For cybersecurity, it is pivotal for the next decade to be able to operate with a decreasing time window to act.
The accelerated execution of cyber attacks and an increased ability to at machine-speed identify vulnerabilities for exploitation compress the time window cybersecurity management has to address the unfolding events. In reality, we assume there will be time to lead, assess, analyze, but that window might be closing. It is time to raise the issue of accelerated cyber engagements.
Limited time to lead
If there is limited time to lead, how do you ensure that you can execute a defensive strategy? How do we launch counter-measures at speed beyond human ability and comprehension? If you don’t have time to lead, the alternative would be to preauthorize.
In the early days of the “Cold War” war planners and strategists who were used to having days to react to events faced ICBM that forced decisions within minutes. The solution? Preauthorization. The analogy between how the nuclear threat was addressed and cybersecurity works to a degree – but we have to recognize that the number of possible scenarios in cybersecurity could be on the hundreds and we need to prioritize.
The cybersecurity preauthorization process would require an understanding of likely scenarios and the unfolding events to follow these scenarios. The weaknesses in preauthorization are several. First, the limitations of the scenarios that we create because these scenarios are built on how we perceive our system environment. This is exemplified by the old saying: “What gets us into trouble is not what we don’t know. It’s what we know for sure that just ain’t so.”
The creation of scenarios as a foundation for preauthorization will be laden with biases, assumption that some areas are secure that isn’t, and the inability to see attack vectors that an attacker sees. So, the major challenge becomes when to consider preauthorization is to create scenarios that are representative of potential outfalls.
One way is to look at the different attack strategies used earlier. This limits the scenarios to what has already happened to others but could be a base where additional scenarios are added to. The MITRE Att&ck Navigator provides an excellent tool to simulate and create attack scenarios that can be a foundation for preauthorization. As we progress, and artificial intelligence becomes an integrated part of offloading decision making, but we are not there yet. In the near future, artificial intelligence can cover parts of the managerial spectrum, increasing the human ability to act in very brief time windows.
The second weakness is the preauthorization’s vulnerability against probes and reverse-engineering. Cybersecurity is active 24/7/365 with numerous engagements on an ongoing basis. Over time, and using machine learning, automated attack mechanisms could learn how to avoid triggering preauthorized responses by probing and reverse-engineer solutions that will trespass the preauthorized controls.
So there is no easy road forward, but instead, a tricky path that requires clear objectives, alignment with risk management and it’s risk appetite, and an acceptance that the final approach to address the increased velocity in the attacks might not be perfect. The alternative – to not address the accelerated execution of attacks is not a viable option. That would hand over the initiative to the attacker and expose the organization for uncontrolled risks.
Repeatedly through the last year, I have read references to the OODA-loop and the utility of the OODA-concept för cybersecurity. The OODA-loop resurface in cybersecurity and information security managerial approaches as a structured way to address unfolding events. The concept of the OODA (Observe, Orient, Decide, Act) loop developed by John Boyd in the 1960s follow the steps of observe, orient, decide, and act. You observe the events unfolding, you orient your assets at hand to address the events, you make up your mind of what is a feasible approach, and you act.
The OODA-loop has become been a central concept in cybersecurity the last decade as it is seen as a vehicle to address what attackers do, when, where, and what should you do and where is it most effective. The term has been “you need to get inside the attacker’s OODA-loop.” The OODA-loop is used as a way to understand the adversary and tailor your own defensive actions.
Retired Army Colonel Tom Cook, former research director for the Army Cyber Institute at West Point, and I wrote 2017 an IEEE article titled “The Unfitness of Traditional Military Thinking in Cyber” questioning the validity of using the OODA-loop in cyber when events are going to unfold faster and faster. Today, in 2019, the validity of the OODA-loop in cybersecurity is on the brink to evaporate due to increased speed in the attacks. The time needed to observe and assess, direct resources, make decisions, and take action will be too long to be able to muster a successful cyber defense.
Attacks occurring at computational speed worsens the inability to assess and act, and the increasingly shortened time frames likely to be found in future cyber conflicts will disallow any significant, timely human deliberation.
I have no intention of being a narrative impossibilist, who present challenges with no solutions, so the current way forward is preauthorizations. In the near future, the human ability to play an active role in rapid engagement will be supported by artificial intelligence decision-making that executes the tactical movements. The human mind is still in charge of the operational decisions for several reasons – control, larger picture, strategic implementation, and intent. For cybersecurity, it is pivotal for the next decade to be able to operate with a decreasing time window to act.
Jan Kallberg, PhD