Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 1
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy THE NATIONAL ACADEMIES Advisers to the Nation on Science, Engineering, and Medicine National Academy of Sciences National Academy of Engineering Institute of Medicine National Research Council March 25, 2010 Mr. Brian Overington Assistant Deputy Director of National Intelligence Office of the Director of National Intelligence Washington, DC 20511 Dear Mr. Overington: This letter report from the National Research Council’s (NRC’s) Committee on Deterring Cyberattacks is the first deliverable for Contract Number HHM-402-05-D-0011, DO#12. This committee (biographies of committee members are provided in Attachment 1) was created to help inform strategies for deterring cyberattacks and to develop options for U.S. policy in this area. The project statement of task is provided below: An ad hoc committee will oversee an activity to foster a broad, multidisciplinary examination of deterrence strategies and their possible utility to the U.S. government in its policies toward preventing cyberattacks. In the first phase, the committee will prepare a letter report identifying the key issues and questions that merit examination. In the next phase, the committee will engage experts to prepare papers that address key issues and questions, including those posed in the letter report. The papers will be compiled in a National Research Council publication and/or published by appropriate journals. This phase will include a committee meeting and a workshop to discuss draft papers, with authors finalizing the papers following the workshop. This letter report satisfies the deliverable requirement of the first phase of the project by providing basic information needed to understand the nature of the problem and to articulate important questions that can drive research regarding ways of more effectively preventing, discouraging, and inhibiting hostile activity against important U.S. information systems and networks. (Attachment 2 acknowledges the reviewers of this letter report.) The second phase of this project will entail selection of appropriate experts to write papers on questions raised in this report. Much of the analytical framework of this letter report draws heavily on reports previously issued by the NRC.1 In particular, it builds in large part on the work of a 1 National Research Council (NRC), Technology, Policy, Law, and Ethics Regarding U.S. Acquisition and Use of Cyberattack Capabilities (William Owens, Kenneth Dam, Herbert Lin, editors), The National Academies Press, Washington, D.C., 2009;
OCR for page 2
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy previous NRC panel (the NRC Committee on Offensive Information Warfare), which issued a report entitled Technology, Policy, Law, and Ethics Regarding Acquisition and Use of U.S. Cyberattack Capabilities in April 2009, and extracts without specific attribution sections from Chapters 2, 9, and 10 of that report. In addition and as requested by the Office of the Director of National Intelligence (ODNI), the committee reviewed the ODNI-provided compendiums on three summer workshops conducted by the ODNI,2 and incorporated insights and issues from them into this report as appropriate. This report consists of three main sections. Section 1 describes a broad context for cybersecurity, establishing its importance and characterizing the threat. Section 2 sketches a range of possible approaches for how the nation might respond to cybersecurity threats, emphasizing how little is known about how such approaches might be effective in an operational role. Section 3 describes a research agenda intended to develop more knowledge and insight into these various approaches. As for the second phase of this project, a workshop will be held in June 2010 to discuss a number of papers that have been commissioned by the committee and possibly additional papers received through the NRC’s call for papers. This call for papers is at the heart of a competition sponsored by the NRC to solicit excellent papers on the subject of cyberdeterrence. The call for papers can be found at http://sites.nationalacademies.org/CSTB/CSTB_056215. 1. The Broad Context for Cybersecurity3 Today, it is broadly accepted that the U.S. military and economic power is ever more dependent on information and information technology. Accordingly, maintaining the security of important information and information technology systems against hostile action (a topic generally referred to as “cybersecurity”) is a problem of increasing importance to policy makers. Accordingly, an important policy goal of the United States is to prevent, discourage, and inhibit hostile activity against these systems and networks. This project was established to address cyberattacks, which refer to the deliberate use of cyber operations—perhaps over an extended period of time—to alter, disrupt, deceive, degrade, usurp, or destroy adversary computer systems or networks or the information and/or programs resident in or transiting these systems or networks.4 Cyberattack is NRC, Toward a Safer and More Secure Cyberspace (Seymour Goodman and Herbert Lin, editors), The National Academies Press, Washington, D.C., 2007. 2 These workshops addressed the role of the private sector, deterrence, and attribution. 3 The discussion in this section is based on Chapter 1, NRC, Technology, Policy, Law, and Ethics Regarding U.S. Acquisition and Use of Cyberattack Capabilities, 2009; and Chapter 2, NRC, Toward a Safer and More Secure Cyberspace, 2007. 4 This report does not consider the use of electromagnetic pulse (EMP) attacks. EMP attacks typically refer to nonselective attacks using nuclear weapons to generate an intense electromagnetic pulse that can destroy all unprotected electronics and electrical components within a large area, although a tactical EMP weapon intended to
OCR for page 3
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy not the same as cyber exploitation, which is an intelligence-gathering activity rather than a destructive activity and refers to the use of cyber operations—perhaps over an extended period of time—to support the goals and missions of the party conducting the exploitation, usually for the purpose of obtaining information resident on or transiting through an adversary’s computer systems or networks. Cyberattack and cyber exploitation are technically very similar, in that both require a vulnerability, access to that vulnerability, and a payload to be executed. They are technically different only in the nature of the payload to be executed. These technical similarities often mean that a targeted party may not be able to distinguish easily between a cyber exploitation and a cyberattack. Because of the ambiguity of cyberattack and cyber exploitation from the standpoint of the targeted party, it is helpful to have a word to refer to a hostile cyber activity where the nature of the activity is not known (that is, an activity that could be either a cyberattack or a cyber exploitation)—in this report, the term cyberintrusion is used to denote such activity. The range of possibilities for cyberintrusion is quite broad.5 A cyberattack might result in the destruction of relatively unimportant data or the loss of availability of a secondary computer system for a short period of time—or it might alter top-secret military plans or degrade the operation of a system critical to the nation, such as an air traffic control system, a power grid, or a military command and control system. Cyber exploitations might target the personal information of individual consumers or critical trade secrets of a business, military war plans, or design specifications for new weapons. Although all such intrusions are worrisome, some of these are of greater significance to the national well-being than others. Intrusions are conducted by a range of parties, including disgruntled or curious individuals intent on vandalizing computer systems, criminals (sometimes criminal organizations) intent on stealing money, terrorist groups intent on sowing fear or seeking attention to their causes, and nation-states for a variety of national purposes. Moreover, it must be recognized that nation-states can tolerate, sponsor, or support terrorist groups, criminals, or even individuals as they conduct their intrusions. A state might tolerate individual hackers who wish to vandalize an adversary’s computer systems, perhaps for the purpose of sowing chaos. Or it might sponsor or hire criminal organizations with special cyber expertise to carry out missions that it did not have the expertise to undertake. Or it might provide support to terrorist groups by looking the other way as those groups use the infrastructure of the state to conduct Internet-based operations. In times of crisis or conflict, a state might harbor (or fail to discourage, or encourage, or control) “patriotic hackers” or “cyber patriots” who conduct hostile cyberintrusions against a putative adversary. Note that many such actions would also be plausibly deniable by the government of the host state. selectively target such components on a small scale is possible to imagine. For a comprehensive description of the threat from EMP attacks, see Report of the Commission to Assess the Threat to the United States from Electromagnetic Pulse (EMP) Attack, available at http://www.globalsecurity.org/wmd/library/congress/2004_r/04-07-22emp.pdf. 5 Chapter 1, NRC, Technology, Policy, Law, and Ethics Regarding U.S. Acquisition and Use of Cyberattack Capabilities, 2009.
OCR for page 4
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy The threats that adversaries pose can be characterized along two dimensions—the sophistication of the intrusion and the damage it causes. Though these two are often related, they are not the same. Sophistication is needed to penetrate good cyberdefenses, and the damage an intrusion can cause depends on what the adversary does after it has penetrated those defenses. As a general rule, a greater availability of resources to the adversary (e.g., more money, time, talent) will tend to increase the sophistication of the intrusion that can be launched against any given target and thus the likelihood that the adversary will be able to penetrate the target’s defenses. Two important consequences follow from this discussion. First, because nation-state adversaries can bring to bear enormous resources to conduct an intrusion, the nation-state threat (perhaps conducted through intermediaries) is the most difficult to defend against. Second, stronger defenses reduce the likelihood but cannot eliminate the possibility that even less sophisticated adversaries can cause significant damage. 2. A Range of Possibilities The discussion below focuses primarily on cyberattacks as the primary policy concern of the United States, and addresses cyber exploitation as necessary. 2.1 THE LIMITATIONS OF PASSIVE DEFENSE AND SOME ADDITIONAL OPTIONS The central policy question is how to achieve a reduction in the frequency, intensity, and severity of cyberattacks on U.S. computer systems and networks currently being experienced and how to prevent the far more serious attacks that are in principle possible. To promote and enhance the cybersecurity of important U.S. computer systems and networks (and the information contained in or passing through these systems and networks), much attention has been devoted to passive defense—measures taken unilaterally to increase the resistance of an information technology system or network to attack. These measures include hardening systems against attack, facilitating recovery in the event of a successful attack, making security more usable and ubiquitous, and educating users to behave properly in a threat environment.6 Passive defenses for cybersecurity are deployed to increase the difficulty of conducting the attack and reduce the likelihood that a successful attack will have significant negative consequences. But experience and recent history have shown that they do not by themselves provide an adequate degree of cybersecurity for important information systems and networks. A number of factors explain the limitations of passive defense. As noted in previous NRC reports,7 today’s decision-making calculus regarding cybersecurity 6 As an example, see NRC, Toward a Safer and More Secure Cyberspace, 2007. 7 National Research Council, Cybersecurity Today and Tomorrow: Pay Now or Pay Later, The National Academies Press, Washington, D.C., 2002; NRC, Toward a Safer and More Secure Cyberspace, 2007.
OCR for page 5
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy excessively focuses vendor and end-user attention on the short-term costs of improving their individual cybersecurity postures to the detriment of the national cybersecurity posture as a whole. As a result, much of the critical infrastructure on which the nation depends is inadequately protected against cyberintrusion. A second important factor is that passive defensive measures must succeed every time an adversary conducts a hostile action, whereas the adversary’s action need succeed only once. Put differently, attacks can be infinitely varied, whereas defenses are only as strong as their weakest link. This fact places a heavy and asymmetric burden on a defensive posture that employs only passive defense. Because passive defenses do not eliminate the possibility that an attack might succeed, it is natural for policy makers to seek other mechanisms to deal with threats that passive defenses fail to address adequately. Policy makers understandably aspire to a goal of preventing cyberattacks (and cyber exploitations as well), but most importantly to a goal of preventing serious cyberattacks—cyberattacks that have a disabling or a crippling effect on critical societal functions on a national scale (e.g., military mission readiness, air traffic control, financial services, provision of electric power). In this context, “deterrence” refers to a tool or a method used to help achieve this goal. The term “deterrence” itself has a variety of connotations, but broadly speaking, deterrence is a tool for dissuading an adversary from taking hostile actions. Adversaries that might conduct cyberintrusions against the United States span a broad range and may well have different objectives. Possible adversaries include nation-states that would use cyberattacks to collect intelligence, steal technology, or “prepare the battlefield” for use of cyberattacks either by themselves or as part of a broader effort (perhaps involving the use or threat of use of conventional force) to coerce the United States; sophisticated elements within a state that might not be under the full control of the central government (e.g., Iranian Revolutionary Guards); criminal organizations seeking illicit monies; terrorist groups operating without state knowledge; and so on. In principle, policy makers have a number of approaches at their disposal to further the broad goal of preventing serious cyberattacks on the United States. In contrast to passive defense, all of these approaches depend on the ability to attribute hostile actions to specific responsible parties (although the precise definition of “responsible party” depends to a certain extent on context). The first approach, and one of the most common, is the use of law enforcement authorities to investigate cyberattacks, and then identify and prosecute the human perpetrators who carry out these attacks. Traditionally, law enforcement actions serve two purposes. First, when successful, they remove such perpetrators from conducting further hostile action, at least for a period of time. Second, the punishment imposed on perpetrators is intended to dissuade other possible perpetrators from conducting similar actions. However, neither of these purposes can be served if the cyberattacks in question cannot be attributed to specific perpetrators. In a cyber context, law enforcement investigations and prosecutions have had some success, but the time scale on which such activities yield results is typically on the order of months, during which time cyberattacks often continue to plague the victim. As a result, most victims have no way to stop an attack that is causing ongoing damage or loss of information. In addition, the likelihood that any given attack will be successfully investigated and prosecuted is low, thus reducing any potential deterrent effect.
OCR for page 6
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy Notwithstanding the potential importance of law enforcement activities for the efficacy of possible deterrence strategies, law enforcement activities are beyond the scope of this report and will not be addressed further herein. A second approach relies on deterrence as it is classically understood. The classical model of deterrence (discussed further in Section 2.2) seeks to prevent hostile actions through the threat of retaliation or responsive action that imposes unacceptable costs on a potential adversary or denies an adversary the benefits that may result from taking those hostile actions. Deterrence thus includes active defense, in which actions can be taken to neutralize an incoming cyberattack. A third approach takes note of the fact that the material threat of retaliation underlying deterrence is not the only method of inhibiting undesirable behavior. Behavioral restraint (discussed further in Section 2.3) is more often the result of formal law and informal social norms, and the burden of enforcement depends a great deal on the robustness of such rules and the pressures to conform to those rules that can be brought to bear through the social environment that the various actors inhabit. These approaches—and indeed an approach based on passive defense—are by no means mutually exclusive. For example, some combination of strengthened passive defenses, deterrence, law enforcement, and negotiated behavioral restraint may be able to reduce the likelihood that highly destructive cyberattacks would be attempted and to minimize the consequences if cyberattacks do occur. But how well any of these approaches can or will work to prevent cyberattacks (or cyberintrusions more broadly) is open to question, and indeed is one primary subject of the papers to be commissioned for this project. 2.2 CLASSICAL DETERRENCE8 Many analysts have been drawn to the notion of deterring hostile activity against important IT systems and networks, rather than just defending against such activity. Deterrence seems like an inevitable choice in an offense-dominant world—that is, a world in which offensive technologies and tactics are generally capable of thwarting defensive efforts. As noted in Section 2.1, a major difficulty of defending against hostile actions in cyberspace arises from the asymmetry of offense versus defense. Deterrence was and is a central construct in contemplating the use of nuclear weapons and in nuclear strategy. Because effective defenses against nuclear weapons are difficult to construct, using the threat of retaliation to persuade an adversary to refrain from using nuclear weapons is regarded by many as the most plausible and effective alternative to ineffective or useless defenses. Indeed, deterrence of nuclear threats in the Cold War establishes the paradigm in which the conditions for successful deterrence are largely met. Although the threat of retaliation is not the only possible mechanism for practicing deterrence, such a threat is in practice the principal and most problematic 8 The discussion in Section 2.2 is based on Chapter 9, NRC, Technology, Policy, Law, and Ethics Regarding U.S. Acquisition and Use of Cyberattack Capabilities, 2009.
OCR for page 7
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy method implied by use of the term.9 Extending traditional deterrence principles to cyberattack (that is, cyberdeterrence) would suggest an approach that seeks to persuade adversaries to refrain from launching cyberattacks against U.S. interests, recognizing that cyberdeterrence would be only one of a suite of elements of U.S. national security policy. But it is an entirely open question whether cyberdeterrence is a viable strategy. Although nuclear weapons and cyber weapons share one key characteristic (the superiority of offense over defense), they differ in many other key characteristics, and the section below discusses cyberdeterrence and when appropriate contrasts cyberdeterrence to Cold War nuclear deterrence. What the discussion below will suggest is that nuclear deterrence and cyberdeterrence do raise many of the same questions, but indeed that the answers to these questions are quite different in the cyber context than in the nuclear context. The U.S. Strategic Command formulates deterrence as follows:10 Deterrence [seeks to] convince adversaries not to take actions that threaten U.S. vital interests by means of decisive influence over their decision-making. Decisive influence is achieved by credibly threatening to deny benefits and/or impose costs, while encouraging restraint by convincing the actor that restraint will result in an acceptable outcome. For purposes of this report, the above formulation will be used to organize the remainder of this section, by discussing at greater length the words in bold above. Nevertheless, the committee does recognize that there are other plausible formulations of the concept of deterrence, and that these formulations might differ in tone and nuance from that provided above. 2.2.1 “Convince” At its root, convincing an adversary is a psychological process. Classical deterrence theory assumes that actors make rational assessments of costs and benefits and refrain from taking actions where costs outweigh benefits. But it assumes unitary actors (i.e., a unitary decision maker whose cost-benefit calculus is determinative for all of the forces under his control), and also that the costs and benefits of each actor are clear, well-defined, and indeed known to all other actors involved, and further that these costs and benefits are sufficiently stable over time to formulate and implement a deterrence strategy. Classical deterrence theory bears many similarities to neoclassical 9 Analysts also invoke the concept of deterrence by denial, which is based on the prospect of deterring an adversary through the prospect of failure to achieve its goals—facing failure, the adversary chooses to refrain from acting. But denial is—by definition—difficult to practice in an offense-dominant world. 10 U.S. Department of Defense, Deterrence Operations: Joint Operating Concept, Version 2.0, December 2006, available at http://www.dtic.mil/futurejointwarfare/concepts/do_joc_v20.doc.
OCR for page 8
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy economics, especially in its assumptions about the availability of near-perfect information (perfect in the economic sense) about all actors. Perhaps more importantly, real decisions often take place during periods of crisis, in the midst of uncertainty, doubt, and fear that often lead to unduly pessimistic assessments. Even a cyberattack conducted in peacetime is more likely to be carried out under circumstances of high uncertainty about the effectiveness of technology on both sides, the motivations of an adversary, and the effects of an attack. In addition, cyber conflict is relatively new, and there is not much known about how cyber conflict would or could evolve in any given situation. History shows that when human beings with little hard information are placed into unfamiliar situations in a general environment of tension, they often substitute supposition for knowledge. In the words of a former senior administration official responsible for protecting U.S. critical infrastructure, “I have seen too many situations where government officials claimed a high degree of confidence as to the source, intent, and scope of a [cyber]attack, and it turned out they were wrong on every aspect of it. That is, they were often wrong, but never in doubt.”11 As an example, cyber operations that would be regarded as unfriendly during normal times may be regarded as overtly hostile during periods of crisis or heightened tension. Cyber operations X, Y, and Z undertaken by party A (with a history of neutrality) may be regarded entirely differently if undertaken by party B (with a history of acting against U.S. interests). Put differently, reputations and past behavior matter—how we regard or attribute certain actions that happen today will depend on what has happened in the past. This point has particular relevance as U.S. interest in obtaining offensive capabilities in cyberspace becomes more apparent. The United States is widely regarded as the world leader in information technology, and such leadership can easily be seen by the outside world as enabling the United States to conceal the origin of any offensive cyber operation that it might have conducted. That is, many nations will find it plausible that the United States is involved in any such operation against it, and even if no U.S.-specific “fingerprints” can be found, such a fact can easily be attributed to putative U.S. technological superiority in conducting such operations. Lastly, a potential adversary will not be convinced to refrain from hostile action if it is not aware of measures the United States may take to retaliate. Thus, some minimum of information about deterrence policy must be known and openly declared. This point is further addressed in Section 2.2.4. 2.2.2 “Adversaries” In the Cold War paradigm of nuclear deterrence, the world is state-centric and bipolar. It was reasonable to presume that only nation-states could afford to assemble the substantial infrastructure needed to produce the required fissile material and develop nuclear weapons and their delivery vehicles. That infrastructure was sufficiently visible that an intelligence effort directed at potential adversaries could keep 11 See NRC, Technology, Policy, Law, and Ethics Regarding Acquisition and Use of U.S. Cyberattack Capabilities, 2009, page 142.
OCR for page 9
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy track of the nuclear threat that possible adversaries posed to the United States. Today’s concerns about terrorist use of nuclear weapons arise less from a fear that terrorists will develop and build their own nuclear weapons and more from a fear that they will be able to obtain nuclear weapons from a state that already has them. These characteristics do not apply to the development of weapons for cyberattack. Many kinds of cyberattack can be launched with infrastructure, technology, and background knowledge easily and widely available to nonstate parties and small nations. Although national capabilities may be required for certain kinds of cyberattack (such as those that involve extensive hardware modification or highly detailed intelligence regarding truly closed and isolated system and networks), substantial damage can be inflicted by cyberattacks based on ubiquitous technology. A similar analysis holds for identifying the actor responsible for an attack. In the nuclear case, an attack on the United States would have been presumed to be Soviet in origin because the world was bipolar. In addition, surveillance of potential launch areas provided high-confidence information regarding the fact of a launch, and also its geographical origin—a missile launch from the land mass of any given nation could be safely attributed to a decision by that nation’s government to order that launch. Sea-based or submarine-based launches are potentially problematic in this regard, although in a bipolar world, the Soviet Union would have been deemed responsible. In a world with three potential nuclear adversaries (the United States, Soviet Union, and China), intensive intelligence efforts have been able to maintain to a considerable extent the capability for attributing a nuclear attack to a national power, through measures such as tracking adversary ballistic missile submarines at sea. Identification of the distinctive radiological signatures of potential adversaries’ nuclear weapons is also believed to have taken place. The nuclear deterrence paradigm also presumes unitary actors, nominally governments of nation-states—that is, it presumes that the nuclear forces of a nation are under the control of the relevant government, and that they would be used only in accordance with the decisions of national leaders. These considerations do not hold for cyberattack, and for many kinds of cyberattack the United States would almost certainly not be able to ascertain the source of such an attack, even if it were a national act, let alone hold a specific nation responsible. For example, the United States is constantly under cyberattack today, and it is widely believed (though without conclusive proof) that most of these cyberattacks are not the result of national decisions by an adversary state, though press reports have claimed that some are. In general, prompt technical attribution of an attack or exploitation—that is, identification of the responsible party (individual? subnational group? nation-state?) based only on technical indicators associated with the event in question—is quite problematic, and any party accused of launching a given cyberintrusion could deny it with considerable plausibility. Forensic investigation might yield the identity of the responsible party, but the time scale for such investigation is often on the order of weeks or months. (Although it is often quite straightforward to trace an intrusion to the proximate node, in general, this will not be the origination point of the intrusion. Tracing an intrusion to its actual origination point past intermediate nodes is what is most difficult.)
OCR for page 10
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy Three factors mitigate to some (unknowable) degree this bleak picture regarding attribution. First, for reasons of its own, a cyberattacker may choose to reveal to its target its responsibility for a cyberattack. For example, it may conduct a cyberattack of limited scope to demonstrate its capability for doing so, acknowledge its responsibility, and then threaten to conduct a much larger one if certain demands are not met.12 Second, over time a series of cyberintrusions might be observed to share important technical features that constitute a “signature” of sorts. Thus, the target of a cyberattack may be able to say that it was victimized by a cyberattack of type X on 16 successive occasions over the last 3 months. An inference that the same party was responsible for that series of attack might under some circumstances have some plausibility. Third, the target of a cyberattack may have nontechnical information that points to a perpetrator, such as information from a well-placed spy in an adversary’s command structure or high-quality signals intelligence. If such a party reports that the adversary’s forces have just launched a cyberattack against the United States, or if a generally reliable communications intercept points to such responsibility, such information might be used to make a plausible inference about the state responsible for that attack. Political leaders in particular will not rely only on technical indicators to determine the state responsible for an attack—rather, they will use all sources of information available to make the best possible determination. Nevertheless, it is fair to say that absent unusually good intelligence information, high confidence in the attribution of a cyberattack to a nation-state is almost certain to be unattainable during and immediately after that attack, and may not be achievable for a long time afterward. Thus, any retaliatory response to a cyberattack using either cyber or kinetic weaponry may carry a significant risk of being directed improperly, perhaps with grave unintended consequences. 2.2.3 “Actions that threaten U.S. vital interests” What actions is the United States trying to deter, and would the United States know that an action has occurred that threatens its vital interests? A nuclear explosion on U.S. territory is an unambiguously large and significant event, and there is little difficulty in identifying the fact of such an explosion. The United States maintains a global network of satellites that are capable of detecting and locating nuclear explosions in the air and on the ground, and a network of seismic sensors that provide additional information to localize nuclear explosions. Most importantly, a nuclear explosion would occur against the very quiet background of zero nuclear explosions happening over time. But U.S. computer and communications systems and networks are under constant cyberintrusion from many different parties, and against this background noise, 12 Of course, a forensic investigation might still be necessary to rule out the possibility that the putative attacker was only claiming responsibility for the attack when in fact it had no real ability to conduct the attack on its own. To mitigate the possibility that it might not be believed, the party claiming responsibility could leave a “calling card” in the wake of an attack whose contents only it could know.
OCR for page 11
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy the United States would have to notice that critical systems and networks were being attacked and damaged. A cyberattack on the United States launched by an adversary might target multiple sites—but correlating information on attacks at different sites against a very noisy background to determine a common cause is today technically challenging. Target sets may be amorphous and complex, especially when massively complex and globally scaled supply chains are involved. And the nature of a questionable event (an intrusion) is often in doubt—is it an attack or an exploitation? If an attack, does a destructive cyberattack take place when the responsible software agent is implanted in a critical U.S. system, or when it is activated? Even knowing the effect or impact of an attack or exploitation is difficult, as the consequences of some intrusions will play out only over an extended period of time. (For example, an attack may be designed to have no immediate impact and only later to show destructive consequences.) Another profound difference between the nuclear and cyber domains is that nuclear weapons are not thought to target individual private sector entities—it would be highly unusual for a major corporation, for example, to be the specific target of a nuclear weapon. By contrast, major corporations are subject to cyberattacks and cyber exploitations on a daily basis. This difference raises the question of whether deterrence of such intrusions on individual private sector entities (especially those that are regarded as a part of U.S. critical infrastructure) is an appropriate goal of U.S. policy—as suggested by recent allegations of Chinese cyberintrusions against human rights activists using Google’s gmail.com service and against multiple private sector companies in the United States seeking important intellectual property of these companies.13 The question is important, because targeted private entities might seek to defend themselves by retaliating against attackers or cyber spies, notwithstanding criminal prohibitions, with consequences damaging to U.S. national interests. The question is important for a number of reasons. First, U.S. military forces have not been used in recent years to support the interests of specific private sector entities, at least not as a matter of declared public policy. Thus, an explicit threat to respond with force, whether cyber or otherwise, to a cyberattack on an individual private sector entity would constitute a major change in U.S. policy. Second, targeted private entities might seek to defend themselves by retaliating against attackers or cyber spies, even though such actions are currently illegal under U.S. law, and such retaliation by these entities might well have consequences damaging to U.S. national interests. 2.2.4 “Credible threat” A credible threat is one that an adversary believes can and will be executed with a sufficiently high probability to dissuade the adversary from taking action. (The definition of “sufficiently high” is subject to much debate and almost certainly depends on the specific case or issue in question. In some cases, even a low absolute 13 See, for example, Ariana Eunjung Cha and Ellen Nakashima, “Google China Cyberattack Part of Vast Espionage Campaign, Experts Say,” Washington Post, January 14, 2010.
OCR for page 22
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy postures and the credibility of the regime? What meaningful capabilities would the United States be giving up if it were to agree to such a regime? These and other related questions find few answers in the literature. The feasibility of these or other regimes to limit use of cyberattack is unclear, especially in light of the difficulties of working out the details of how the regime would actually operate. It is for this reason that research is needed to explore their feasibility. Agreements in a cyber context might also usefully address important collateral issues, such as criminal sanctions or compensation for damages sustained under various circumstances. They might also require signatories to pass national laws that criminalize certain kinds of cyber behavior undertaken by individuals and to cooperate with other nations in prosecuting such behavior, much as the Convention on Cyber Crime has done.21 There are a number of major complications associated with arms control regimes for cyberattack. These include: The functional similarity between cyber exploitation and cyberattack. That is, from the target’s perspective, it may be difficult or impossible to distinguish between a cyber operation intended for attack and one intended for exploitation. Restrictions on cyberattack will almost certainly restrict cyber exploitation to a large degree, and nations—including the United States—may well be loath to surrender even in principle any such capability for gaining intelligence. The lack of state monopoly over cyber weapons. For kinetic weaponry, the destructiveness and potency of any given weapon has some significant correlation with the extent to which it is only available to nation-states—almost everyone has access to rifles, whereas jet fighters and submarines are mostly restricted to nations. For cyber weapons, this correlation is far less strong, and private parties can and do wield some cyber weapons that can be as destructive and powerful as some of those wielded by nation-states. Although as a rule nation-states do have major operational advantages in conducting cyberattacks (e.g., intelligence agencies that can support cyberattack), nonstate actors are certainly capable of acquiring cyber weaponry that can cause enormous damage. “Positive inspection” arrangements to increase the confidence that each side is abiding by an agreement not to engage in proscribed activities could be easily thwarted or circumvented. One primary reason is that the footprint of personnel and equipment needed to conduct cyber operations is small, and thus could be located virtually anywhere in a nation (or even in another nation). In contrast to nuclear weapons, the private sector has essentially unlimited access to most of the technology that underlies cyberattack weapons, and the scope for destructive use varies over a much wider range. Thus, an extraordinary degree of intrusiveness would be required to impose controls on the private acquisition and use of cyber weapons. It would be impractical and unacceptable, not to mention futile, to subject every personal computer and all forms of electronic communication to inspection to ensure that cyber weapons are not present on computers or concealed within e-mails. On the other hand, special rules might help to regulate access to the operations of critical social 21 See http://conventions.coe.int/Treaty/EN/Treaties/html/185.htm.
OCR for page 23
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy infrastructure in order to improve the attribution of parties that come into contact with them. The inherent anonymity of cyberattacks, mentioned above, greatly complicates the attribution of responsibility for an attack, and thus it is difficult to hold violators of any agreement accountable. Any alleged violation could simply be met with a strongly worded denial, and unambiguous evidence supporting the allegation would be hard to provide. Moreover, behavioral norms are generally much harder to instill and enforce in an environment in which actors can act anonymously. Suggestions are often made to create a parallel Internet (call it an SAI, for strongly authenticated Internet) that would provide much stronger authentication of users than is required on today’s Internet and would in other ways provide a much more secure environment.22 If important facilities, such as power grids and financial institutions, migrated to an SAI, accountability for misbehavior would be much greater (because of the lack of anonymity) and the greater security of the environment would mean that only very sophisticated parties could mount attacks on it or within it. Although the availability of an SAI would certainly improve the security environment over that of today, it is not a panacea. Perhaps most importantly, SAI users would immediately become high-priority targets to be compromised by nontechnical cyberattacks. A compromised SAI user would then become an ideal platform from which to launch IT-based cyberattacks within the SAI—and in particular, would become an ideal jumping-off point for slowly and quietly assembling an array of computing resources that can be used for attack—all of which would be on the SAI. In addition, experience with large networks indicates that maintaining an actual air-gap isolation between an SAI and the standard Internet or dial-up or wireless connections would be all but impossible—not for technical reasons but because of a human tendency to make such connections for the sake of convenience. Subnational groups can take action independently of governments. Subnational groups may be particularly difficult to identify, and are likely to have few if any assets that can be targeted. Some groups (such as organized hacker groups) regard counterattacks as a challenge to be welcomed rather than a threat to be feared. Finally, a subnational group composed of terrorists or insurgents might 22 For example, the White House Cyberspace Policy Review of May 2009 called for the nation to “implement, for high-value activities (e.g., the Smart Grid), an opt-in array of interoperable identity management systems to build trust for online transactions.” White House, Cyberspace Policy Review, 2009, available at http://www.whitehouse.gov/assets/documents/Cyberspace_Policy_Review_final.pdf. More recently, a trade press article reported on the intent of the Defense Information Systems Agency of the U.S. Department of Defense to establish an enclave for its unclassified networks that is isolated from public Internet access (Amber Corrin, “DISA to Establish Safe Haven Outside the Internet,” DefenseSystems.com, February 12, 2010, available at http://defensesystems.com/articles/2010/02/12/disa-dmz.aspx?s=ds_170210).
OCR for page 24
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy seek to provoke retaliation in order to galvanize public support for it or to provoke anti-American sentiments in its supporting public. This last point is particularly relevant to any international agreements or regime that the United States might deem helpful in reducing cyberattacks against it—any legal agreement or regime must be respected by all parties, including the United States. If the United States wishes other nations to eschew certain actions or to abide by certain behavioral requirements or to grant it certain rights under certain circumstances, it too must be willing to do the same with respect to other nations. As an example, some analysts have suggested that it is an appropriate strategy for the United States to seek the right to retaliate against a nation for offensive acts emanating from within its borders, even if that nation’s government denies responsibility for those attacks and asserts that those responsible are nonstate actors. Doing so, they argue, would give states an incentive to crack down on harmful private offensive actors in its borders. On the other hand, it is not clear that it is in the U.S. interest for the United States to be subject to such a regime, given that parties within the United States are themselves responsible for conducting many cyberattacks against the rest of the world. Any solution proposed for other nations must (most probably) be tolerable to the United States as well, but accepting such consequences may be politically, or economically, or legally infeasible. It should also be noted that the traditional arms control agreements are not the only form of agreement that might be helpful.23 For example, nations have sometimes agreed on the need to protect some area of international activity such as airline transport, telecommunications, maritime activities, and so on, and have also agreed on standards for such protection. They may declare certain purposes collectively with regard to a given area of activity on which they agree, often in the form of a multilateral treaty, and then establish consensus-based multilateral institutions (generally referred to as "specialized agencies" composed of experts rather than politicians) to which to delegate (subject to continuous review) the task of implementing those agreed purposes. It has sometimes been easier to obtain agreement among the nations involved on standards and methods concerning the civilian (commercial) aspects of a given activity than to obtain agreement on the military (governmental) aspects of the same activity.24 For example, civil aviation is regulated internationally through agencies that have promulgated numerous agreements and regulations, all by consensus. Over the years, some precedents, and some forms of regulation, have been established, again largely by consensus, that have enhanced the protection of civilian aviation and reduced the uncertainties regarding governmental (military) aviation. A similar pattern of international regulation has resulted in increased maritime safety. 23 Chapter 10, NRC, Technology, Policy, Law, and Ethics Regarding U.S. Acquisition and Use of Cyberattack Capabilities, 2009. 24 Abraham D. Sofaer and Seymour E. Goodman, A Proposal for an International Convention on Cyber Crime and Terrorism, Center for International Security and Cooperation, Stanford University, August 2000.
OCR for page 25
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy In both areas, states have agreed to criminalize terrorist attacks, and to prosecute or extradite violators. These commitments have not uniformly been kept, but security has been enhanced in these areas of international commerce because of the virtually universal support given to protecting these activities from identified threats. It is an open question whether such an approach might enhance cybersecurity internationally, whether or not it excludes any direct application or restriction on the national security activities of signatories. 2.4 DOMESTIC REGIMES TO PROMOTE CYBERSECURITY Law enforcement regimes to prosecute cyber criminals are not the only ones possible to help promote cybersecurity. As noted in Toward a Safer and More Secure Cyberspace, the nation’s cybersecurity posture would be significantly enhanced if all owners and operators of computer systems and networks took actions that are already known to improve cybersecurity. That is, the nation needs to do things that the nation already knows how to do. What that report identified as a critical problem in cybersecurity was a failure of action. That report attributed the lack of adequate action to two factors—the fact that decision makers discount future possibilities of disaster so much that they do not see the need for present-day action (that is, they weigh the immediate costs of putting into place adequate cybersecurity measures, both technical and procedural, against the potential future benefits (actually, avoided costs) of preventing cyber disaster in the future—and systematically discount the latter as uncertain and vague) and the additional fact that the costs of inaction are not borne by the relevant decision makers (that is, the nation as a whole bears the cost of inaction, whereas the cost of action is borne by the owners and operators of critical infrastructure, which are largely private-sector companies). Accordingly, that report called for changes in the decision-making calculus that at present excessively focuses vendor and end-user attention on the short-term costs of improving their cybersecurity postures. The report did not specify the nature of the necessary changes, but rather noted the need for more research in this area to assess the pros and cons of any given change. The present report reiterates the importance of changing the decision-making calculus described above, but suggests that developing the necessary domestic regime (including possibly law, regulation, education, culture, and norms) to support a new calculus will demand considerable research. 3. A Possible Research Agenda Although the preceding section seeks to describe some of the essential elements of cyberdeterrence, it is sobering to realize the enormity of intellectually unexplored territory associated with such a basic concept. Thus, the committee believes that considerable work needs to be done to explore the relevance and applicability of deterrence and prevention/inhibition to cyber conflict. At the highest level of abstraction, the central issue of interest is to identify what combinations of posture,
OCR for page 26
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy policies, and agreements might help to prevent various actors (including state actors, nonstate actors, and organized criminals) from conducting cyberattacks that have a disabling or a crippling effect on critical societal functions on a national scale (e.g., military mission readiness, air traffic control, financial services, provision of electric power). The broad themes described below (lettered A-H) are intended to constitute a broad forward-looking research agenda on cyberdeterrence. Within each theme are a number of elaborating questions that are illustrative of those that the committee believes would benefit from greater exploration and analysis. Thoughtful research and analysis in these areas would contribute significantly to understanding the nature of cyberdeterrence. A. Theoretical Models for Cyberdeterrence Is there a model that might appropriately describe the strategies of state actors acting in an adversarial manner in cyberspace? Is there an equilibrium state that does not result in cyber conflict? How will any such deterrence strategy be affected by mercenary cyber armies for hire and/or patriotic hackers? How does massive reciprocal uncertainty about the offensive cyberattack capabilities of the different actors affect the prospect of effective deterrence? How might adversaries react technologically and doctrinally to actual and anticipated U.S. policy decisions intended to strengthen cyberdeterrence? What are the strengths and limitations of applying traditional deterrence theory to cyber conflict? What lessons and strategic concepts from nuclear deterrence are applicable and relevant to cyberdeterrence? How could mechanisms such as mutual dependencies (e.g., attacks that cause actual harm to the attacker as well as to the attacked) and counterproductivity (e.g., attacks that have negative political consequences against the attacker) be used to strengthen deterrence? How might a comprehensive deterrence strategy balance the use of these mechanisms with the use of traditional mechanisms such as retaliation and passive defense? B. Cyberdeterrence and Declaratory Policy What should be the content of a declaratory policy regarding cyberintrusions (that is, cyberattacks and cyberintrusions) conducted against the United States? Regarding cyberintrusions conducted by the
OCR for page 27
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy United States? What are the advantages and disadvantages of having an explicit declaratory policy? What purposes would a declaratory policy serve? What longer-term ramifications accompany the status quo of strategic ambiguity and lack of declaratory policy? What is the appropriate balance between publicizing U.S. efforts to develop cyber capabilities in order to discourage/deter attackers and keeping them secret in order to make it harder for others to foil them? What is the minimum amount and type of knowledge that must be made publicly available regarding U.S. government cyberattack capabilities for any deterrence policy to be effective? To the extent that a declaratory policy states what the United States will not do, what offensive operational capabilities should the United States be willing to give up in order to secure international cooperation? How and to what extent, if at all, does the answer vary by potential target (e.g., large nation-state, small nation-state, subnational group, and so on)? What declaratory policy might help manage perceptions and effectively deter cyberattack? C. Operational Considerations in Cyberdeterrence On what basis can a government determine whether a given unfriendly cyber action is an attack or an exploitation? What is the significance of mistaking an attack for an exploitation or vice versa? How can uncertainty and limited information about an attacker’s identity (i.e., attribution), and about the scope and nature of the attack, be managed to permit policy makers to act appropriately in the event of a national crisis? How can overconfidence or excessive needs for certainty be avoided during a cyber crisis? How and to what extent, if at all, should clear declaratory thresholds be established to delineate the seriousness of a cyberattack? What are the advantages and disadvantages of such clear thresholds? What are the tradeoffs in the efficacy of deterrence if the victim of an attack takes significant time to measure the damage, consult, review options, and most importantly to increase the confidence that attribution of the responsible party is performed correctly? How might international interdependencies affect the willingness of nations to conduct certain kinds of cyberattack on other nations? How can
OCR for page 28
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy blowback be exploited as an explicit and deliberate component of a cyberdeterrence strategy? How can the relevant feedback loops be made obvious to a potential attacker? What considerations determine the appropriate mode(s) of response (cyber, political, economic, traditional military) to any given cyberattack that calls for a response? How should an ostensibly neutral nation be treated if cyberattacks emanate from its territory and that nation is unable or unwilling to stop those attacks? Numerous cyberattacks on the United States and its allies have already occurred, most at a relatively low level of significance. To what extent has the lack of a public offensive response undermined the credibility of any future U.S. deterrence policy regarding cyberattack? How might credibility be enhanced? How and to what extent, if at all, must the United States be willing to make public its evidence regarding the identity of a cyberattacker if it chooses to respond aggressively? What is the appropriate level of government to make decisions regarding the execution of any particular declaratory or operational policy regarding cyberdeterrence? How, if at all, should this level change depending on the nature of the decision involved? How might cyber operations and capabilities contribute to national military operations at the strategic and tactical levels, particularly in conjunction with other capabilities (e.g., cyberattacks aimed at disabling an opponent’s defensive systems might be part of a larger operation), and how might offensive cyber capabilities contribute to the deterrence of conflict more generally? How should operational policy regarding cyberattack be structured to ensure compliance with the laws of armed conflict? How might possible international interdependencies be highlighted and made apparent to potential nation-state attackers? What can be learned from case studies of the operational history of previous cyberintrusions? What are the lessons learned for future conflicts and crises? Technical limitations on attribution are often thought to be the central impediment in holding hostile cyber actors accountable for their actions. How and to what extent would a technology infrastructure designed to
OCR for page 29
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy support high-confidence attribution contribute to the deterrence of cyberattack and cyber exploitation, make the success of such operations less likely, lower the severity of the impact of an attack or exploitation, and ease reconstitution and recover after an attack? What are the technical and nontechnical barriers to attributing cyberintrusions? How might these barriers be overcome or addressed in the future? D. Regimes of Reciprocal/Consensual Limitations What regimes of mutual self-restraint might help to establish cyberdeterrence (where regimes are understood to include bilateral or multilateral hard-law treaties, soft-law mechanisms [agreements short of treaty status that do not require ratification], and international organizations such as the International Telecommunication Union, the United Nations, the Internet Engineering Task Force, the Internet Corporation for Assigned Names and Numbers, and so on)? Given the difficulty of ascertaining the intent of a given cyber action (e.g., attack or exploitation) and the scope and extent of any given actor’s cyber capabilities, what is the role of verification in any such regime? What sort of verification measures are possible where agreements regarding cyberattack are concerned? What sort of international norms of behavior might be established among like-minded nations collectively that can help establish cyberdeterrence? What sort of self-restraint might the United States have to commit to in order to elicit self-restraint from others? What might be the impact of such self-restraint on U.S. strategies for cyber conflict? How can a “cyberattack taboo” be developed (perhaps analogous to taboos against the use of biological or nuclear weapons)? How and to what extent, if any, can the potency of passive defense be meaningfully enhanced by establishing supportive agreements and operating norms? How might confidence-building and stability measures (analogous to hotline communications in possible nuclear conflict) contribute to lowering the probability of crises leading to actual conflict? How might agreements regarding nonmilitary dimensions of cyberintrusion support national security goals? How and to what extent, if at all, should the United States be willing to declare some aspects of cyberintrusion off limits to itself? What are the tradeoffs involved in foreswearing offensive operations, either unilaterally or as part of a multilateral (or bilateral) regime?
OCR for page 30
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy What is an act of war in cyberspace? Under what circumstances can or should a cyberattack be regarded as an act of war.25 How and to what extent do unique aspects of the cyber realm, such as reversibility of damage done during an attack and the difficulty of attribution, affect this understanding? How and to what extent, if any, does the Convention on Cyber Crime (http://conventions.coe.int/Treaty/EN/Treaties/html/185.htm) provide a model or a foundation for reaching further international agreements that would help to establish cyberdeterrence? How might international and national law best address the issue of patriotic hackers or cyber patriots (or even private sector entities that would like to respond to cyberattacks with cyber exploitations and/or cyberattacks of their own), recognizing that the actions of such parties may greatly complicate the efforts of governments to manage cyber conflict? E. Cyberdeterrence in a Larger Context How and to what extent, if at all, is an effective international legal regime for dealing with cyber crime a necessary component of a cyberdeterrence strategy? How and to what extent, if at all, is deterrence applicable to cyberattacks on private companies (especially those that manage U.S. critical infrastructure)? How should a U.S. cyberdeterrence strategy relate to broader U.S. national security interests and strategy? F. The Dynamics of Action/Reaction What is the likely impact of U.S. actions and policy regarding the acquisition and use of its own cyberattack capabilities on the courses of action of potential adversaries? How and to what extent, if at all, do efforts to mobilize the United States to adopt a stronger cyberdefensive posture prompt potential adversaries to believe that cyberattack against the United States is a viable and effective means of causing damage? 25 The term “act of war” is a colloquial term that does not have a precise international legal definition. The relevant terms from the UN Charter are “use of force,” “threat of force,” and “armed attack,” although it must be recognized that there are no internationally agreed-upon formal definitions for these terms either.
OCR for page 31
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy G. Escalation Dynamics How might conflict in cyberspace escalate from an initial attack? Once cyber conflict has broken out, how can further escalation be deterred? What is the relationship between the onset of cyber conflict and the onset of kinetic conflict? How and under what circumstances might cyberdeterrence contribute, if at all, to the deterrence of kinetic conflict? What safeguards can be constructed against catalytic cyberattack? Can the United States help others with such safeguards? H. Collateral Issues How and to what extent do economics and law (and regulation) affect efforts to enhance cybersecurity in the private sector? What are the pros and cons of possible solution elements that may involve (among other things) regulation, liability, and standards-setting that could help to change the existing calculus regarding investment strategies and approaches to improve cybersecurity? Analogies from other “protection of the commons” problem domains (e.g., environmental protection) may be helpful. What are the civil liberties implications (e.g., for privacy and free expression) of policy and technical changes aimed at preventing cyberattacks, such as systems of stronger identity management for critical infrastructure? What are the tradeoffs from a U.S. perspective? How would other countries see these tradeoffs? How can the development and execution of a cyberdeterrence policy be coordinated across every element of the executive branch and with Congress? How should the U.S. government be organized to respond to cyber threats? What organizational or procedural changes should be considered, if any? What roles should the new DOD Cyber Command play? How will the DOD and the intelligence community work together in accordance with existing authorities? What new authorities would be needed for effective cooperation? How and to what extent, if any, do private entities (e.g., organized crime, terrorist groups) with significant cyberintrusion capabilities affect any government policy regarding cyberdeterrence? Private entities acting outside government control and private entities acting with at least tacit government approval or support should both be considered. How and to what extent are current legal authorities to conduct cyber operations (attack and exploitation) confused and uncertain? What standards should govern whether or not a given cyber operation takes
OCR for page 32
Letter Report from the Committee on Deterring Cyberattacks: Informing Strategies and Developing Options for U.S. Policy place? How does today’s uncertainty about authority affect the nation’s ability to execute any given policy on cyberdeterrence? Cyberattack can be used as a tool for offensive and defensive purposes. How should cyberattacks intended for defensive purposes (e.g., conducted as part of an active defense to neutralize an incoming attack) differ from those intended for offensive purposes (e.g., a strategic cyberattack against the critical infrastructure of an adversary)? What guidelines should structure the former as opposed to the latter? Research contributions in these areas will have greater value if they can provide concrete analyses of the offensive actors (states, criminal organizations, patriotic hackers, terrorists, and so on), motivations (national security, financial, terrorism), actor capacities and resources, and which targets require protection beyond that afforded by passive defenses and law enforcement (e.g., military and intelligence assets, critical infrastructure, and so on). 4. Conclusion The research agenda described in the questions above is intellectually challenging and fundamentally interdisciplinary. The committee hopes that a variety of scholarly communities, including those in political science, psychology, and computer science and information technology, are able to find ways of working together to address the very important question of deterring cyberattacks against the societal interests of the United States. Moving forward and in accordance with the requirements of the relevant contract, the committee has commissioned a number of papers that address some of the questions articulated above. Drafts of these papers will be discussed in a workshop to be held in June 2010. Although resource limitations will constrain the number of papers commissioned, the committee is of the belief that all of these questions are important and deserve further significant attention. Respectfully, John D. Steinbruner, Chair Committee on Deterring Cyberattacks Computer Science and Telecommunications Board Division on Engineering and Physical Sciences Division on Policy and Global Affairs