Customer Reviews

Security Metrics: Replacing Fear, Uncertainty, and Doubt

5 star:		(14)
4 star:		(4)
3 star:		(3)
2 star:		(0)
1 star:		(1)

 

 

I read Security Metrics right after finishing Managing Cybersecurity Resources, a book by economists arguing that security decisions should be made using cost-benefit analysis. On the face of it, cost-benefit analysis makes perfect sense, especially given the authors' analysis. However, Security Metrics author Andy Jaquith quickly demolishes that approach (confirming the problem I had with the MCR plan). While attacking the implementation (but not the idea) of Annual Loss Expectancy for security events, Jaquith writes on p 33 "[P]ractitioners of ALE suffer from a near-complete inability to reliably estimate probabilities [of occurrence] or losses." Bingo, game over for ALE and cost-benefit analysis. It turns out the reason security managers "herd" (as mentioned in MCR) is that they have no clue what else to do; they seek safety in numbers by emulating peers and then claim that as a defense when they are breached.

Fortunately, Security Metrics offers another solution. The book gives readers three sets of information: theory, metrics, and tools (concepts, not programs). The theory chapters (1 and 2) were so concise yet insightful I was tempted to underline every sentence. (I am not kidding.) Even the Preface made me glad to be reading the book when it associated "security ROI" with "the Macarena" and called it a "needless distraction." I laughed in agreement when I saw Andy call "security enablement" the "Abominable Snowman: it is rarely spotted, but legions of people swear it exists. After all, as my friend Dan geer puts it, 'You don't usually see airlines advertising how their planes fall out of the sky less often than their competitors.'" Why is that? My answer is simple: security is assumed and expected. Advertising anything else has no effect or makes people suspicious. I knew this book would be good.

The metrics chapters probably list hundreds of metrics you can extract verbatim and apply to your own environment. To the reviewer who wanted to reprint them in an appendix: they're called chapters 3 and 4. My main concern with the metrics was the focus on input-centric measurements instead of results. I would have liked to read more metrics on measuring whether security programs are working, rather than what techniques and tools are applied up front.

The tools chapters were helpful to anyone needing a statistics refresher. The visualization sections were especially helpful. (Feel free to dismiss yet another ignorant review from WB, who thinks a "review" means writing a few paragraphs after flipping through the pages of five books a day.) Andy's examples of turning lousy graphs and charts into information visualization vehicles should be followed by all managers.

Security Metrics is strengthened by the many stories from the author's consulting experience. I sensed that his techniques work and are not the product of the thought laboratory alone. I found his "Balanced Scorecard" approach to be interesting, especially to the degree it ties real metrics to business operations.

I had a few issues with terminology, such as using the term "threats" on p 231 when "attacks" is more accurate. (The football analogy is correct, however.) I semi-agreed with the author's suggestion to abandon "risk management" in favor of metrics-based approaches, but I didn't think two pages (4-5) were really enough to make the case. On p 264, threats are not risks, but they help instantiate risks. On pp 78-7, "risk of exploit" should be "ease of exploitation."

These are minor concerns, given the overwhelming concentration of practical and implementation-worthy pieces of information in Security Metrics. You must read this book if you care to measure security progress. Now we need Dan Geer to extend beyond writing wise forewords and articles into the world of his own book!

It's difficult to imbue a book on metrics with something other than academic theories, but Jaquith offers the working security professional a tangible lifeline. Nearly all of his suggested metrics are within easy reach, thanks to a commonsense approach and a tie-in to the instrumentation you're most likely to have in your data center.

Don't be scared off by the term "metrics," either; it's an easy read, chock full of amusing stories and turns of phrase (I thought my 80-year-old father was the only one who said "'pert near"). Jaquith focuses on the practical, from What Not to Draw (a graphics primer for charts and tables) to a Balanced Scorecard Makeover that actually looks achievable from outside the C-suite.

If your boss likes metrics, and your budget request is in jeopardy, you can't do better than this guide to making your case. Now, if only we had a practical, lightweight risk analysis methodology to go along with it ...

The goal of security metrics is to replace fear, uncertainty, and doubt (FUD) with a more formalized and meaningful system of measurement. The FUD factor is the very foundation upon which much of information security is built, and the outcome is decades of meaningless statistics and racks of snake oil products. Let's hope that Andrew Jaquith succeeds, but in doing so, he is getting in the way of many security hardware and software vendors whose revenue streams are built on FUD.

One could write a book on how FUD sells security products. One of the most memorable incidents was in 1992 when John McAfee created widespread panic about the impending Michelangelo virus. The media was all over him as he was selling solutions for the five million PCs worldwide he said would be affected. The end result is that the Michelangelo virus was a non-event. Nonetheless, it was far from the last time that FUD was used to sell security.

The allure of FUD is that companies can spend huge amounts of money fighting nebulous digital adversaries and feel good about it. They can then put all of that fancy hardware in dedicated racks in their data center, impressing the auditors with the flashing lights giving off an aroma of security and compliance.

And that is the chaos that security metrics comes to solve. Security metrics, if done right, can help transform a company from a nebulous perspective on security to an effective one based on formal security risk metrics.

Security Metrics is a fabulous book that should be in the hands of every security professional. The book demonstrates that companies must establish metrics based on their unique requirements, as opposed to simply basing their requirements on imprecise industry polls, best-practices and other ill-defined methods.

So why don't companies do that in the first place? If security metrics can provide even a quarter of the benefits that Jaquith states, companies should run to implement them. Real security metrics require an organization to open up their security hood and dig deep into the engine that runs their security infrastructure. It necessitates understanding the internal requirements, unique organizational risks, myriad strengths and weaknesses, and much more. Very few companies are willing to dedicate the time and resources for that, and would rather build their security infrastructure on thick layers of FUD. History has shown that the security appliance of the month almost always beats a formal risk and needs assessment.

Chapter 1 lays out the problem with approaches that most companies take to risk management. The main problem is that traditional risk management is far too dependant on identification and fixing, as opposed to quantification and triage based on value. Quantifying and valuing risk is much more difficult than simply identifying, since the software tools used do not have an organization context or knowledge of the specific business domain.

Chapter 2 sets out the foundation of security metrics. The goal of these metrics are to provide a framework in which organizations can quantify the likelihood of danger, estimate the extent of possible damage, understand the performance of their security organizations and weigh the costs of security safeguards against their expected effectiveness.

The time has come for security metrics since information security is one of the few management disciplines that have yet to submit itself to serious analytical scrutiny. The various chapters provide many different metrics that can be immediately used in most organizations to address that.

The author defines various criteria for what makes a good metric. One of his pet peeves is the use of the traffic light as a metaphor for compliance. Jaquith feels that traffic lights are not metrics at all, since they don't contain a unit of measure or are a numerical scale. He suggests using traffic lights colors sparingly, and only to supplement numerical data or draw attention to outliers. He astutely notes that if your data contains more precision than three simple gradations, why dilute their value by obscuring them with a traffic light.

The chapter concludes on what makes a bad metric, defined as any metric that relies too much on the judgment of a person. These metrics can't be relied on since the results can't be guaranteed to be the same from person to person. Also, security frameworks such as ISO-17799 should not be used for metrics. The book also tackles the sacred cow of risk management, namely ALE (annualized loss expectancy), and how it is significantly misused and misunderstood in the industry.

The book states that in developing metrics, there must be formal collaboration between the business units and the security staff. This collaboration serves to increase awareness and acceptance of security. In addition, it ensures that security requirements are incorporated into the lifecycle early on. This is needed as business units generally have no clue as to what the needed security requirements are.

Chapter 5 is a short course on analysis techniques and statistics. The author quotes George Colony who stated that "any idiot can tell you what something is. It is much harder to say what that thing means". With that, the book details a number of techniques for analyzing security data (average, median, time series, etc.) and how each one should be used.

Chapter 6 is about visualization and notes that most information security professionals have no real idea how to show security, both literally and figuratively. Part of the problem is that security is proliferated with esoteric terminology and concepts, and the lack of understanding risk management amongst the masses. Part of the reason for this difficulty in sharing the security message with management is that many security practitioners lack simple metaphors for communicating priorities. This is compounded by the fact that the message is often focused exclusively on technical security issues, as opposed to the underlying business issues, which is was management is concerned with. The chapter is invaluable as it weans one off the malevolent pie chart and traffic light PowerPoint presentation.

Marcus Ranum notes that people seem to want to treat computer security like its rocket science or black magic. In fact, computer security is nothing but attention to detail and good design. FUD is all about emphasizing the black magic aspect of hackers and other rogue threats. Metrics are all about the attention to detail that FUD lives to obfuscate.

Security Metrics: Replacing Fear, Uncertainty, and Doubt is one of the more important security books of the last few years. Jaquith turns much of the common security wisdom on its head, and the world will be a better place for it. Security metrics are a necessity whose time has come and this invaluable book shows how it can be done.

I'd really like to give this "3 ½" stars, but I rounded up to 4 stars. There is, currently, no book that is the "last word" on security metrics. The field is just not mature enough for that. However, this is certainly a very good and useful book for most people.

This book is for you if you are a practicing information security professional and you want to know the latest ideas about how to define, deploy, and use security metrics to improve security management. Written in an informal, personal style, Andrew's book reads like "letters from the front lines" (by analogy) than a treatise on military strategy.
The informal style makes the reading, at times, both fun and funny.

He's up front about his preferences and biases, so you know where he's coming from. But he's not bombastic. If you disagree with him on some points (as I do), so be it. His writing invites open debate on the important issues. He's also generous in quoting and crediting various members of the security metrics communities that he participates in.

Andrew falls into the "bag-o-metrics" school of thought, as contrasted from the "risk modeling" school. (This is currently a raging debate within the community.) Basically, Andrew is pessimistic about the possibility of defining any models that integrate security metrics into an overall assessment of business risk. He's especially caustic in his comments about "asset valuation" and other related approaches. Given their current state of development, I don't blame him.

Given this philosophy, Andrew proposes a long list of operational security metrics, each of which measure something very specific (and quantitative), but don't necessarily aggregate. With enough of these "point metrics", some correlations may emerge, he reasons. To help give structure to the "bag", he offers some material relating metrics to various control frameworks (eg. COBIT) and also the Balanced Scorecard. The latter was a noble attempt to fit security metrics into enterprise performance management, but I don't think he really succeeded. But he chews on the right issues and questions, so it help people who are willing to do your own research into corporate scorecards.

Two of most notable parts of the book are the "Introduction: Escaping the Hamster Wheel of Pain", and "Chapter 7 Automating Metrics Calculations". The first is good because he talks in plain, blunt language about the current dismal state of security management in most organizations (who don't effectively use metrics to drive decisions). Ch. 7 is good because it gives a good framework and snapshot in time for a fast-emerging field.

I do have some criticisms of the book, even accepting Andrew's philosophies and premises.

First, there should have been an Appendix that compiled all the suggested metrics into one place. [...]

Second, there is inadequate coverage of how security metrics can mesh with other security and risk management needs - privacy, digital rights, IP protection, forensics, fraud prevention, physical security, and business continuity, to name a few. InfoSec should not be an island. Therefore the metrics system needs to "play well with others", so to speak.

Third, there is inadequate attention to measuring knowledge of attackers and attack strategies. Basically, over the time scale of months or years, InfoSec is an evolutionary strategic "game" between attackers and defenders. This makes information security an arms race, so you need to know if you are falling behind, fighting the last war, etc. Every organization needs to constantly learn about potential attackers and new attack strategies. At least a few metrics in these areas would add big value.

Finally, there is a 49 page chapter devoted "Visualization", which I don't think is the best use of that space. While I think visualization and reporting are *critical*, I think there are plenty of other books and guides that provide guidance. I didn't see anything in this chapter that was specific to InfoSec. That said, the material is useful and valuable if you aren't skilled at visual design.

In conclusion -- a good book that has plenty of useful material, as reported from the front lines.

Andrew Jaquith's book on security metrics is refreshing in its approach. Instead of a neverending cycle of risk assessments and vulnerability patching (a process which the author humorously calls the "hamster wheel of pain"), we are told to focus on core operational security processes and measurement of key indicators.

The central premise of the book is that a "risk management" approach, as promoted by many security vendors, doesn't work. The reason it doesn't work is that it is extremely difficult to get a good handle on the true value of assets, and an accurate estimate of risk. As the author puts it, "identifying problems is easy ... quantifying and valuing risk is much harder."

The thorough discussion of information security metrics makes this book worthwhile reading. However, there is a hint of sloppy thinking sprinkled throughout, which tends to undermine one's trust in the author's intellectual honesty. For example, when discussing the importance of tracking not only inbound viruses, but outbound as well, the author makes the following statement:

BEGIN QUOTE -
Another twist I have added to the traditional antivirus statistics is a simple metric documenting the number of outbound viruses or spyware samples caught by the perimeter mail gateway's content filtering software. Why it matters is simple--it is an excellent indicator of how "clean" the internal network is. Organizations that practice good hygiene don't infect their neighbors and business partners. My friend Dan Geer relates this quote from the CSO of a Wall Street investment bank:

"Last year we stopped 70,000 inbound viruses, but I am prouder of having stopped 500 outbound."

In other words, the bank's internal network is cleaner than the outside environment by a factor of 140 to 1.
- END QUOTE

Certainly, the conclusion in the last sentence cannot be supported without additional information. The volume of inbound email is likely to be drastically higher, which may account for the difference. The bank's outbound detection/prevention mechanism also may not be as efficient as the inbound.

Moreover, the metrics analysis chapter is very rudimentary and incomplete. Basic concepts like mean, median, and standard deviation are briefly discussed, but there is no mention of statistical random sampling techniques and confidence levels, which would surely be of significant importance when measuring key indicators across large populations, where a complete enumeration is either impossible, or too expensive and time-consuming. Sometimes, metrics which are "meaningful", are not the ones that are "tangible" and "easy to measure". A certain degree of statistical sophistication can be helpful in such situations.

In summary, the book offers some useful insight and practical advice for those who are charged with running an information security management program, but a healthy skepticism of the assumptions underlying the author's conclusions is warranted. In order to develop truly meaningful information security metrics, a much more sophisticated approach than what is described in this book will likely be needed.

Security Metrics condenses a lot of useful knowledge on data analysis, visualization, and automated calculation. I've successfully applied Jaquiths's concepts over the past month with fantastic results. The results have helped me illustrate my data in more meaningful ways and also direct my research with more insight. Highly recommended!

Mr. Jaquith's down to earth style and sense of humor make reading this book a very pleasant experience. It is an excellent first book in this field, and I predict it will become a standard. The security field needs this kind of critical insight to promote a higher level of professionalism and service. There are many great ideas in these pages that should be useful whether you're just starting a security program, overhauling an old program, or just want to try something new for your next project. The treatment of mathematics is right on the mark - defensible without being so rigorous it loses people. I'm going to be keeping this book within arm's reach for a long time.

I wrote some kind of pithy quote for this excellent book that ended up on the front cover. If you are over 40 like me, you can almost read it if you get out a magnifying glass.

Instead of thinking up a new pithy quote, I want to keep things simple. Buy this book. Not only is the subject timely, critical, and important, but this has to be one of the best written security books ever. The prose is sparkling and witty, making reading it a joy.

There is plenty of room for more work in security metrics, just as there is plenty of room for more work in my subfield of software security. If we use this work as a guide, we should be in great shape. Thanks Andrew for pointing the way so eloquently.

gem
[...]

Author of "Software Security," "Exploiting Software," and "Building Secure Software"

I'm not sure what I can write to sway you to buy or read the book if 5 star reviews from Ben Rothke and Richard Bejtlich don't sway you but I'll throw my likes and dislikes in here anyway. I'm not a "metrics guy" in fact, I'm still not , but I do think the book puts the concept of using them into perspective for the person that may not use any metrics in their security work.

I've been summing up the book to people at work by using the example (and I'll badly paraphrase) from the book of "if your spam gateway blocks 100,000 spam messages a day is that a good metric?" Initially you may say yes, that is a good metric. In fact most people at work said the same thing. But, as the author explains it is a poor metric. Better metrics are useful percentages like the percentage of missed spam or the percentage of false positives. Saying that 100,000 spam message are being stopped only tells us that you have a ton of spam on your network.

Some of the things I liked about the book were the author's discussions on how to make charts more readable and efficient at portraying information. I had to read the Tufte books in college and have to admit that I got more out of chapter 6 (visualization) than I feel I learned that whole semester of class. Chapter 2 discussing what makes good metrics was extremely useful, as well were chapters 3 & 4 because they gave good examples of metrics you can use to measure an organizations various defenses like perimeter security or application security. The discussion of using COBIT, ITIL and Security Frameworks in Chapter 4 was also good.

I only had two minor gripes. First was that toward the end of the book the author talks about colors of slides and charts which obviously doesn't do us any good since the book is in black and white and second, that he does use some big words throughout the book and I did find myself having to go back and reread things. Could he have put it into simpler terms, probably, but that doesn't make the book bad, just means I need to work on my vocab :-)

Overall it was a good entrance to the world of security metrics for me and took and away some of the perceived boredom of them. It definitely gave me some tools to look more critically at the numbers and stats that some of the vendors throw our way as well as how to deliver data and information in a more useful matter.

There's plenty of other smaller works on metrics, but this book is very much the most comprehensive (and best) work regarding the subject that I've seen. It's really a must have for the library of every security organization and every security analyst.

Consultants would do well to read it and modify their deliverables. Security departments would do well to read it and modify their reporting based on Jaquith's suggestions. Auditors would do well to change their discussions with security based on the contents of the book.

Jaquith's style is light and entertaining - makes reading on the subject that much more fun.

A couple of personal nitpicky issues with the book:

First, I personally don't believe that his strict empiricist approach is a luxury most CISO's have. A subjectivist approach can be valid if adequate rigor is applied (sue me, I'm a Bayesian).

Second, he advocates tying metrics to an ISMS - which is fine and completely appropriate, but limits the usefulness of metrics from a risk management program standpoint. Fortunately, the rest of the metric development content is just as relevant to use outside the "control of controls".

--
However, don't let my small academic issues with the book dissuade you. CISO's should buy several copies and distribute them among their managers and analysts. Buy the book, it will not disappoint.