Data Destruction Policies in the Age of Cloud Computing

These days, most big companies and many midsize ones have some form of a data-governance program, typically including policies for data retention and destruction. They have become an imperative because of increasing attacks on customer data and also state and national laws mandating protection of customer data. The old mind set of “Keep everything, forever” has changed to “If you don’t have it, you can’t breach it.”

In some ways, managing data-retention policies has never been easier to implement in the cloud. Cloud vendors often have easy templates and click-box settings to retain your data for a specific period and then either move it to quasi-offline cold digital storage or straight to the bit bucket (deletion). Just click, configure, and move on to the next information security priority.

Just Click Delete?

However, I’m going to ask an awkward question, one that has been burning in my mind for a while. What really happens to that data once you click “delete” on a cloud service? In the on-premises, hardware world, we all know the answer; it would simply be deregistered on the disk it resides on. The “deleted” data still sits on the hard drive, gone from the operating system view and waiting to be overwritten when the space is needed. To truly erase it, extra steps or special software are needed to overwrite the bits with random zeros and ones. In some cases, this needs to be done multiple times to truly wipe out the phantom electronic traces of the deleted data.

And if you do business with the US government or other regulated entities, you may be required to comply with Department of Defense standard 5220.22-M, which contains specifics on data destruction requirements for contractors. These practices are common, even if not required by regulations. You don’t want data you don’t need any more coming back to haunt you in the event of a breach. The breach of the Twitch game-streaming service, in which hackers were able to gain access to basically all of its data going back almost to the inception of the company — including income and other personal details about its well-paid streaming clients — is a cautionary tale here, along with reports of other breaches of abandoned or orphaned data files in the last few years.

Lack of Access to Verify

So, while the policies are easier to set and manage in most cloud services versus on-premises servers, assuring it is properly done to the DoD standard is much harder or impossible on cloud services. How do you do a low-level disk overwrite of data on cloud infrastructure where you don’t have physical access to the underlying hardware? The answer is that you can’t, at least not the way we used to do it — with software utilities or outright destruction of the physical disk drive. Neither AWS, Azure, or Google Cloud Services offer any options or services that do this, not even on their dedicated instances, which run on separate hardware. You simply don’t have the level of access needed to do it.

Outreach to the major services either was ignored or answered with generic statements about how they protect your data. What happens to data that is “released” in a cloud service such as AWS or Azure? Is it simply sitting on a disk, nonindexed and waiting to be overwritten, or is it put through some kind of “bit blender” to render it unusable before being returned to available storage on the service? No one, at this point, seems to know or be willing to say on the record.

Adjust to New Reality

We must develop a cloud-compatible way of doing destruction that meets the DoD standards, or we must stop pretending and adjust our standards to this new reality.

Maybe cloud providers can come up with a service to provide this capability, since only they have direct access to the underlying hardware. They have never been shy about inventing new services to charge for, and certainly plenty of companies would be eager to pay for such a service, if the appropriate certificates of destruction were provided. It would probably be cheaper than fees charged by some of the companies providing certified physical-destruction services.

Amazon, Azure, Google, and any major cloud service (even software-as-a-service providers) need to address these issues with real answers, not obfuscation and marketing-speak. Until then, we will just be pretending and hoping, praying some brilliant hacker doesn’t figure out how to access this orphaned data, if they haven’t already. Either way, the hard questions on cloud data destruction need to be asked and answered, sooner rather than later.