U.S. strategy has not kept pace with the evolving cyber threat. Recent proposals ignore key strategic features of the cyber domain, resulting in overly narrow policies. We must take a global approach to cyber-deterrence, and we must blend aggressive retaliation when the origins of attacks are clear with forbearance when they aren’t.
Here’s a scenario that should trouble America’s political leaders:
Top-secret plans for a next-generation fighter jet are stolen from a U.S. defense contractor’s computer. It appears the intrusion originated in China. Then again, it’s easy for other actors to make it look as if the culprit is China. Also, some signs point to North Korea. Ultimately, the United States blames China. It launches a retaliatory cyber strike that paralyzes Chinese military computer networks for a week. U.S. diplomats tell their counterparts that they’ve been warned against future incursions, but the move backfires.
It turns out the initial attack came from Iranian Revolutionary Guard operatives routing the attack through a server in China and using North Korean coding techniques. Chinese leaders are incensed—viewing claims about misattribution of an Iranian
attack as mere pretext for U.S. aggression. They respond with a cyberattack on the electrical grid in the Washington, DC area, causing rolling blackouts amid a summer heat wave. Several people die. As the dust-up becomes public, U.S.
officials feel pressure to retaliate further. The president orders a missile strike against a Chinese military computer facility. Two-dozen Chinese coders are killed. What started as a limited cyber-intrusion now threatens to turn into a full-fledged military conflict.
Each element of this hypothetical has occurred in the recent past in one form or another. And it shows how complicated the strategic landscape has become in the cyber age. Unfortunately, our strategic thinking has not kept pace. The United States needs a new strategy for deterrence in cyberspace.
Attribution problems are the chief strategic complication. As former U.S. Deputy Secretary of Defense William Lynn put it: “Whereas a missile comes with a return address, a computer virus generally does not.” In congressional testimony, former NSA Director Michael Hayden warned that “[c]asually applying well-known concepts from physical space like deterrence, where attribution is assumed, to cyberspace, where attribution is frequently the
problem, is a recipe for failure.”
During the Cold War, game theory played a key role in helping policymakers formulate strategies appropriate to the challenges of the nuclear age. Following this tradition, we have taken up General Hayden’s challenge by developing a new game theoretic analysis of deterrence appropriate for the cyber age, one that takes the problem of attribution seriously.
Our analysis yields four key insights:
- Deterrence in cyberspace is fundamentally global and interconnected, not
- Optimal cyber-deterrence blends aggressive retaliation when attacks are clearly
attributable with forbearance when they aren’t, rather than across-the-board
- Retaliatory efforts should be focused on our most deterrable, rather than most
- Technological improvements in attribution will not always improve deterrence.
A Global Landscape
Our call for a new cyber strategy comes in the context of other schemes we view as dangerously simplistic. The latest Department of Defense Cyber Strategy focuses on our most belligerent and capable cyber adversaries: China and Russia. The 2018 National Defense Strategy acknowledges North Korea and Iran as rogue nations to contend with and notes the emergence of threats from non-state actors. Nonetheless the 2018 DOD Cyber Strategy emphasizes America’s two biggest rivals. “Our focus will be on the States that can pose strategic threats to U.S. prosperity and security, particularly China and Russia,” it reads. But we cannot be effective in the cyber domain, which contains a much larger number of cyber-capable adversaries, if we narrow our lens in this way.
Consider a country weighing how to respond to a cyberattack. That country will only retaliate if it is sufficiently confident in its assessment of which adversary is responsible. Of course, it is the anticipation of such retaliation that creates deterrence. This means that an adversary will be more aggressive in cyberspace when it believes the defending country is less likely to reach the
confidence threshold necessary for retaliation.
One important input to blame assessment concerns features of the attack itself—e.g., the location of servers, the language and style of malicious code, or the identity of likely beneficiaries. Another input concerns the more general strategic environment. Adversaries believed to be particularly active or capable in the cyber domain will be more suspect following any hard-to-attribute attack. It is this latter fact that makes cyber-deterrence fundamentally global, rather than bilateral.
Suppose some adversary, say China, is believed to have become more aggressive in the
cyber domain. Then China is now more suspect whenever a hard-to-attribute cyberattack occurs. China is thus more likely to face retaliation. But if China becomes more suspect, other adversaries, say North Korea, Russia or Iran, must become less suspect. And so these other adversaries are less likely to face retaliation. This reduced risk of retaliation tempts these adversaries to become more aggressive. And so, in cyberspace, if we become worse at deterring any one adversary, we become worse at deterring them all.
This interconnectedness is reflected in several recent cyberattacks. According to American authorities, a Russian cyberattack disrupted the opening ceremony of the PyeongChang Winter Olympics. The GRU routed the intrusion through North Korean IP addresses to deflect blame. The North Koreans were an attractive target for this “false flag” operation precisely because they were already highly suspect due the Sony Pictures hack and other cyber operations. Or consider the “GhostNet” plot, a worldwide infiltration of government and commercial networks, originating in China. A report by the Information Warfare Monitor identifies the Chinese government and military as leading suspects. But it notes that another plausible explanation is “a state other than China, but [operating] physically within China…perhaps in an effort to deliberately mislead observers.”
Here we see why attribution problems mean we must think and act globally. If we narrow our focus to China and Russia, then we encourage belligerence by other actors. And this increased aggressiveness will create new opportunities in cyberspace for the Chinese and Russians.
Toward a More Effective Cyber Deterrence Strategy
In any deterrence setting, the optimal response after an attack may not be the same as
the threat that optimizes deterrence prior to an attack. This familiar problem of credible commitment necessitates that governments articulate and commit to a deterrence doctrine. Serious discussions are underway about how countries might pre-commit in cyberspace.
But, for such pre-commitment to be of use, we must know what the optimal doctrine
is. Recent arguments call for a more aggressive retaliatory regime—for instance by declaring governments responsible for cyberattacks originating in their territory, regardless of the
perpetrator. Such calls are consistent with the general theory of deterrence: heightened retaliatory aggressiveness deters more attacks.
But our analysis shows that matters are less clear-cut in the cyber domain. There is a
vanishingly small chance that we will engage in, say, nuclear retaliation against the wrong adversary. As William Lynn reminds us, missiles come with a return address. But cyberattacks do not. As such, cyber-deterrence doctrine must balance a fundamental trade-off. Committing to a more aggressive retaliatory policy deters more attacks. But it also entails greater risk of
mistaken retaliation. Thus, in cyber warfare, where attribution problems loom large, full deterrence is infeasible. And increased aggressiveness across-the-board is unlikely to be optimal.
The optimal deterrence doctrine for cyber warfare is more nuanced. We should commit
ourselves—through policy declarations, treaties, and standing military orders—to retaliating more aggressively than we otherwise would following clearly attributable attacks. But we should also commit ourselves to retaliating less aggressively following attacks whose attribution is
particularly ambiguous. Such forbearance will reduce the risk of erroneous retaliation and dangerous escalatory spirals, with only limited costs for deterrence and security. And let’s return to the global nature of cyber conflict. Despite the 2018 Department of Defense Cyber Strategy’s focus on Russia and China, the optimal cyber doctrine doesn’t call for increased aggressiveness against our most aggressive adversaries. Rather, it calls for increased aggressiveness against our most deterrable adversaries. An adversary is deterrable if its attacks are particularly easy to attribute (e.g., it is technologically limited, other countries aren’t trying to mimic it) and if it is particularly responsive to retaliation (e.g., because of its own cyber vulnerability or because of
domestic political considerations). When we improve deterrence against these adversaries, we improve deterrence against all our adversaries, who will have fewer other cyber aggressors to hide behind.