Wednesday, November 25, 2009

Review of Martin Libicki's Cyberdeterrence and Cyberwar

Amazon.com just posted my three star review of Martin Libicki's Cyberdeterrence and Cyberwar. I've reproduced the review in its entirety here because I believe it is important to spread the word to any policy maker who might read this blog or be directed here. I've emphasized a few points for readability.

As background, I am a former Air Force captain who led the intrusion detection operation in the AFCERT before applying those same skills to private industry, the government, and other sectors. I am currently responsible for detection and response at a Fortune 5 company and I train others with hands-on labs as a Black Hat instructor. I also earned a master's degree in public policy from Harvard after graduating from the Air Force Academy.

Martin Libicki's Cyberdeterrence and Cyberwar (CAC) is a weighty discussion of the policy considerations of digital defense and attack. He is clearly conversant in non-cyber national security history and policy, and that knowledge is likely to benefit readers unfamiliar with Cold War era concepts. Unfortunately, Libicki's lack of operational security experience undermines his argument and conclusions. The danger for Air Force leaders and those interested in policy is that they will not recognize that, in many cases, Libicki does not understand what he is discussing. I will apply lessons from direct experience with digital security to argue that Libicki's framing of the "cyberdeterrence" problem is misguided at best and dangerous at worst.

Libicki's argument suffers five key flaws. First, in the Summary Libicki states "cyberattacks are possible only because systems have flaws" (p xiii). He continues with "there is, in the end, no forced entry in cyberspace... It is only a modest exaggeration to say that organizations are vulnerable to cyberattack only to the extent they want to be. In no other domain of warfare can such a statement be made" (p. xiv). I suppose, then, that there is "no forced entry" when a soldier destroys a door with a rocket, because the owners of the building are vulnerable "to the extent they want to be"? Are aircraft carriers similarly vulnerable to hypersonic cruise missiles because "they want to be"? How about the human body vs bullets?

Second, Libicki's fatal understanding of digital vulnerability is compounded by his ignorance of the role of vendors and service providers in the security equation. Asset owners can do everything in their power to defend their resources, but if an application or implementation has a flaw it's likely only the vendor or service provider who can fix it. Libicki frequently refers to sys admins as if they have mystical powers to completely understand and protect their environments. In reality, sys admins are generally concerned about availability alone, since they are often outsourced to the lowest bidder and contract-focused, or understaffed to do anything more than keep the lights on.

Third, this "blame the victim" mentality is compounded by the completely misguided notions that defense is easy and recovery from intrusion is simple. On p 144 he says "much of what militaries can do to minimize damage from a cyberattack can be done in days or weeks and with few resources." On p 134 he says that, following cyberattack, "systems can be set straight painlessly." Libicki has clearly never worked in a security or IT shop at any level. He also doesn't appreciate how much the military relies on civilian infrastructure from everything to logistics to basic needs like electricity. For example, on p 160 he says "Militaries generally do not have customers; thus, their systems have little need to be connected to the public to accomplish core functions (even if external connections are important in ways not always appreciated)." That is plainly wrong when one realizes that "the public" includes contractors who design, build, and run key military capabilities.

Fourth, he makes a false distinction between "core" and "peripheral" systems, with the former controlled by users and the later by sys admins. He says "it is hard to compromise the core in the same precise way twice, but the periphery is always at risk" (p 20). Libicki is apparently unaware that one core Internet resource, BGP, is basically at constant risk of complete disruption. Other core resources, DNS and SSL, have been incredibly abused during the last few years. All of these are known problems that are repeatedly exploited, despite knowledge of their weaknesses. Furthermore, Libicki doesn't realize that so-called critical systems are often more fragile that user systems. In the real world, critical systems often lack change management windows, or are heavily regulated, or are simply old and not well maintained. What's easier to reconfigure, patch, or replace, a "core" system that absolutely cannot be disrupted "for business needs," or a "peripheral" system that belongs to a desk worker?

Fifth, in addition to not understanding defense, Libicki doesn't understand offense. He has no idea how intruders think or the skills they bring to the arena. On pp 35-6 he says "If sufficient expenditures are made and pains are taken to secure critical networks (e.g., making it impossible to alter operating parameters of electric distribution networks from the outside), not even the most clever hacker could break into such a system. Such a development is not impossible." Yes, it is impossible. Thirty years of computer security history have shown it to be impossible. One reason why he doesn't understand intruders appears on p 47 where he says "private hackers are more likely to use techniques that have been circulating throughout the hacker community. While it is not impossible that they have managed to generate a novel exploit to take advantage of a hitherto unknown vulnerability, they are unlikely to have more than one." This baffling statement shows Libicki doesn't appreciate the skill set of the underground.

Libicki concludes on pp xiv and xix-xx "Operational cyberwar has an important niche role, but only that... The United States and, by extension, the U.S. Air Force, should not make strategic cyberwar a priority investment area... cyberdefense remains the Air Force's most important activity within cyberspace." He also claims it is not possible to "disarm" cyberwarriors, e.g., on p 119 "one objective that cyberwar cannot have is to disarm, much less destroy, the enemy. In the absence of physical combat, cyberwar cannot lead to the occupation of territory." This focus on defense and avoiding offense is dangerous. It may not be possible to disable a country's potential for cyberwar, but an adversary can certainly target, disrupt, and even destroy cyberwarriors. Elite cyberwarriors could be likened to nuclear scientists in this respect; take out the scientists and the whole program suffers.

Furthermore, by avoiding offense, Libicki makes a critical mistake: if cyberwar has only a "niche role," how is a state supposed to protect itself from cyberwar? In Libicki's world, defense is cheap and easy. In the real world, the best defense is 1) informed by offense, and 2) coordinated with offensive actions to target and disrupt adversary offensive activity. Libicki also focuses far too much on cyberwar in isolation, while real-world cyberwar has historically accompanied kinetic actions.

Of course, like any good consultant, Libicki leaves himself an out on p 177 by stating "cyberweapons come relatively cheap. Because a devastating cyberattack may facilitate or amplify physical operations and because an operational cyberwar capability is relatively inexpensive (especially if the Air Force can leverage investments in CNE), an offensive cyberwar capability is worth developing." The danger of this misguided tract is that policy makers will be swayed by Libicki's misinformed assumptions, arguments, and conclusions, and believe that defense alone is a sufficient focus for 21st century digital security. In reality, a kinetically weaker opponent can leverage a cyber attack to weaken a kinetically superior yet net-centric adversary. History shows, in all theatres, that defense does not win wars, and that the best defense is a good offense.

9 comments:

DeLepster said...

With such a devastating review, how come he still gets three stars?

Richard Bejtlich said...

Just because his argument is flawed doesn't mean it's not worth reading. :)

Bryon said...

I think your argument against "no forced entry" is flawed. A soldier knocking down a door is inherently different than anything in a 'cyber' realm. This kind of 'real life' allusion/metaphor does not translate or equate to how computer systems work and really just confuses the issue.

I think all Libicki is trying to say is design flaws are what allow system compromises.

Now, countering his argument by saying that design flaws are provably never going to be completely removed (halting problem), there you might have something...

Richard Bejtlich said...

Bryon, I don't agree. Is the fact that you can destroy a door a "design flaw" in the door? What if you create an app and take every precaution you can given a certain threat model, and a new threat model appears that renders the old one useless?

The point is any attack is "forced entry". The word "attack" implies force by itself.

Bryon said...

Your threat model analysis is an accurate critique of the impossible nature of actually trying to build a functional and secure system that operates under Libicki's model (and I agree). It does not however show how Libicki's model is incorrect or flawed.

The word 'attack' when considered in a 'cyber' world needs to have its physical connotations set aside. But again I agree that Libicki should refrain from using the word and choose 'exploit' or something similar.

only.Samurai said...

I haven't read all Libicki's work, so I may have missed this. But his claim that systems are only as flawed as their technical components completely disregards the less-technical methods an attack may employ. Social engineering and phishing come to mind. I've worked penetration tests before where we were unable to compromise the systems directly, but were able to gain access through exploiting an employee. A cyberattacker could use similar methods to gather valid credentials, at which point the good majority of defensive technology is useless.

6p01157148a11a970c said...

It's hard to take Libricki's analysis seriously when his underlying knowledge of cyber security is so weak.

For example, on page 143 he says: "Cyberattacks are about deception, and the essence of deception is the difference between what you expect and what you get: surprise. This is why operational cyberwar is tailor-made for surprise attack and a poor choice for repeated attacks: It is difficult to surprise the same sysadmin twice in the same way."

Why would one assume that an attacker would have only one method of attack and use it repeatedly? The breadth of attack vectors is huge and has the same sort of asymmetry as terrorist attacks. At this point I think it's accepted wisdom that anti-terrorism must be proactive and have major offensive components.

6p01157148a11a970c said...

I just realized I left out the last sentence of the comment I posted at 6:50pm. Here it is:

By analogy, the asymmetric nature of cyber attacks leads one to conclude that offensive capabilities are needed, although I am not saying that it's the sole justification for developing offensive cyber war capabilities.

Richard Bejtlich provided several other arguments in his post.

Also I noticed I spelled the author's name wrong. It should be Libicki.

Stiennon said...

Richard: I usually find myself in agreement with your analysis so it surprised me to see you come down on Libiki so hard. I read his book cover to cover and found it to contain cogent thinking on the issue of deterrence, and defense versus offense. Thinking that is much better than what has been produced of late by the hand wringers and chicken littles that have been crying cybergeddon! and cyberkatrina!

Libiki shines a light on the deterrence argument that needed shining. There appears to be a mad scramble for dollars from the various branches of the military as they reach for new money for their pet projects. But why the sudden interest in developing offensive capability when the US military has demonstrated that it is having trouble protecting critical systems from attackers. The Pentagon, NIPRNet, Sandia, and others have all suffered major breaches.

I am still in the camp that says the best cyber defense is a good cyber defense. Until the US can demonstrate proficiency in defense I will not expect them to effectively manage offensive cyber war capabilities.