Action MAY be executed. Note that each clause
may in general be a complex Boolean expression.

Discussed the various tables and grammar.

Next steps:
    solicit comments, and add more detail on capabilities, constraints,
    and how to extend the associated information model.

User-group-based Security Policy for Service Layer
John Strassner,   draft-you-i2nsf-user-group-based-policy-01

Motivation for this construct was described. Key point: this is an
extensible identifier to identify user groups. It will be generated by
policies under operator control, and takes the form of roles plus
additional criteria. This provides operators flexibility in making
policy decisions by decoupling what identifies a user from a
static representation of a user in the network.

The UAPC functional entities were described, along with a sample
implementation.


Information Model for Security Policy Exchange
Luyuan Fang,   draft-fang-i2nsf-inter-cloud-ddos-mitigation-api

We currently lack an efficient, automated, standard way to exchange
security information between providers. The problem is at the boundary
between providers. If this point is compromised, then both providers,
and especially inter-cloud operations, are also compromised.

Note that this is much harder to handle. It is very difficult to identify the
attacker as well as the status of a provider's partners; there is a distinct
lack of automated tools to exchange attack-related data, as well as to
support coordinated remediation.

We need a standardized set of inter-provider APIs for network security
policy exchange, so that providers can create and deploy their tools on
top of this standard framework. This in turn requires an information
model. Four types of information: mitigation capabilities, mitigation
request and response, monitoring and reporting, and knowledge sharing.
Want to try and define knowledge sharing objects.

John said that an info model is vital, but in order to define semantics,
you MAY need ontologies and/or some formal type of logic (such as the
ISO Common Logic (24707) work).


Information Model for Monitoring NSFs
DaCheng,   draft-zhang-i2nsf-info-model-monitoring-00

This draft specifies the information model for the monitoring part of
the capability interface. Want to concentrate on alarm and report
messages. The model will include common information that should be
included in all alarms and reports (e.g., NSF name, vendor name,
timestamps, type of NSF, NSF model,... ); preliminary definitions of
each were shown.

--- Other Work ---

Remote Attestation Procedures for Virtualized NSFs
Diego Lopez,   draft-pastor-i2nsf-vnsf-attestation

Showed a diagram explaining what the principles of attestation are.
Create a trusted channel, then users and the security controller
mutually authenticate to establish a trusted connection with the security
controller, which then makes the attestation available to the user.
Next version will include a deeper discussion on procedures,

Dan: why isn't this merged into the framework?
Diego: it could be; at the least, a reference should be there. However,
this draft contains additional information that goes beyond the scope of
the framework.


SDN-Based Security Functions
Jaehoon (Paul) Jeong,  draft-jeong-i2nsf-sdn-security-services-04

Described updates from -03 to -04.
   - Added new use case (VoIP/VoLTE) and two new requirements.

Described an OpenFlow-based architecture for a VoIP IPOS based
on ODL. The experiment in general follows the layers described in the
framework draft.  Next steps: provisioining of the service- and
capability layers, and provide more details about the prototypes.


I2NSF Data Flow Requirements & Secure Session Layer Services
Sue Hares,   draft-hares-i2nsf-mgtflow-reqs
Working on management data flow. Thought about different types of
attacks, such as TCP-SYN and ICMPO attacks. What could I2RS do to
help, and what could simplify the protocol?

DOTS and Mile have a set of management traffic requirements. These
place a set of important requirements on the flow of management data.
I2RS has some complementary requirements. A potential solution (in
draft-hares-i2nsf-ssls) describes placing a function above the many
transports being used so that a management entity can choose the
**best** transport to use.

Security alerts over the first MILE
Robert Moskowitz,     draft-moskowitz-firstmile

No standard mechanism exists to inform NSF about the policies
for dealing with security alerts in the monitoring system. Also, no
mechanism exists for NSF to report these alerts/events to the
monitoring system.

Note that this is different than DOTS (DDoS alerting/mitigation) and
MILE (inter-admin defense coordination). There are many other
attacks (e.g., Ping of death, TCP SYN, port scan, ...). In addition,
we need to be able to report events when the network itself is
under attack.

We need a pub-sub reporting system and a registration of
defense system monitoring entities.

Sue stated that the pub-sub work is being done in I2RS.