[Full-disclosure] Getting Off the Patch
Thor (Hammer of God)
thor at hammerofgod.com
Fri Jan 14 19:08:03 GMT 2011
>From: Pete Herzog [mailto:lists at isecom.org]
>Sent: Friday, January 14, 2011 10:19 AM
>To: Thor (Hammer of God)
>Cc: Valdis.Kletnieks at vt.edu; phocean; full-disclosure at lists.grok.org.uk; Zach
>Subject: Re: [Full-disclosure] Getting Off the Patch
>> It's brilliant! Where do I sign up?
>What you run a patch management company? What's your problem with
>trying to improve the way we do things? If we find patching isn't a good nor
>necessary solution for better security then why shouldn't we propose a new
No, I do not run a patch management company, but despite that, I successful patch on an ongoing basis without experiencing any of your claimed wastes of money, time, and resources. And within the context of this conversation, since you are the one saying that you don't have to patch, it should be you that illustrates a level of patch management expertise
Coming up with some way of creating a dependency on new, additional security in depth requirements that on their own create additional administration in order to consciously stop patching is ridiculous Pete. If your controls are good enough to obviate the need for patching, then they should ALREADY BE in place, and part of the model which includes patching. This is why you are seeing the "wtf is new or different about this?" posts.
>Maybe you misunderstood this? If you need empirical evidence that patches
>change code then please do a diff yourself between two apps, one patched
>and one not. Here I was writing of the cost of functional testing and
>remediation of the operational security which scales exponentially as the
>operations scale. One doesn't need a server farm to prove as more servers
>are introduced into an operation that the number of connections between
>them grows. 2 servers each with 1 connection has 2. Add 2 more servers and
>now you have 4 servers but 8 connections to verify. And it goes on like that. If
>you don't do any testing and don't care then you don't have that work or
>money to lose with patching. But I said that already.
The fact that patching changes code is a point so obvious that it doesn't need to made. What I asked for is empirical supporting your claim that your Get Off The Patch model actually saves time and money, while ensuring that your security is strong enough so that you can decide purposefully not to patch. Having a server farm to perform an ongoing cost analysis of the two models is absolutely required if you are going to present this idea to even the most basic of management personnel.
When you go to management with a paradigm shift that will require clearance from legal, policy, engineering and development teams, you will have to show them a clear and unambiguous reduction in costs and risks that will justify the organization assuming the overall risk of not patching. When you make claims such as "patching is a waste of money" and that it causes costs to spiral exponentially, you are going to have to show that. I submit in this case that you can't provide that because you don't have it, and haven't done it. If the patching process truly is a budget-sucking, workflow blocking, administrative nightmare as you state, then the evidence of that fact should be trivial to illustrate. And nowhere in the model do you address the costs of the new model. You said, and I quote (which I probably don't have to say since I am actually using quotes), "We find that that the right balance of operational controls at each interactive point within a vector can provide protection against 100% of the threats including unknown threats." How did your "we" find that? You found it HOW? This statement clearly states that YOU HAVE DONE THIS, but I'm confused as to why you would then respond with "I don't need a server farm to prove this." You are stating that you have found a way to protect against 100% of threats, including unknown threats. That statement alone wins you a spot on "The Wackiest Things Said on an ITSec List Show" but it also illustrates that you completely miss the point about illustrating risk. Qualifying threats does no good if you have not quantified risks.
How exactly is this going to be presented to management? "Hey, the million dollars we spent to whack the servers with a rubber chicken to scare away the vulnerabilities has been a complete waste of money, and though that was our idea in the first place, we now have a new idea where we are going to do "other things" and ignore the vulnerabilities altogether. By doing so, we will take care of 100% of all threats known and unknown. However, we don't know how much that will cost, how much it will save, or what we will do for jobs once we set it up so that 100% of all threats known and unknown are protected against."
How is anyone supposed to actually consider this when you have no data to show that it works since you've not actually done it in production environment of consequence? Is the expectation that management is going to OK this (as well as legal, engineering, etc) when you've already illustrated that you can't manage patches in the first place without costs spiraling out of control?
I know this is all a harsh response, but your continued dialog on the subject illustrates that you seem to actually believe this is viable in the absence of any examples of it working.
Full-Disclosure is hosted and sponsored by Secunia.