Can the best time to launch a cyberattacks be predicted? Researchers at the University of Michigan say they have developed a mathematical model that can do just that.

The model analyzes when a potential attacker is most likely to hit. Robert Axelrod, professor of political science and public policy at Michigan’s Ford School, told news media that the model he developed with postdoctoral research fellow Rumen Illiev creates some new concepts for dealing with computer attacks by pinpointing the best time to use specific cyber methods. Axelrod compared their work, which is focused on the strategic dimensions of cyberattacks, to the strategic considerations for the use of nuclear weapons.

The concepts are built around stealth, analyzing the ability of a resource to exploit a vulnerability in a computer system Relevant Products/Services but remain undiscovered if it is used, and persistence, the ability of a vulnerability to remain undiscovered if it is not used. Illiev noted that a successful attack Relevant Products/Services would be built around both stealth and persistence.

Like a Double Agent

“The question of timing is analogous to the question of when to use a double agent to mislead the enemy, where it may be worth waiting for an important event but waiting too long may mean the double agent has been discovered by the target and becomes useless," the researchers noted in their paper, published last month by the Proceedings of the National Academy of Sciences.

Other variables include the weapon’s value, a function in part of its stealth and persistence; the current and future stakes; the threshold of stakes that would cause an attacker to use the weapon; and what the researchers call the “discount rate” -- the fact that using the weapon today could be more valuable to the attacker than doing so later.

The model shows that it is better to use stealthier weapons sooner rather than later, and that the use of more persistent weapons can be postponed. Those observations may seem to be fairly obvious, but the model fits several instances of actual attacks.

‘Rational Timing of Surprise’

These instances included the famed Stuxnet worm, which the model described as having low persistence -- because it employed four zero-day exploits -- but a high level of stealth. The weapon was much more valuable being used sooner rather than later, since the aim was to delay the ability of Iran to develop nuclear-grade uranium.

The researchers said that working from their model, Stuxnet was “expected to have poor persistence and comparatively good stealth,” which is how it was used.

Similarly, a cyberattack by Iran on Saudi Aramco (Saudi Arabian Oil Co.) was not stealthy and the stakes required quick action. By contrast, the researchers said that Chinese cyberattacks are not usually conducted at the most optimal times, but they said it’s hard to understand why a nation-state might choose some less-than-effective occasion.

This, of course, points to the model’s weakness, in that it expects attackers to be rational actors, attacking at the most opportune moment for the greatest leverage, given the stakes and the choice of weapons.

In fact, the researchers describe their model as a way to study “the rational timing of surprise.”