You are currently browsing the category archive for the ‘Risk Management Theory’ category.
So, I said I would do a “Risk Analysis” of Trump’s executive order using an actual risk framework to prove the order was dumb and where/why/how it was dumb. Here it is! I used FAIR! (Note: I used a single data point here, because that’s all this analysis needed – I just go through control relationships and impact types mostly – but despite the snarky tone, the analysis is legit and if someone runs with this and fills in all the factors with actual numbers, that would likely be a fun exercise – I’d love to see what the actual estimated ALE is.)
In FAIR (Factor Analysis of Information Risk), the expression of risk is in the form of the “Probable frequency and magnitude of loss” due to a weakness. It is a useful model for many things, including identifying whether a perceived weakness is actually a weakness in terms of your own appetites and, interestingly, what effect controls have on your risk and how they affect each other.
I’m not going to introduce the whole model here and lot of this is paraphrased, so a quick overview is probably helpful. If you’re not familiar with FAIR, start with Risk Lens’ “FAIR on a Page” (it’s also an open standard, which you can read about at the Open Group’s website).
Moving on: In the model (an ontology, see above), you have a “Loss Frequency” measure and a “Loss Magnitude” measure. Each of those measures is determined by a number of factors. These factors include things like “Threat Event Frequency” and “Vulnerability”. Each of these factors are themselves composed of other factors (e.g. “Contact Frequency” and “Probability of Action” combine to estimate “Threat Event Frequency”).
On the right side of the ontology, “Loss Magnitude” is split into “Primary Loss” (costs that happen every time a threat event is successful) and “Secondary Loss” (costs which could occur for a subset of Loss Events and which come from impacts to other stakeholders not accounted for in Primary Loss). Think: A bank loses customer data and has to do incident response (primary loss) and sometimes must also provide free credit monitoring to those customers (secondary loss).
While this model is clearly geared for “information security”, it’s actually pretty useable for reasoning through other “risk” scenarios.
For instance, why is the Trump executive order on banning immigrants….dumb? Can we express this clearly and show the interplay between factors that gets us to “dumb” in a rational way? Yes! Ill walk you through it! Refer back to Risk Lens’ “FAIR on a Page” if you need to:
The first thing we need to do is determine threat event, threat community, loss events, and assets. Based on the executive order, we can say this:
- Asset: US interests inside its borders
- Threat Community: Anyone from one of the 7 countries
- Threat Event: Attempted Terrorist Attack inside US borders
- Loss Event: Successful Terrorist Attack inside US borders
- Secondary Loss: The consequences and responses to a successful terrorist attack inside US borders (this is important)
So, with that said, let’s run through the ontology:
1. Loss Event Frequency: None. We have never experienced any successful attacks inside US borders from this threat community here and there’s no evidence to think it will soon. “0 Loss Events” x “Any magnitude of loss” = “No Loss, No Risk”. Normally, if we just wanted to know the immediate and probably future risk, we could stop here: We know how often loss occurs and we know that an executive order cant reduce “0” any further. But we also want to examine what could change and if the executive order will help manage THAT totally speculative future problem, so we will look below the surface to…
2. Threat Events (Loss Event SubFactor 1): Because Loss Event Frequency is so low, it doesn’t really matter how often folks TRY to cause loss if theyre rarely successful (NO control can keep the level at “Never successful”). But perhaps of Threat Events go up enough, we will experience a noticeable increase in Loss Events? This makes sense…but to understand whether this will happen and whether the executive order will help, we have to understand the two Threat Event Factors:
A. Contact Frequency (Threat Event Subfactor 1): How often is the threat community in contact with US assets in the border? Pretty often! We have lots of folks from all over the world in this country. Trump’s Executive order keys in on this factor in particular. If we reduce contact frequency, we will reduce the number of threat events, which will reduce the number of loss events! (From..again… “0” to…less than 0?)
B. Probability of Action (Threat Event SubFactor 2): We do have a Threat Community that comes in FREQUENT contact, but they rarely, if ever take action. In fact, given how high the contact frequency is vs successful threat events (ie, loss events), we can say the Probability of Action is VERY low (if it weren’t low, we’d be seeing more successful attacks) – so low that, unless something changes to increase probability of action, we are at very low risk from this threat community (remember this for later: What could change the probability of action?)
3. Vulnerability (Loss Event SubFactor 2): Ok, so…that’s Threat Event Frequency…. What about our Vulnerability in the face of Threat Events? Maybe theyre trying a lot and just not successful? If that’s the case, then Trump’s executive order STILL doesn’t have an impact, because it attempts to control for “Contact Frequency” and does not affect either of “Vulnerability”’s two factors: Required “Threat Capability” and “Resistance Strength”. The Executive order doesn’t increase the tools/skills needed by bad guys and it doesn’t make our Assets (people, infrastructure) particularly more resistant to attacks in the fact of someone who chooses to take action. We MAY be able to improve this, but Trump’s order doesn’t speak to it and given the low Loss Event Frequency it probably isn’t necessary for this Threat Community
There we have it: Loss Event Frequency – very low and Trump’s order doesn’t really speak to it anyway. What about Loss MAGNITUDE? Here is gets interesting:
1. Primary Loss Magnitude: This is what the immediate aftermath of an attack would be for the US. If anyone gets in and does damage, Trump’s order does nothing to minimize that magnitude or impact. Once it happens, it happens. Having a smaller Threat Community doesn’t make the pain or cost less later. So this is a null factor.
2. Secondary Loss Frequency and Magnitude: Woah! :) Here we have a problem. Because it turns out Trump’s Order IS A SECONDARY LOSS in FAIR terms. It is a RESPONSE to prior terrorist attacks. Because we were attacked, someone used FEAR (not rationality) to justify keeping folks out of the country. Families. Children. Scientists. Injured. It was insensitive to entire nations of people who were Not already likely to take action (based on the Loss Event Frequency analysis). But, what happens if you find your son or daughter couldn’t get medical treatment because of this? What happens if you find your family split up over this? What happens if your radical organization can turn to you and say “Look, we were right, these people are assholes and don’t want you”? What happens is THE PROBABILITY OF ACTION INCREASES FROM NOT ONLY THE ORIGINAL THREAT COMMUNITY, BUT OTHERS. In other words, people otherwise unlikely to take action before are angrier as a result of trumps actions and are now more likely to act against the US, thereby increasing future primary and secondary losses.
Summary (in case it wasn’t clear):
- Trump’s order does not minimize our vulnerability
- Trump’s order reduces the “Contact Frequency” of a threat community who has never been a demonstrated source of losses to the US internally
- The only significant impact on US risk that Trump’s order actually has is that it likely INCREASES the probable future frequency and magnitude of loss to the united states (it’s risk) by increasing the probability of action factor without affecting the others either way
Conclusion: After a FAIR Risk Analysis of Trump’s order, it turns out it was indeed DUMB.
Ill have a longer discussion of SIRACon later, maybe, but for now, you can find my talk slides here:
Some of it is old material, but some of it is new. I really like how it’s fitted together and ordered here.
A couple of people have asked me to clarify what I mean by Sieges (and parasites) in terms of the first Siege post and the subsequent strategy/problem space framework post. Here’s a quick email I wrote that might help:
Sieges and Parasites:
From a collective non-aggressor entity perspective, cybersecurity “conflict” is functionally a siege of the collective environment: Non-combatants trying to maintain a minimum level of survivability while they’re surrounded, being drained of resources, and lack sufficient environmental influence/position to make effective risk decisions.
Compare/Contrast Siege and Parasitic Environment as conflict types to: crime, espionage, battlefield warfare, natural events. These latter tend to be incident/event driven, where the risk and responses to a siege are more environmental over time, with incidents to individuals happening but being largely irrelevant except as they contribute to the overall lack of stability/freedom to operate.
This though process got kicked off for me while reading about the siege of Sarajevo in particular. Imagine – you (a private org standing in as a citizen for this narrative) are in a city surrounded by artillery and snipers and you have to decide how best to keep getting water, which involves cross several streets through town. Some streets are vaguely safer than others, usually, but not necessarily. You occasionally can see or have insight into the people on the hills, but not usually. There are dedicated defenders around, but theyre not well positioned and lack the capacity to defend everyone all the time. Your resources are limited and your freedom to operate is constrained further over time as resources diminish. You can be hit at any time once you move from a standstill from your base/home (and even then, without change, you are at some risk). You sort of make up criteria for decisions that help you feel safer (has anyone crossed that street recently? Were they shot at?) but aren’t really indicative of actual risk.
In this case, trying to decide how and when to get water as a risk based decision is almost a nonsensical proposition: You don’t control your environment, you have a lot of exposure, and you lack relevant information that would change your situation significantly (this isn’t the same as lacking data, just helpful data).
This scenario is substantially different from how we look at cybersecurity and infosec today: Individual defenders, with sufficient skill and competency, access to resources indefinitely and as needed, on a relatively level playing field, trying to prevent, manage, or mitigate individual events on their own.
Ultimately, right now, we’re asking a bunch of non-combatants (you know, most businesses) to have the capacity to effectively and sustainable participate in what is becoming a low level global conflict (inclusive of state to state, criminal, hacktivist, etc activity) while under siege.
This is a broken model and will never, ever get us where we want to be (for more reasons than I’ll lay out here). We have to break the siege (thoughts on that being out of scope for the moment), which involves a level of strategic cooperation and unity that present culture, politics, business realities, and law do not allow.
(The Parasitic environment analogy is more specific to single-organizations, as it allows for specific targeting: https://sintixerr.files.wordpress.com/2015/01/hackervaluechain2.jpg )
Aside: Interestingly, though, from an aggressor standpoint, it might or *might not* look like either a siege or a parasitic environment – ie, aggressors acting individually and *without* coordination are contributing to creating a separate conflict type for defenders (Siege).
I’ve been known, now and again, to mouth off sarcastically that we don’t have any idea what “Cybersecurity” is, strategically – that we have no real concept of what “it” is. So, as a preface to my upcoming talks, I’ve sketched out a very, very draft and incomplete framework off the top of my head that is, I think, STILL more complete than anything else out there. It’s done in terms of “Environments” that must be managed or that pose describable, discernible, solvable problem spaces that pertain to cybersecurity risk. Note how different this looks than the NIST Framework, NERC CIP, SANS guidance, what you hear panels talk about, etc. Just remember, I have a lot to add here, which I’ll do after my upcoming talks have been given.
A STRATEGIC CYBERSECURITY PROBLEM SPACE FRAMEWORK
- Sieges & Unity (Defense problem of community siege-breaking, not incidents)
- Game Theory & International Relations
- Norms, Stabilization, and Confidence Building Measures
- Parasite Management
- Single Organization Value Control
- Competition for use of shared, not owned infrastructure
- Information vs Kinetic Warfare
- Long term abuse of misplaced cultural, political, and legal redlines
- Complexity (exposure rising directly and infinitely with complexity)
- Competency (technical competency required by all, who cannot maintain)
- Security Express-ability (lower layers are approximating upper layer expressions)
- Geography & Power Delegation (Targets are Geography, cannot insert gov between industry and adversary)
- Geography & Proximity (Everyone is a Neighbor)
Single Organizational Environment
- Developing Sustainable Practices without requiring core Competency
- Decision Making Capacity Building
- Full System (Human) Threat Modeling
- Self Awareness
- Vulnerability/Exposure Identification & Management
- Exploitation Opportunity Identification & Management
- Stakeholder psychology requires targeted action to achieve desired behavior change
- Exceptional Distance between decisions, actions and risk limits involvement
- Ability to Process sufficient incoming knowledge tangential to core
- Common Problem Space Consensus Development/Socialization
- Development and Engagement of Appropriate Regimes
- Stabilizing vs Developing managed Environments
- Business Value Production is inherently and completely tied to exposure creation/mgt, how does gov manage?
- Entrenched Industry is sucking needed resources away uselessly, needs derailment (fail, iterate, improve)
- Abstract, tenuous connection between market and risk
- We Need Generals: Now Guys with Guns Espousing Tactical Requirements in Place of Strategies to Win (Win = Desired level of risk for desired investment over time)
- Formal Roles limiting Routing of Knowledge/Capability into available levers (ie, if you’re not selling something, you’re not participating)
Getting to be a busy fall/winter schedule. If you’re interested in catching up with me, learning, or just discussing security, check me out at one of these venues:
- Sept 14 | Washington DC | EnergySec Summit
- Giving workshop on Frameworks and the Discipline of Cybersecurity
- Sept 28-29 | Krakow, Poland | CYBERSEC EU
- Panelist in the State Stream
- Oct 8-9 | Detroit, MI | SIRAcon
- Speaking on risk: Yours, Anecdotally”
- Oct 13-14 | Dallas, TX | “Reframing Cybersecurity”
- Teaching my 2-day class
- Nov 7 | Jackson, MS | B-Sides Jackson
- Nov 10-11 | Nashville, TN | “Reframing Cybersecurity”
- Teaching my 2-day class
- Jan | Florida
- To Be Announced
Cant make these? Interested in having my somewhat unusual viewpoints represented at your security, industry, coffee, hiking, or other event? Let me know so I can get it on the calendar! :)
So the results of the Mozilla Delphi project are out. I was one of the panelists – alongside some pretty well known names like Jane Hall Lute, Bruce Schneier, and some other big etc.’s. You can find it here:
And some background here:
“Mozilla’s Cybersecurity Delphi 1.0 is a step to address this gap, by identifying and prioritizing concrete threats and solutions. Through the iterative structure of the Delphi method, we will build expert consensus about the priorities for improving the security of the Internet—infrastructure to protect public safety, sustain economic growth, and foster innovation. The Delphi method offers unique benefits in this context because it aggregates the input of a diverse, broad set of voices, using a discrete and defined process with a clear, fixed end point and a mechanism for non-attribution to encourage open and through engagement. “
Im still processing the results, many of which I adamantly disagree with, but what I think the report mainly shows is that “cybersecurity” isn’t a thing that exists outside of specific sets of contexts and perspectives and goals. It just goes…poof…and disappears as a concept if it’s not bracketed by material constraints. The all over the board nature of the responses seems to demonstrate that (even though Mozilla did a good job creating a narrative around them).
That said, I think there are some interesting points in the document and that it’s worth a read – at the very least you’ll get to see some of the filter biases of some very smart people (obviously including my own). And those are worth knowing, because very often our human fears and backgrounds and perceptions are not reflective of actual risks and needs.