Time for new network security certification
Since GIAC messed up
Comment I wrote a column for Securityfocus some time ago that aired my concerns over GIAC dropping the practical portion of their certification process. That column resulted in a lot of feedback, with most agreeing about how GIAC bungled what was up till then, the best certification around.
GIAC was until then known for a stringent practical paper that had to be written prior to being authorized to write two online exams to complete the certification process. Well since the great hubbub that GIAC caused over dropping the practical, it has since been reinstated by Stephen Northcutt.
That said, you now have the option of doing it after you have written the online exams to achieve what is now called the Gold standard. Writing only the online exams will give you the Silver standard. It still does not have the same impact as requiring the practical be done first. It was via the practical that you showed prospective employers you could actually perform the actions. The vast amount of people will only write the exams in order to get the GIAC cert. What was accomplished by having the practical section placed after the exams was to lower the bar for people desiring a GIAC certificate. Not many people realize the difference between Silver and Gold. So with that in mind, why bother going for the Gold via a practical paper? Many managers and HR people do not know the difference.
Why is having a practical portion to a network security certification so important? Some excellent examples are seen in the archives for the pen-test list that Securityfocus hosts. There are posts there from people asking for tips on how to pen-test some pretty basic systems for one of their client engagements. I cringe to think of the poor customer who contracted out a pen-test of their network to a person who needs to ask for advice on how to do it, via a public mailing list. It really does boggle the mind. Had these people actually written a rigorous certification exam they might not be on the mailing list asking for help in pen-testing a client network.
So far I have only detailed what I perceive as the problem with network security certifications, and not yet given what I believe is the solution to it. The solution that I will detail shortly was borne out of several conversations I had with a friend who is also in the business of network security as a consultant and trainer, Mike Sues. We both came to the conclusion that it really isn't good enough to simply churn out endless certifications based on multiple choice questions. This proves little beyond your memorization skills. Classic examples of this were the horror stories of some years ago about the MCSE boot camps. Here you would have people certified as MCSE's after a week, with little prior experience, yet they couldn't install a printer driver. This is a perfect example of good memory skills in place of understanding the underlying theory.
This lack of understanding when it comes to network security theory, in its many forms, is crucial to why we have so many network security practitioners out there today who really should be unemployed. Network security theory comes in many forms as mentioned. It will range from understanding TCP/IP, to programming and scripting concepts, to network architecture and beyond. Like many other fields of study such as engineering, network security is composed of a huge body of knowledge. There are very few, if any, experts who know it all. Therefore, it only makes sense to understand the theory behind these various areas. It is only by understanding the theory that you truly understand something. An example of this is why it is good to deny inbound TCP Port 53 on your firewall. Regurgitating something that you heard on a course or in an IRC chat room isn't good enough. You would only know why the above example is a good firewall policy by having a passing knowledge of the DNS protocol.
So what should the cert be based on then?
There are presently a large number of certifications out there. The bulk of these are vendor specific while some others are theme specific, such as pen-test certifications. To answer what an ideal certification should be, we need to ask ourselves just what the ideal network security analyst should know. That body of knowledge would encompass the areas I listed above plus others. While I listed programming and scripting concepts, it is not absolutely necessary to be able to actually program or script - although it is extremely helpful. Of the two, scripting is becoming more and more important to know for the security analyst. Where knowledge of programming and scripting comes in, is when an analyst needs to do cursory reverse engineering of a piece of malware, or bang out a quick script. When doing such a task it is important to understand and recognize programming or scripting functions for what they are. What an analyst should really know though would be several articles in and of itself. I have written articles in the past on these topics, which are available here. Those articles give a good overview of the skills required to be successful in the network security field.
As I chronicled in those articles there are a lot of skills that an analyst requires. More importantly, the analyst also needs to understand the underlying theory behind those skills. This is even more true in today's network security environment. It would be extremely unlikely for any of us to remain with the same company for our entire career. Therefore, it is rather likely that we will work on other corporate networks, and odds are that we will then be using other vendor's products. This is once again where one needs a certification that will test an analyst's knowledge of theory, vice specific vendor product knowledge. After all, a firewall GUI is a firewall GUI. Firewalls all work in the same fashion, the vendor specific window dressing of their GUI is really rather immaterial.
So it appears then that one should test an analyst on theory in both a practical and realistic environment. This is not as unrealistic as it sounds. All of this can be done using open source programs. Yes, that means Linux or BSD. In today's competitive job market it is extremely unlikely that a network security analyst does not have at least a passing familiarity with either Linux or BSD, and it is assumed that they know Microsoft Windows. All of the programs needed to then test an analyst on their knowledge can be had via open source offerings such as Snort, IP Tables, and BIND to name a few.
The certification itself would then be scenario based in a computer lab. It would encompass a simulated network security incident that would then test the person across various bodies of knowledge. During the testing the student would also be asked to explain why it is that they are taking specific actions. They are then being tested for not only the right answer, but also why it is that they did it. That would demonstrate their understanding of the underlying theory. If the student can explain their reasons for a specific course of action, then learning how to use vendor offering of IDS ABC would be trivial. This certification framework would not only be restricted to IDS's and firewalls, it would also encompass testing people for knowledge of reverse engineering, security and application architecture, conducting Threat Risk Assessments (TRA) and more. The whole driving force of this new certification would be to confirm that the person has a grasp of network security fundamentals, as well as other specialized areas. Should they possess this critical mass of knowledge then there is very little that they would not understand. That is a certification that would have true value.
This article originally appeared in Security Focus.
Copyright © 2007, SecurityFocus
Don Parker, GCIA GCIH, specializes in intrusion detection and incident handling. In addition to writing about network security he enjoys a role as guest speaker for various security conferences.