September 16, 2009

Cloud security through control vs.ownership

Cloud computing makes auditors cringe. It's something we hear consistently from enterprise customers: it was hard enough to make virtualization "palatable" to auditors; cloud is going to be even harder. By breaking the links between hardware and software, virtualization liberates workloads from the physical constraints of a single machine. Cloud takes that a step further making the physical location irrelevant and even obscure.

Traditionally, control of information flows directly from ownership of the underlying platform. In the traditional security model location implies ownership, which in turn implies control. You build the layers of trust with the root of trust anchored to the specific piece of hardware. Virtualization breaks the link between location and application. Cloud (at least "public cloud") further breaks the link between ownership and control.

As we've examined in many previous columns we are rapidly moving from a location-centric security model to a more identity- and data-centric model. The unstoppable forces of ubiquitous connectivity and mobility have broken the location-centric security model and perimeter strategy, and left us searching for a better model for security. In the process, certain fundamental assumptions have also changed. When security is location-centric, then location, ownership and control are aligned. The logical security model coincides with the physical security model and a perimeter separates trusted (owned, local) from untrusted (other, remote). As we move beyond this model we have to examine the links between location, ownership and control.

Control of information is not in fact dependent on total ownership or a fixed location. An easy example is public key encryption. I maintain ownership of a private key and I control access to it. Usually the private key is stored in a secure location. But from the ownership of the key I can exert control over the information without having to own the rest of the infrastructure. I can build a trusted VPN over an untrusted infrastructure.

The key here is that public cloud computing requires us to exert control without ownership of the infrastructure. We can exert control and secure the information through a combination of encryption, contracts with service-level agreements and by (contractually) imposing minimum security standards on the providers. If those are in place, then there is no inherent reason why a cloud computing environment cannot be made secure and compliant. We do not need to own the assets in order to exert security, anymore than we need to own the Internet in order to trust a VPN.

Auditors and regulators are continuously adapting to new technologies and business models. As long as we can clearly demonstrate control through technology and contracts we should be able to make a cloud computing environment as compliant and as secure as a privately owned facility.

Talk to your auditors about their understanding of risk as it relates to location, ownership and control. Once you clearly separate the concepts you might find it easier to have that discussion.

---

DONATION: www.anti-keylogger.org and www.keylogger.org is an independent research projects supported by a team of enthusiasts. If you find this project useful and would like to help foster its continued development, please consider making a donation.

Thanks in advance for your support!