Re: [OPSEC] OPSEC control plane protection draft

[Rodney Dunn <[e-mail]>]

The current proposal on the table is the additional paragraph in Section 
3.3 "Design Trade-offs" section to add the following paragraph between 
*-*'s. The purpose of this paragraph is to highlight that the end goal 
would be to have a COPP policy that is as explicit as possible on what 
to permit and dropping all other traffic. Does the list agree that this 
paragraph accurately reflects that desire?


Section 3.3 "Design Trade-offs"


...
The goal of the method for protecting the router control plane is to
minimize the possibility for disruptions by reducing the vulnerable
surface.  The latter is inversely proportional to the granularity of
the filter design.  The finer the granularity of the filter design
(e.g., isolating a more targeted subset of traffic from the rest of
the policed traffic, or isolating valid source addresses into a
different class or classes) the smaller the probability of
disruption.

*In addition to the traffic matching explicit classes care should be 
taken on the policy decision that governs the handling of traffic that 
would fall through the policy. Typically that traffic is referred to as 
traffic that gets matched in a default class. It may also be traffic 
that matches a protocol specific class where previous classes that have 
more granular matching did not match all packets for that specific 
protocol. The ideal policy would have explicit classes to match only the 
traffic specifically required at the control plane and drop all other 
traffic that does not mach a predefined class. As most vendor 
implementations permit all traffic hitting the default class an explicit 
drop action would need to be configured in the policy such that the 
traffic hitting that default class would be dropped versus permitted and 
then delivered to the control plane. This approach requires rigorous 
traffic pattern identification such that a default drop policy does not 
break existing functionality of the device. The approach defined in this 
document allows the default traffic and rate limits it rather than just 
dropping it. This approach was highlighted as a way to give time for the 
implementer to evaluate the traffic in a production scenario prior to 
dropping all traffic not explicitly matched and permitted. However, it 
is highly recommended that after monitoring the traffic matching the 
default class that explicit classes be defined to catch the legitimate 
traffic. After all legitimate traffic is being explicitly matched the 
default class should be configured to drop any remaining traffic. *


Additionally, the baselining and monitoring of traffic flows to the
router's control plane are critical in determining both the rates and
granularity of the policies being applied.  This is important to
validate the existing policies and rules or update them as the
network evolves and its traffic dynamics change.  Some possible ways
to achieve this include individual policy counters that can be
exported or retrieved for example via SNMP, and logging of filtering
actions.

....




On 8/17/10 4:50 PM, Rodney Dunn wrote:
>
>>> What about if we add this in the Design
>>> Trade-Off section to address the concern so we will be more precise.
>>>
>>> *In addition to the traffic matching explicit classes care should be
>>> taken on the policy decision that governs the handling of
>>> traffic that
>>> would fall through the policy.
>> Do we need to say that by convention there is no default deny at the
>> end of a cpp policy?
>>
>>
>
> Another good catch...how about this:
>
> ....Typically that traffic is referred to as traffic falling in the
> default class. It may also be traffic that matches a generic protocol
> specific class where previous classes that have more granular matching
> did not match all packets for that specific protocol. The ideal policy
> would have explicit classes to match only the traffic specifically
> required at the control plane and drop all other traffic. *As most
> vendor implementations permit all traffic hitting the default class an
> explicit drop action would need to be configured in the policy such that
> the traffic hitting that default class would be dropped versus being
> delivered up to the control plane.* This approach requires rigorous
> traffic pattern identification such that a default drop policy does not
> break existing functionality of the device.....
>
>
>
>>> after monitoring the traffic matching the default class
>>> explicit classes
>>> should be defined such that the default class could be configured to
>>> drop traffic falling through the policy.*
>>
>> Can we add a COMMENTED OUT deny ip any any in the default policy with a
>> "do not enable until throughly tested in production ..." type statement.
>>
>
> I would prefer we cover it in the text section because that would make
> it harder to link the Legitimate Traffic (Section 3.1) definition to the
> actual configuration in the document.
>
> Rodney
>
>
>>>
>>>
>>> Rodney
>>>
>>>
>>>
>>> On 8/16/10 9:58 PM, Christopher Morrow wrote:
>>>> On Mon, Aug 16, 2010 at 9:34 PM, Rodney
>>> Dunn<[e-mail]> wrote:
>>>>> Donald,
>>>>>
>>>>> First thanks for the comment. It's a good one. We actually
>>> originally had it
>>>>> with a default drop for the all IP and default classes.
>>> However, after a
>>>>> good bit of discussion we (both Cisco and Juniper) felt
>>> that we should
>>>>> soften it up just a bit. We agreed to add the explicit
>>> match for the ALLIP
>>>>> class so it could be monitored and then tightened down further.
>>>>>
>>>>> We realized there were various opinions on how that should be done.
>>>>
>>>> can we get a 'first verify complete COPP coverage, then deny all
>>>> remaining traffic with $INSERT_PROPER_DENY_HERE' paragraph?
>>>>
>>>> It sounds like someone with a legal degree got to your final
>>>> recommendation :) that, operationally, leaves the network
>>> with a whole
>>>> to plug, and I can guarantee that someone with a scanning virus is
>>>> gonna fill it for you :(
>>>>
>>>> -chris
>>>>
>>>>> ie:
>>>>>
>>>>> http://www.cisco.com/web/about/security/intelligence/coppwp_gs.html
>>>>>
>>>>>
>>>>> Thanks,
>>>>> Rodney
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>>> On 8/16/10 5:16 PM, Smith, Donald wrote:
>>>>>>
>>>>>> For undesirables in JTK's paper here he specifically did
>>> a deny ip any any
>>>>>> at the end of the cpp policy for that.
>>>>>>
>>>>>> http://aharp.ittns.northwestern.edu/papers/copp.html
>>>>>>
>>>>>> The default term for juniper is log and discard.
>>>>>>
>>>>>> There isn't a deny ip any any in the draft.
>>>>>>
>>>>>>
>>>>>>
>>>>>> (coffee != sleep)& (!coffee == sleep)
>>>>>> [e-mail] gcia
>>>>>>
>>>>>>> -----Original Message-----
>>>>>>> From: [e-mail] [mailto:[e-mail]]
>>>>>>> On Behalf Of Rob Bird
>>>>>>> Sent: Friday, March 26, 2010 11:28 AM
>>>>>>> To: David Dugal
>>>>>>> Cc: [e-mail];
>>>>>>> [e-mail]
>>>>>>> Subject: Re: [OPSEC] OPSEC control plane protection draft
>>>>>>>
>>>>>>> This is most excellent. I was just advising a customer this
>>>>>>> morning on this very issue (again).
>>>>>>>
>>>>>>> I look forward to working on this.
>>>>>>> Rob
>>>>>>>
>>>>>>> -
>>>>>>> Rob Bird, Chief Technology Officer
>>>>>>> Red Lambda, Inc.
>>>>>>> "Network security at global scale"
>>>>>>> www.redlambda.com
>>>>>>>
>>>>>>> On Mar 26, 2010 1:03 PM, "David Dugal"
>>>>>>> <[e-mail]> wrote:
>>>>>>>
>>>>>>> -----BEGIN PGP SIGNED MESSAGE-----
>>>>>>> Hash: SHA1
>>>>>>>
>>>>>>> Hi Richard.
>>>>>>>
>>>>>>> Thank you very much for the scrutiny, analysis
>>> and feedback. As
>>>>>>> mentioned during my brief presentation, our hope
>>> is that this
>>>>>>> recommendation by example will provide awareness of a
>>>>>>> possible attack
>>>>>>> surface occasionally overlooked, especially by
>>> smaller or newer
>>>>>>> installations.
>>>>>>>
>>>>>>> I appreciate the feedback and will enhance the draft to
>>>>>>> make reference
>>>>>>> to cryptographic security, as well as attempt to make
>>>>>>> the document IP
>>>>>>> version agnostic.
>>>>>>>
>>>>>>> Thank you for your support, both in carefully reading
>>>>>>> the document, and
>>>>>>> for your willingness to have our draft taken under the
>>>>>>> OPSEC WG wing.
>>>>>>>
>>>>>>> - ---
>>>>>>> David G. Dugal Support:
>>>>>>> [phone]
>>>>>>> Security Incident Response Team Direct:
>>>>>>> [phone]
>>>>>>> Juniper Networks Mobile:
>>>>>>> [phone]
>>>>>>> Westford, MA, USA PGP Key:
>>> 0xAB6E02A5
>>>>>>>
>>>>>>>
>>>>>>> On Fri Mar 26 2010 09:06:40 GMT-0700 (Pacific Daylight
>>>>>>> Time), Richard
>>>>>>> Graveman<[e-mail]> proclaimed ...
>>>>>>>
>>>>>>>
>>>>>>> > David,
>>>>>>> >
>>>>>>> > I read the draft carefully after the meeting and
>>>>>>> realize that my
>>>>>>> > comments missed the...
>>>>>>>
>>>>>>> > .
>>>>>>> >
>>>>>>> -----BEGIN PGP SIGNATURE-----
>>>>>>> Version: GnuPG v1.4.10 (MingW32)
>>>>>>>
>>>>>>>
>>> iEYEARECAAYFAkus53cACgkQh59lzatuAqVE9wCgh53mgxNRPWUztlI27aOITHRr
>>>>>>> 2zMAoPb5y3phm260P1zSoDu0LSbUjNcN
>>>>>>> =kitD
>>>>>>> -----END PGP SIGNATURE-----
>>>>>>>
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> OPSEC mailing list
>>>>>>> [e-mail]
>>>>>>> https://www.ietf.o...
>>>>>>>
>>>>>>>
>>>>>>
>>>>>> This communication is the property of Qwest and may
>>> contain confidential
>>>>>> or
>>>>>> privileged information. Unauthorized use of this
>>> communication is strictly
>>>>>> prohibited and may be unlawful. If you have received
>>> this communication
>>>>>> in error, please immediately notify the sender by reply
>>> e-mail and destroy
>>>>>> all copies of the communication and any attachments.
>>>>>> _______________________________________________
>>>>>> OPSEC mailing list
>>>>>> [e-mail]
>>>>>> https://www.ietf.org/mailman/listinfo/opsec
>>>>>
>>>>> _______________________________________________
>>>>> OPSEC mailing list
>>>>> [e-mail]
>>>>> https://www.ietf.org/mailman/listinfo/opsec
>>>>>
>>>
>>
>> This communication is the property of Qwest and may contain
>> confidential or
>> privileged information. Unauthorized use of this communication is
>> strictly
>> prohibited and may be unlawful. If you have received this communication
>> in error, please immediately notify the sender by reply e-mail and
>> destroy
>> all copies of the communication and any attachments.
> _______________________________________________
> OPSEC mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/opsec