Estonia’s new ambassador-at-large for cyber security, Heli Tiirmaa-Klaar, recently explained to the Wall Street Journal that “compared to many other security fields, in cyber we have reached maybe 10 percent of total readiness to understand the threats, to respond to threats and also to prevent the threat or maybe deter the threat. We have lots of room for development.” She’s right; just look at the most basic of metrics: How do governments count cyber attacks? How do they classify them?
The problems — imprecision of language, and a lack of policy — can be seen in a trio of official quotes from a single month last year. On Jan. 7, French Defense Minister Jean-Yves Le Drian warned that 2016 had seen 24,000 cyberattacks against French defense targets, and that the attacks were doubling every year. On Jan. 8, the Financial Times reported off an interview with EU security commissioner Sir Julian King that “there were 110 separate attempts to hack the European Commission’s servers in 2016, a 20 percent rise on the year before.” And on Jan. 19, NATO Secretary General Jens Stoltenberg told Die Welt that “there was a monthly average of 500 threatening cyber attacks last year against NATO infrastructure that required intensive intervention from our experts. That’s an increase of 60 percent compared to 2015.”
Clearly, the figures were all over the place. But why? Did all three officials count cyberattacks differently? And if so, what standards and metrics did they apply?
So in October, I emailed their institutions to ask what incidents were included in their numbers (pings, port scans, phishing emails, malware infections, DDoS, etc.) and whether their standards and metrics were public. The French MoD never got back to me. The NATO press office said it could not process my question, because the alliance does “not comment on the nature of attacks or the methodology that [NATO] use[s] to qualify some incidents as attacks.” The European Commission’s IT Security Directorate politely explained that “we report internally on these figures but we do not publish this detailed information.”
Related: How to Survive a Russian Hack
But without published standards and discernable metrics, such warnings are of no real value to the public. We simply do not know whether 6,000 annual attacks against NATO’s infrastructure is a lot or whether any of the 24,000 attacks against the French MoD were serious. All we know is that something was counted by someone somehow to somewhat explain the threat environment.
To widen my inquiry, I also got in touch with the Dutch National Cybersecurity Center and Estonia’s Information System Authority, or RIA. The Dutch center coordinates the government response to cyber crises in the Netherlands and also serves as the Dutch central government’s Computer Emergency Response Team. Similarly, RIA coordinates the development and administration of Estonia’s information system and handles security incidents that have occurred in Estonian computer networks. Both adhere to certain baseline standards and metrics to count and categorize cyber incidents that are reported to them, and summarize their findings in annual reports.
But when I asked these organizations whether their respective governments had a single set of reporting standards and metrics, they said no. Officials with the Netherlands center emailed to say that “there is no single definition which applies to all Dutch ministries on what constitutes a cyber attack or critical incident” and that “Ministries are responsible for their own incident registration, including definitions and escalation procedures.”
RIA responded similarly — “there is no formal, universally applicable classification criteria for cyberattack/incident in Estonia that would apply across all government agencies or private sector parties” — but also noted that the government’s computer emergency response team has “an internally defined classification that allows for a reasonable level of consistency.” This is borne out, somewhat, on the quantitative side by RIA’s 2017 Cyber Security Assessment, which indicates that the CERT team handled 9,135 incidents in 2016, of which 1,687 related to government institutions.
In contrast, the 2017 Cyber Security Assessment Netherlands reported a mere 623 incidents, of which 254 occurred under the more general category of “public organizations.” However, the key difference between the annual reports is that Estonia’s notes whether incidents were low priority, medium, high, or critical, while the Netherlands’ does not.
The next question: does a “critical cyber incident” constitute a “cyberattack”? The Tallinn Manual, a collection of expert analyses on international cyber law, offers the widely accepted definition that a cyberattack is “a cyber operation, whether offensive or defensive, that is reasonably expected to cause injury or death to persons or damage or destruction to objects.” And according to the RIA report, “There were no critical cyber incidents that would have posed a threat to people’s life or health in 2016.” While this might seem like a clear-cut case for equating the terms, there is a caveat. The Estonian report says there were also “348 high-priority incidents that affected the functioning of a service or website considered important for the state,” including “interruptions or attacks against vital service providers’ information systems.” From a government perspective, those 348 incidents are attacks that have to be resolved in a matter of minutes to contain their destructive effects. Based on that report, then Estonia’s president could have told the public that the government had faced 9,135, 348, or zero cyberattacks in 2016.
So why is this a serious problem that needs fixing?
The first major concern is that when government officials, such as the NATO Secretary General or the French Minister of Defense, are presenting cyberattack figures, they are bound to significantly over- or under-report the occurrence of relevant cyber incidents. Clearly, the French MoD did not experience 24,000 critical cyber incidents in 2016, nor can we simply assume that any of the 500 critical cyberattacks against NATO were expected to cause injury or death to persons or damage or destruction to objects.
Imprecision therefore severely hinders the public’s ability to understand the threat environment. As a writer for Forbes asked in 2010: “Just how big is the cyber threat to the [US] Department of Defense?” The article cites the then-leaders of U.S. Cyber Command as drawing a line between probes and scans, while then-Deputy Defense Secretary William J. Lynn III called them all attacks. “What’s a probe? What’s a scan? How do they differ?… How serious is each type of incident? How many of each type of event are we seeing on a daily basis?”
Imprecision also hinders cyberdefense efforts within governments and between militaries. If NATO and EU member states lack common standards and metrics for reporting and categorizing cyber incidents, then statistics on national threat landscapes are destined to be both incomplete and non-comparable.
And third, imprecision blurs the rules of engagement for responding to a cyberattack. Just because Estonia categorizes an incident as critical – which might prompt Tallinn to invoke NATO’s Article 5 – hardly means the other 28 allies will evaluate the incident in the same way. We have already seen this playing out during the DDoS attacks against Estonia in 2007. Essentially, policy analysts divided into two sides: Those who believed that the attacks were the beginning of war, and those who argued that such attacks were already commonplace.
The bottom line is this: While NATO member states are embroiled in discussing cyber deterrence frameworks, offensive operations, and creating norms and rules for state behavior in cyberspace, they have still not reached consensus on how to actually count and categorize cyber incidents across the alliance. And two things are for certain even in cyberspace: The alliance cannot manage what it does not measure, and it has to understand what it is trying to solve.