As of right now, I hope I have captured people's views correctly, here is the summary that I seen from email (the current majority is 6 to 4 for "Against 2 objects":
Against 2 objects
Jason Keirstead - IBM
Bret Jordan - Symantec
Rich Shok - US Bank
Nicholas Hayden - Anomali
Pat Maroney - Wapack Labs
In the Middle
Allan Thompson - Looking Class
Dave Cridland - Survive
For 2 objects
John Wunder - MITRE
Sarah Kelley - CIS
Nathan Reller - JH APL
If you have not spoken up, please do so.
From: email@example.com <firstname.lastname@example.org> on behalf of Nicholas Hayden <email@example.com>
Sent: Wednesday, April 12, 2017 11:28:21 AM
To: Jason Keirstead
Cc: Wunder, John A.; firstname.lastname@example.org
Subject: [EXT] Re: [cti] Intel note and opinion
I also agree with Jason, having two separate objects will muddy the water for implementation.
Nicholas Hayden, CISSP, GICSP, CNDA, CEH, Sec+
808 Winslow St Redwood City, CA 94063
Phone: (650) 257-0867 | Twitter: @anomali
I have a very strong opinion either #2 or #3 must be done, and that #1 is not workable.
In order to fully understand why, you have to consider the entire life-cycle of the objects - from production in one piece of software, to consumption and storage/display in another. You also have to understand that,
at their core, both of these objects are objects are things that originate from humans, and carry human-entered facts about an object (either as human-entered text, or as "opinion", or other). Finally, you have to understand that the folks writing STIX software
do not normally expose the data model to the operators/analysts.... operators and analysts should not be seeing "note", or "opinion", or any other names of objects in most software that deals with STIX, it should be totally abstracted away. The analysts should
not know or care about this.
In any piece of software dealing with creating CTI by humans, one can imagine you will have to have some UI where one would enter these "things", and in any piece of software dealing with displaying CTI to humans, one
can imagine you will have to have some UI where one will display these "things". If we have two different objects, then you can immediately see how this presents a problem for the software creators
- Consider the producer software When someone wants to simply enter text - which object do I encode it as in STIX? Since both can convey the information, it is totally ambiguous which to use - out of the gate, we are
now at "two ways to do one thing", something we said we are trying to get away from in STIX 2. What if they enter text, it gets encoded as a "note", and then later on the same user goes in and and add an "opinion" flag? Should I revoke the "note" object and
add an "opinion" object? Leave it and issue an "opinion" and duplicate all the text? Again, totally ambiguous. This points to the fact that these two things are different ideas - voting and commenting - and should be kept fully separate (#2)
- Consider now the consumer software. Any piece of consumer software who is going to display notes and opinions to a user is going to want to have some kind of comment-trail.. some type of timeline. It will simply
not be possible to construct this comment trail without unionizing these two objects and treating them as one... as viewing a timeline of "note" without including "opinion", or vice-versa, will have the potential to leave a large number of human-created comments
dropped on the floor. I can't reasonably see any valid use case to have software where one shows a set of "opinion" without including "note", or vice-versa. This again points to the fact that you need a single source of truth for a comment timeline... either
option #2, or option #3 alternatively.
STSM, Product Architect, Security Intelligence, IBM Security Systems
Without data, all you are is just another person with an opinion - Unknown
From: "Wunder, John A." <email@example.com>
To: "firstname.lastname@example.org" <email@example.com>
Date: 04/10/2017 05:30 PM
Subject: [cti] Intel note and opinion
Sent by: <firstname.lastname@example.org>
After a lot of conversation on intel note and opinion, we’ve narrowed down a lot of the questions on these two objects but have one big one remaining. Specifically, with both intel note and opinion existing as separate
objects a few people (notably Jason and Bret) have noted that there may be overlap and in fact the objects should be merged into one. The thinking is that giving an opinion is essentially the same as giving extra analysis about something (or is at least handled
the same way most of the time) and having two separate objects will be confusing for people. So, here’s how I would outline the questions:
1. Should opinion and intel note remain separate objects?
a. Merging them would provide a single object to provide a simple opinion on a scale (agree/disagree), an opinion on a scale with a text explanation (agree and here’s why), and added analysis w/ no opinion scale
(here’s extra info about this object).
b. Separating them would distinguish providing an opinion (agree/disagree) from providing extra analysis
2. If we go with option b and we have two separate objects, should opinion have an optional description field?
a. Having a description on opinion keeps all information about the opinion in a single object.
b. Not having a description on opinion would mean that opinions are just the agree/disagree statements. People would use the intel note object to capture their explanation and therefore all text commentary would
be provided by intel note.
It seems like the key thing people are wrestling with is whether there’s a distinction between giving extra analysis or context to something and giving an opinion about something. I.e., when people are doing shared analysis
is it important to distinguish me providing an opinion on your object (agree/disagree/neutral) from me adding extra context (human-readable notes) to your data?
So, combining those questions, we have three options:
1. Opinion and intel note are separate objects, and opinion has a description. To have a text explanation of an opinion, you would use the description field on the opinion object.
2. Opinion and intel note are separate objects, and opinion does not have a description. To have a text explanation of an opinion, you would use an intel note and link it to the opinion.
3. Opinion and intel note are merged (likely calling it intel note, since not all of them would be opinions) and you would use that object to describe opinions, opinions w/ descriptions, and added analysis
People can reply with their reasoning and pros/cons, but I’m particularly interested in hearing people who have not chimed in yet. What is your preferred option? Any thoughts on the reasoning?
FYI, here are the latest working versions of intel note and opinion, in Google Docs. These are roughly option #1, based on the recent working call and a poll in Slack.
- Intel note: https://docs.google.com/document/d/15qD9KBQcVcY4FlG9n_VGhqacaeiLlNcQ7zVEjc8I3b4/edit#heading=h.74spnst8naxc
- Opinion: https://docs.google.com/document/d/15qD9KBQcVcY4FlG9n_VGhqacaeiLlNcQ7zVEjc8I3b4/edit#heading=h.haeazu2sh3sq
My own opinion (sorry I know this pun is getting old) is that giving an opinion is distinct from adding analyst notes or extra context and therefore I prefer #1. My second choice would be #2, because I think #3 results
in an ambiguous object that does too many things and can have completely orthogonal sets of fields, which to me is an indication that it really should be two objects.