Tuesday, February 09, 2016

[ldhiqzie] Deduplication with encryption

Encryption and deduplication seem fundamentally incompatible, so we consider workarounds, i.e., "cheating", for the application of cloud storage.

One simple method is to charge for storage, and charge less for deduplicated (so unencrypted) stored content.  Pass along the cost savings of deduplication to the consumer.  This places the burden on the user to decide whether to store an object encrypted or not, based on whether or not he or she expects to be able to save money by amortizing the storage cost among many users who are also storing the same content.  The user has knowledge of whether other users might be storing copies of a given content (e.g., for original work and private information, probably not).  The service could provide hints of what content can be stored cheaper (and how much cheaper) if stored unencrypted.  One tricky issue is, if other users delete their copy of the content, then one's cost of storing it increases unexpectedly as it is shared by less people.

Also, content explicitly shared between multiple users could be encrypted with a common key.  There might be devilish details.

No comments :