"Cyberwar Is Already Upon Us."

Think Again: Cyberwar

Don't fear the digital bogeyman. Virtual conflict is still more hype than reality.

BY THOMAS RID | MARCH/APRIL 2012, Foreign affairs

No way. "Cyberwar is coming!" John Arquilla and David Ronfeldt predicted in a celebrated Rand paper back in 1993. Since then, it seems to have arrived -- at least by the account of the U.S. military establishment, which is busy competing over who should get what share of the fight. Cyberspace is "a domain in which the Air Force flies and fights," Air Force Secretary Michael Wynne claimed in 2006. By 2012, William J. Lynn III, the deputy defense secretary at the time, was writing that cyberwar is "just as critical to military operations as land, sea, air, and space." In January, the Defense Department vowed to equip the U.S. armed forces for "conducting a combined arms campaign across all domains -- land, air, maritime, space, and cyberspace." Meanwhile, growing piles of books and articles explore the threats of cyberwarfare, cyberterrorism, and how to survive them.

COMMENTS (8)SHARE:

Twitter

Reddit

Buzz

More...

Time for a reality check: Cyberwar is still more hype than hazard. Consider the definition of an act of war: It has to be potentially violent, it has to be purposeful, and it has to be political. The cyberattacks we've seen so far, from Estonia to the Stuxnet virus, simply don't meet these criteria.

Take the dubious story of a Soviet pipeline explosion back in 1982, much cited by cyberwar's true believers as the most destructive cyberattack ever. The account goes like this: In June 1982, a Siberian pipeline that the CIA had virtually booby-trapped with a so-called "logic bomb" exploded in a monumental fireball that could be seen from space. The U.S. Air Force estimated the explosion at 3 kilotons, equivalent to a small nuclear device. Targeting a Soviet pipeline linking gas fields in Siberia to European markets, the operation sabotaged the pipeline's control systems with software from a Canadian firm that the CIA had doctored with malicious code. No one died, according to Thomas Reed, a U.S. National Security Council aide at the time who revealed the incident in his 2004 book, At the Abyss; the only harm came to the Soviet economy.

But did it really happen? After Reed's account came out, Vasily Pchelintsev, a former KGB head of the Tyumen region, where the alleged explosion supposedly took place, denied the story. There are also no media reports from 1982 that confirm such an explosion, though accidents and pipeline explosions in the Soviet Union were regularly reported in the early 1980s. Something likely did happen, but Reed's book is the only public mention of the incident and his account relied on a single document. Even after the CIA declassified a redacted version of Reed's source, a note on the so-called Farewell Dossier that describes the effort to provide the Soviet Union with defective technology, the agency did not confirm that such an explosion occurred. The available evidence on the Siberian pipeline blast is so thin that it shouldn't be counted as a proven case of a successful cyberattack.

Most other commonly cited cases of cyberwar are even less remarkable. Take the attacks on Estonia in April 2007, which came in response to the controversial relocation of a Soviet war memorial, theBronze Soldier. The well-wired country found itself at the receiving end of a massive distributed denial-of-service attack that emanated from up to 85,000 hijacked computers and lasted three weeks. The attacks reached a peak on May 9, when 58 Estonian websites were attacked at once and the online services of Estonia's largest bank were taken down. "What's the difference between a blockade of harbors or airports of sovereign states and the blockade of government institutions and newspaper websites?" asked Estonian Prime Minister Andrus Ansip.

Despite his analogies, the attack was no act of war. It was certainly a nuisance and an emotional strike on the country, but the bank's actual network was not even penetrated; it went down for 90 minutes one day and two hours the next. The attack was not violent, it wasn't purposefully aimed at changing Estonia's behavior, and no political entity took credit for it. The same is true for the vast majority of cyberattacks on record.

Indeed, there is no known cyberattack that has caused the loss of human life. No cyberoffense has ever injured a person or damaged a building. And if an act is not at least potentially violent, it's not an act of war. Separating war from physical violence makes it a metaphorical notion; it would mean that there is no way to distinguish between World War II, say, and the "wars" on obesity and cancer. Yet those ailments, unlike past examples of cyber "war," actually do kill people.

Illustration by Francesco Bongiorni for FP

"A Digital Pearl Harbor Is Only a Matter of Time."

Keep waiting. U.S. Defense Secretary Leon Panetta delivered a stark warning last summer: "We could face a cyberattack that could be the equivalent of Pearl Harbor." Such alarmist predictions have been ricocheting inside the Beltway for the past two decades, and some scaremongers have even upped the ante by raising the alarm about a cyber 9/11. In his 2010 book, Cyber War, former White House counterterrorism czar Richard Clarke invokes the specter of nationwide power blackouts, planes falling out of the sky, trains derailing, refineries burning, pipelines exploding, poisonous gas clouds wafting, and satellites spinning out of orbit -- events that would make the 2001 attacks pale in comparison.

But the empirical record is less hair-raising, even by the standards of the most drastic example available. Gen. Keith Alexander, head of U.S. Cyber Command (established in 2010 and now boasting a budget of more than $3 billion), shared his worst fears in an April 2011 speech at the University of Rhode Island: "What I'm concerned about are destructive attacks," Alexander said, "those that are coming." He then invoked a remarkable accident at Russia's Sayano-Shushenskaya hydroelectric plant to highlight the kind of damage a cyberattack might be able to cause. Shortly after midnight on Aug. 17, 2009, a 900-ton turbine was ripped out of its seat by a so-called "water hammer," a sudden surge in water pressure that then caused a transformer explosion. The turbine's unusually high vibrations had worn down the bolts that kept its cover in place, and an offline sensor failed to detect the malfunction. Seventy-five people died in the accident, energy prices in Russia rose, and rebuilding the plant is slated to cost $1.3 billion.

Tough luck for the Russians, but here's what the head of Cyber Command didn't say: The ill-fated turbine had been malfunctioning for some time, and the plant's management was notoriously poor. On top of that, the key event that ultimately triggered the catastrophe seems to have been a fire at Bratsk power station, about 500 miles away. Because the energy supply from Bratsk dropped, authorities remotely increased the burden on the Sayano-Shushenskaya plant. The sudden spike overwhelmed the turbine, which was two months shy of reaching the end of its 30-year life cycle, sparking the catastrophe.

If anything, the Sayano-Shushenskaya incident highlights how difficult a devastating attack would be to mount. The plant's washout was an accident at the end of a complicated and unique chain of events. Anticipating such vulnerabilities in advance is extraordinarily difficult even for insiders; creating comparable coincidences from cyberspace would be a daunting challenge at best for outsiders. If this is the most drastic incident Cyber Command can conjure up, perhaps it's time for everyone to take a deep breath.

"Cyberattacks Are Becoming Easier."

Just the opposite. U.S. Director of National Intelligence James R. Clapper warned last year that the volume of malicious software on American networks had more than tripled since 2009 and that more than 60,000 pieces of malware are now discovered every day. The United States, he said, is undergoing "a phenomenon known as 'convergence,' which amplifies the opportunity for disruptive cyberattacks, including against physical infrastructures." ("Digital convergence" is a snazzy term for a simple thing: more and more devices able to talk to each other, and formerly separate industries and activities able to work together.)

Just because there's more malware, however, doesn't mean that attacks are becoming easier. In fact, potentially damaging or life-threatening cyberattacks should be more difficult to pull off. Why? Sensitive systems generally have built-in redundancy and safety systems, meaning an attacker's likely objective will not be to shut down a system, since merely forcing the shutdown of one control system, say a power plant, could trigger a backup and cause operators to start looking for the bug. To work as an effective weapon, malware would have to influence an active process -- but not bring it to a screeching halt. If the malicious activity extends over a lengthy period, it has to remain stealthy. That's a more difficult trick than hitting the virtual off-button.

Take Stuxnet, the worm that sabotaged Iran's nuclear program in 2010. It didn't just crudely shut down the centrifuges at the Natanz nuclear facility; rather, the worm subtly manipulated the system. Stuxnet stealthily infiltrated the plant's networks, then hopped onto the protected control systems,intercepted input values from sensors, recorded these data, and then provided the legitimate controller code with pre-recorded fake input signals, according to researchers who have studied the worm. Its objective was not just to fool operators in a control room, but also to circumvent digital safety and monitoring systems so it could secretly manipulate the actual processes.

Building and deploying Stuxnet required extremely detailed intelligence about the systems it was supposed to compromise, and the same will be true for other dangerous cyberweapons. Yes, "convergence," standardization, and sloppy defense of control-systems software could increase the risk of generic attacks, but the same trend has also caused defenses against the most coveted targets to improve steadily and has made reprogramming highly specific installations on legacy systems more complex, not less.

"Cyberweapons Can Create
Massive Collateral Damage."

Very unlikely. When news of Stuxnet broke, the New York Times reported that the most striking aspect of the new weapon was the "collateral damage" it created. The malicious program was "splattered on thousands of computer systems around the world, and much of its impact has been on those systems, rather than on what appears to have been its intended target, Iranian equipment," theTimes reported. Such descriptions encouraged the view that computer viruses are akin to highly contagious biological viruses that, once unleashed from the lab, will turn against all vulnerable systems, not just their intended targets.

But this metaphor is deeply flawed. As the destructive potential of a cyberweapon grows, the likelihood that it could do far-reaching damage across many systems shrinks. Stuxnet did infect more than 100,000 computers -- mainly in Iran, Indonesia, and India, though also in Europe and the United States. But it was so specifically programmed that it didn't actually damage those machines, afflicting only Iran's centrifuges at Natanz. The worm's aggressive infection strategy was designed to maximize the likelihood that it would reach its intended target. Because that final target was not networked, "all the functionality required to sabotage a system was embedded directly in the Stuxnet executable," the security software company Symantec observed in its analysis of the worm's code. So yes, Stuxnet was "splattered" far and wide, but it only executed its damaging payload where it was supposed to.

Collateral infection, in short, is not necessarily collateral damage. A sophisticated piece of malware may aggressively infect many systems, but if there is an intended target, the infection will likely have a distinct payload that will be harmless to most computers. Especially in the context of more sophisticated cyberweapons, the image of inadvertent collateral damage doesn't hold up. They're more like a flu virus that only makes one family sick.

"In Cyberspace, Offense Dominates Defense."

Wrong again. The information age has "offense-dominant attributes," Arquilla and Ronfeldt wrote in their influential 1996 book, The Advent of Netwar. This view has spread through the American defense establishment like, well, a virus. A 2011 Pentagon report on cyberspace stressed "the advantage currently enjoyed by the offense in cyberwarfare." The intelligence community stressed the same point in its annual threat report to Congress last year, arguing that offensive tactics -- known as vulnerability discovery and exploitation -- are evolving more rapidly than the federal government and industry can adapt their defensive best practices. The conclusion seemed obvious: Cyberattackers have the advantage over cyberdefenders, "with the trend likely getting worse over the next five years."

A closer examination of the record, however, reveals three factors that put the offense at a disadvantage. First is the high cost of developing a cyberweapon, in terms of time, talent, and target intelligence needed. Stuxnet, experts speculate, took a superb team and a lot of time. Second, the potential for generic offensive weapons may be far smaller than assumed for the same reasons, and significant investments in highly specific attack programs may be deployable only against a very limited target set. Third, once developed, an offensive tool is likely to have a far shorter half-life than the defensive measures put in place against it. Even worse, a weapon may only be able to strike a single time; once the exploits of a specialized piece of malware are discovered, the most critical systems will likely be patched and fixed quickly. And a weapon, even a potent one, is not much of a weapon if an attack cannot be repeated. Any political threat relies on the credible threat to attack or to replicate a successful attack. If that were in doubt, the coercive power of a cyberattack would be drastically reduced.

"We Need a Cyberarms Control Agreement."

We don't. Cyberwar alarmists want the United States to see cybersecurity as a new challenge on a geopolitical scale. They see cyberspace becoming a new area for military competition with rivals such as Russia and China, and they believe new cyberarms limitation agreements are needed to prevent this. There are some rumblings to establish international norms on this topic: The British government convened a conference in London in late 2011, originally intended to make the Internet more secure by agreeing on new rules of the road, and Russia and China proposed at the U.N. General Assembly last September the establishment of an "international code of conduct for information security." Now, diplomats are debating whether the United Nations should try to forge the equivalent of nuclear arms control in cyberspace.

So, should it? The answer is no. Attempts to limit cyberweapons through international agreements have three principal problems. The first difficulty is drawing the line between cybercrime and potentially political activity in cyberspace. In January, for instance, a Saudi hacker stole about 20,000 Israeli credit card numbers from a shopping website and leaked the information to the public. In retaliation, a group of Israeli hackers broke into Saudi shopping sites and threatened to release private credit card information.

Where is the dividing line? Even if it were possible to distinguish criminal from state-sponsored political activity, they often use the same means. A second hitch is practical: Verification would be impossible. Accurately counting the size of nuclear arsenals and monitoring enrichment activities is already a huge challenge; installing cameras to film programmers and "verify" they don't design malicious software is a pipe dream.

The third problem is political, and even more fundamental: Cyberaggressors may act politically, but in sharp contrast with warfare, they are likely to have a strong interest in avoiding attribution. Subversion has always thrived in cyberspace because preserving one's anonymity is easier to achieve than ironclad attribution. That's the root of the political problem: Having a few states agree on cyberarms limitation is about as realistic as a treaty to outlaw espionage and about as practical as outlawing the general subversion of established order.

"The West Is Falling Behind Russia and China."

Yes, but not how you think. Russia and China are busy sharpening their cyberweapons and are already well steeped in using them. The Russian military clandestinely crippled Estonia's economy in 2007 and Georgia's government and banks in 2008. The People's Liberation Army's numerous Chinese cyberwarriors have long inserted "logic bombs" and "trapdoors" into America's critical infrastructure, lying dormant and ready to wreak havoc on the country's grid and bourse in case of a crisis. Both countries have access to technology, cash, and talent -- and have more room for malicious maneuvers than law-abiding Western democracies poised to fight cyberwar with one hand tied behind their backs.

Or so the alarmists tell us. Reality looks quite different. Stuxnet, by far the most sophisticated cyberattack on record, was most likely a U.S.-Israeli operation. Yes, Russia and China have demonstrated significant skills in cyberespionage, but the fierceness of Eastern cyberwarriors and their coded weaponry is almost certainly overrated. When it comes to military-grade offensive attacks, America and Israel seem to be well ahead of the curve.

Ironically, it's a different kind of cybersecurity that Russia and China may be more worried about. Why is it that those countries, along with such beacons of liberal democracy as Uzbekistan, have suggested that the United Nations establish an "international code of conduct" for cybersecurity? Cyberespionage was elegantly ignored in the suggested wording for the convention, as virtual break-ins at the Pentagon and Google remain a favorite official and corporate pastime of both countries. But what Western democracies see as constitutionally protected free speech in cyberspace, Moscow and Beijing regard as a new threat to their ability to control their citizens. Cybersecurity has a broader meaning in non-democracies: For them, the worst-case scenario is not collapsing power plants, but collapsing political power.

The social media-fueled Arab Spring has provided dictators with a case study in the need to patrol cyberspace not only for subversive code, but also for subversive ideas. The fall of Egypt's Hosni Mubarak and Libya's Muammar al-Qaddafi surely sent shivers down the spines of officials in Russia and China. No wonder the two countries asked for a code of conduct that helps combat activities that use communications technologies -- "including networks" (read: social networks) -- to undermine "political, economic and social stability."

So Russia and China are ahead of the United States, but mostly in defining cybersecurity as the fight against subversive behavior. This is the true cyberwar they are fighting.

Thomas Rid, reader in war studies at King's College London, is author of "Cyber War Will Not Take Place."