Virtualisation - More or Less Secure?

"Everything is virtual,  Nothing is real!".  Sounds like a songwriters lambast against modern day society.  It's not.  Virtualisation in a computing sense has been around a while and is here to stay.  From the virtualisation of physical machines, applications and network infrastructure, being virtual seems like an IT managers idea of heaven: less physical kit = less power = less cash = everyone's happy right?  Maybe...

Virtualisation at the server level is probably the most popular deployment.  By this I refer to the likes of VMWare or Microsoft Hyper-V which creates a hypervisor that sits on the physical tin and basically splices, isolates and distributes the physical components into virtual mini-machines.  These mini-machines can be individual servers running a plethora of different operating systems all using the same underlying physical machine. Neat eh?  Provisioning and de-provisioning a new server takes seconds.  Fault tolerance across applications, servers and infrastructure is simpler and more cost effective.  Virtualised networking allows the virtual machines to talk to each other without touching a physical ethernet cable.  Nice.

But does this concentration of resources increase of decrease or increase security?  I'm thinking from a physical perspective you have less kit to worry about.  Good thing right?  Well, you have this concentration factor which results in a smaller attack space that attackers can focus on.  If previously you had 250 physical DL380 servers racked in 4 server rooms which now becomes a pair of clustered boxes in 2 server rooms, there are less things that need protecting and there are less entry points an attacker could focus on.  However, if an attack were to be successful, the rewards are simply much higher.

If a hacker was able to gain access to the host machine (the physical kit running the hypervisor) they'd then had access to all the virtual machines running underneath?  Well perhaps not directly as they will be some logical security and isolation of those virtual machine files, but it theory you have more chance of creating a  DoS scenario than if you had 250 physical machines which would require 250 separate attacks.

From a practical perspective this virtual world will need to be managed and administered by someone.  This normally falls to a server support team of some description.  Whilst previously your pre-virtualised server world might be managed by different specialists (Unix, Solaris, Windows, Database, Web etc) the post-virtualisation world is managed at a meta level by a single team.  So again, the simplest place for an attack would be via the tools and interfaces that manage the virtual environment as that would give meta-access to the machines underneath - their configuration, their power on status, network configuration and so on.

Another area this virtualised concentration factor might influence is the physical aspect.  So previously you may have had more distributed physical machines.   Even in a concentrated virtual scenario with some sort of bi-locational clustering, you have fewer physical racks and more importantly fewer ethernet cables hosting your environment.  What changes need to be made to gain access to that physical environment?  If you gain access to the physical patching of a host machine, you potentially have access to sniff traffic from multiple machines which may in the past have taken more effort.

Whilst this is all a simplified view and there are many logical controls and processes that aide virtualisation security, many areas of weakness still exist, mainly around resource and administration concentration.  A reduced footprint can make administration and protection easier, but that protection will generally come under greater attack as the prize is now much higher.

(Simon Moffatt)