Captain Kirk lied to us... space wasn't the final frontier, the data is.
All kidding aside, if you had a chance to participate in or follow the June 28th edition of the #ConvCloud chat we started by asking what things keeps our IT leaders up at night when it comes to cloud.
What erupted was a discussion that ranged from application resiliency (being able to tolerate and recover from failure of all kinds) to security and trust across the cloud landscape.
Ultimately though, as the title of the blog post suggests, we started seeing a theme. The theme that emerged over the hour of Twitter conversation was that no matter what, we need to find ways to protect data rather than trying to build elaborate and often difficult architectures to compensate for today's poorly written applications.
This may seem like an obvious point, but it's surprising how many conversations start at needing to "secure the environment" first... which is obviously a silly request when you're thinking about cloud computing.
Consider a public cloud provider where you the consumer are getting an environment that is standard across many, many other tenants. The idea of multi-tenancy is scary to many of us in the information security community - but why?
A closer look at that question of multi-tenancy and security seems to go back to having no control over the infrastructure and the network-based security controls that are required.
Worse yet, because we're in a state where the front-line network security must be amicable to the entire population of tenants, you can't have too complex or specific a configuration - this, some may argue, diminishes the overall value. Sure, I can agree with that on some level...
Next we start to think about what is required to have security in a multi-tenant world where your neighbor may be attacking you - so we collapse the perimeter down to the individual application. Logically, this makes sense in a discussion... except for one thing. We must concede that rarely are applications contained within a single virtual host (server).
This implies that machines interact with each other, or rather the application interacts with other components on different virtual machines, in a potentially hostile environment. Interesting... right?
So the perimeter isn't really around the application itself because the application is like an amoeba, with potentially many extensions to other (possibly) poorly secured or risky components. If the application isn't the perimeter, where do we collapse further? Elementary my dear Watson, the data!
If we as IT professionals and architects acknowledge that the perimeter is now around the data, what solutions do we have for protecting the perimeter? How can we protect data which is mobile, usable, and potentially in a constant state of danger? The answer seems to be some form of protection that involves our old friend, encryption.
In fact, encryption appears to be a reasonable solution to a lot of problems that the cloud presents us with from data residency (otherwise known as data sovereignty) to theft, to compliance, to storage destruction.
Unfortunately, nothing is ever quite this simple. Encryption that solves this many issues, across this many technologies must be managed appropriately. I'm talking, of course, about key management.
Is there an effective way to store, use and manage encryption keys in complex environments like the cloud, and distributed (converged) cloud environments? Yes, I'm confident there are but they are not simple to implement nor fool-proof.
The other challenge to the question of encryption is "What do you encrypt? and how?" This isn't trivial... you can encrypt the volume (virtual data store), the database, or each individual data element - but how do you then make that entire mess useful?
Format-preserving encryption is a fantastic idea and has been proven to work for PCI purposes, but it depends on the application. Can you simply do encryption at "all of the above" places?
Sure... but that adds an unnecessary element of complexity for most organization's needs. This isn't trivial... Let's see what we've just acknowledged:
- the perimeter is now collapsing (rapidly) to the data itself
- a reasonable answer to "how?" requires the use of encryption
- encryption appears to resolve many issues IT and business-wide
- encryption requires proper implementation and key management, which is difficult
In the final analysis of this chat, I think we do need to all agree that the perimeter is collapsing again, from the network, to the application, and now down to the data element itself. This I believe is a liberating realization.
Organizations seeking to reap the benefits of the cloud will need to re-architect many of their applications, re-think their security models, and re-assess their processes/procedures... but in spite of all that work, I believe that the benefit still outweighs the work involved up-front.
Join us every Thursday at 1:00pm Eastern for a different topic on the cloud using the #ConvCloud hashtag on Twitter. I'll see you next time!
Cross-posted from Following the White Rabbit