For those of you who’ve known me or my work for the last decade or more you’ll appreciate that one of my main call to arms is security and in particular enforcement of security enforcing technologies at the gateway and application level, my little hobby (developing publishing and supporting a firewall technology which with variants based on the code) reached millions of homes, offices and enterprises across the globe and allowed me to make a career out of security.
So it’s often a question I get asked at conferences and when speaking about security in Cloud and security enforcement and responsibility in the Cloud and virtualisation arena. Fortunately at Red Hat we take security incredibly seriously and have contributed technologies such as SELinux and sVirt into our architectures and supported versions of our releases, as well as employing the mainstays in the SELinux world on our payroll to ensure that we have continuity and those folk are rewarded for their efforts.
However, to put it bluntly most architects and network guys turn SELinux off when building out platforms and virtualised instances which is quite short sighted. When I do pose the question why a lot of responses are aligned to the fact that SELinux can sometimes due to configuration issues and past experiences where stuff broke and was hard to diagnose so easier to just turn off.
Let’s be blunt, it’s there to help you, it’s a free secure template based technology so turning it off if you haven’t got a full toolkit of other security hardening in your build schema or your platform is at best shortsighted. Did I say it was free ? In this current credit crunch culture can you justify not looking at using it ?
If you’re concerned or you struggle then enable it in permissive mode in the first instance making sure you make relevant mods to /etc/sysconfig/selinux to make it persistent on reboot. Simple boolean logic is the best way (and easiest way) to start experimenting with the functionality you want to add. Then if you want to know more then search for the audit2allow function and remember if you’re concerned with restrictive AVC denials breaking stuff then a quick search through auditd in /var/log/audit/audit.log then aureport is your friend. There are loads of howto’s available or if you’re thinking about large scale SELinux use in anger Red Hat even have a course to upgrade your RHCE to give you a complete comfort blanket in your own capabilities. It’s part of the assurance and certification mode we bring to the whole Linux piece. Belt and braces if you will.
Now this article really isn’t a security masterclass or SELinux howto, I’m actually more interested in getting to grips with culture change and trying to pass on my thoughts of how we need to get traction in influencing how protecting your assets, your data and your reputation in Cloud can take shape.
Over the last three years I’ve been using what I would describe as an almost military approach to building out legacy platforms be they physical or virtual. In days of old people might remember Jay Beale and his Bastille Linux hardening script, which was a great starting point when building simple Linux stacks. I remember vividly when he posted it to newsgroups and Slashdot picked up on it. It represented for the first time really in the Linux Open Source community someone who took a simple exercise but made it mainstream towards security as a standard rather than a retrofit. It enabled many of us to not only run it but get under the hood to find out “how” it worked. What is it they say “a little bit of knowledge is a dangerous thing ?”.
So as we move into provisioning our Cloud environments across one or multiple hypervisor types, or moving applications into hybrid or public Cloud having that “accreditation” process or controls breakdown is invaluable. Mine runs over about 5 tabs of a spreadsheet and would make most auditor feel out of a job. However maybe my way of having a moving spreadsheet of controls that I’ve built up over time for all the certifications / governances that I’ve had to deploy to (including in NATO battlefield accredited above classified environments) probably is going a bit far for standard run of the mill server environments.
So its fortunate that my friends and fellow members of the Cloud Security Alliance started many moons ago to put together an authoritative set of controls to allow you to get to work now building out your platforms or engaging with a Cloud provider regardless of the tenacity or the aggressive nature of your certification or audit model. The controls are designed to get you out the blocks building Cloud platformst that need to meet the regulations around ISO 27001/27002, ISACA COBIT, PCI, NIST, Jericho Forum and NERC CIP. Let’s not mention SAS 70. I still, do not, and believe me I’ve tried, understand why an accounting standard has ANY place in Cloud service provision. CCM will help you there and you can also take a look at the CSA STAR programme while you’re there.
I’ve mentioned the Cloud Security Alliance before here numerous times (lets call them the CSA from now on). The CSA are one of the most critical building blocks of the Cloud community and Jim Reavis and the steering members of the CSA have made the education and communication of security best practices to the community their ethos and commitment since they were founded. Red Hat support the CSA and if you’ve heard me talk you’ll hear me mention them proudly on a regular basis. I am continually mentioning them.
Shortly I am recording an often re-arranged podcast with Jim Reavis of the CSA and we’ll get that out to to you as fast as I can mix it in the coming days and weeks.
Whether you’re playing with Cloud in your dev/test sandpit or migrating to a hybrid cloud understanding what part reputation protection of your app dev environment and your underlying transportation of data is critical. Reputations are lost in minutes as are share prices when a company is seen as damaged by data loss. Simple breaches of major household name organisations are often met with lax fines and investigation by sovereign territory governments and information commissioners, however the risk factors involved are enormous. At the back end of the application architecture – in the trenches – are the technical guys who have to turn the dreams and aspirations of sales people and marketing types into the portals and customer facing Cloud hosted environments that will generate the revenue. If we arm you to do your job better and to do it in a way that allows generic controlled growth of your platforms and your Cloud aspirations then thats a good thing right ?
Do visit the CCM matrixes today and learn how they help you go to work in ways that will make your auditor despair. It’s kinda cool actually because auditing Cloud and typically follow the sun type datacentre clouds has always been a dark art. By following this article and my advice you can actually have a retort to this argument. Cut a huge percentage out your auditors workload (and their resulting invoice) by owning the moral upper ground and in the process maybe think about turning SELinux back on. Blended use of SELinux, sVirt, supported certifed Red Hat subscriptions and technology such as CloudForms gives you everything you need from an IaaS perspective today to go to work. If PaaS security is your thing then listen out soon to another podcast I’m going to record with Tim Kramer of the OpenShift team (in fact if you haven’t already read it go visit Tim’s great security post here).
Also I’m promised a security podcast with Mark Cox at some point in the coming month so if security is your thing you’re going to be kept busy listening to me warble down your earbuds about everything related to CloudSec. If you think that more people could benefit from a primer in Cloud security deployment and the need to think out the box then share this article – I appreciate every Twitter mention I get if it helps educate another Linux user as to how to do things better.
Then get to the CSA website and join. It costs nothing and you’ll learn a lot if you are an active participant. Tell them I sent you.
The Cloud Evangelist 
Hi Richard,
Great article as always. I’d like to add on to your comments about leaving SELinux (and IPtables as well) enabled. When SELinux was first included in Fedora years ago, it was quite easy to build a host that you couldn’t even log into after install if you weren’t careful. When it debuted with RHEL 4, there were only a handful of services that played well with SELinux; RHEL 5 improved upon that, but it was still a black art for many.
Those days are long gone. Folks like Dan Walsh and others have turned it into a security layer that is effective right out of the box. If you need additional configuration, it is prudent to actually learn the steps necessary to create your own policies (or rules in the case of IPtables). There are effective tutorials & books available, in addition to the training that you mentioned.
But perhaps most important, there is support. As in, paid support. There, I said it. I’m all for reducing costs where it makes sense. I’m also in favor of the “physician, heal thyself” Hippocratic approach that many technologists embrace. But no one knows “all of it”. If you have a business or offering that is worth protecting, paying for support shouldn’t be looked down upon or seen as an unnecessary expense.
Jon | @CaptainKVM
Salient words from Jon, Thanks for taking time out to comment