For years computer scientists have talked about the very interesting challenge regarding the desire to be able to operate over encrypted data in an advanced way called Homomorphic Encryption. This is a field that gets nuanced fast, with terms like "partially homomorphic" relevant to many existing systems that are in place, but Fully Homomorphic thought by many to be so far away to be just a dream.
No computer system on the market today can do fully homomorphic processing. Data might be stored on hard drives encrypted, and might be moved around between computers encrypted, and might even be stored in RAM or moved around places on boards in computers while encrypted. But when it comes to processing, the data must always be operated on by the computer in the clear. That is just the way computers work. So, while there have been great advances in security, there is always going to be a place in time and space where data is totally unencrypted and therefore, in theory, a malicious actor could exploit this to their advantage. It might be hard to do, but when the data is valuable bad guys always find a way.
Some very interesting research on theoretical implementations is available. As far as I can tell, most working ideas flow from those written by Craig Gentry, an incredibly bright IBM researcher. You can catch him explaining some of his research at the video at this link and below:
It is clear this research is producing some working homomorphic encryption schemes. But I have seen nothing in any of the research that makes me think a solution can be put in place that cannot be defeated by bad guys. And if that can't be done then the solution will not solve any problems, it will just add processing overhead. So in the end I remain a skeptic regarding any claims of a working fully homomorphic solution.
Said another way, when I read of claims that this can be done I think there must be a catch and wanted to give you that warning before you read the below. I imagine that if this were put to a test by a team of security professionals they would find alarming vulnerabilities. The only thing I have that backs up this assertion is years of experience that says the bad guys can always get in when they try, plus a deep belief that true fully homomorphic encryption can only be done in ways that introduce new vulnerabilities.
That said, I am providing IBM's release below, and plan on tracking this and other reporting on this topic:
Made in IBM Labs: Advancing Privacy and Security in the Cloud
Patented cryptography invention enables unlimited analysis of encrypted data
ARMONK, N.Y. - 23 Dec 2013: IBM (NYSE: IBM) inventors have received a patent for a breakthrough data encryption technique that is expected to further data privacy and strengthen cloud computing security.
The patented breakthrough, called “fully homomorphic encryption,” could enable deep and unrestricted analysis of encrypted information —intentionally scrambled data — without surrendering confidentiality. IBM’s solution has the potential to advance cloud computing privacy and security by enabling vendors to perform computations on client data, such as analyzing sales patterns, without exposing or revealing the original data.
IBM's homomorphic encryption technique solves a daunting mathematical puzzle that confounded scientists since the invention of public-key encryption over 30 years ago.
Invented by IBM cryptography Researcher Craig Gentry, fully homomorphic encryption uses a mathematical object known as an “ideal lattice” that allows people to interact with encrypted data in ways previously considered impossible. The breakthrough facilitates analysis of confidential encrypted data without allowing the user to see the private data, yet it will reveal the same detailed results as if the original data was completely visible.
IBM received U.S. Patent #8,565,435: Efficient implementation of fully homomorphic encryption for the invention, which is expected to help cloud computing clients to make more informed business decisions, without compromising privacy and security.
“Our patented invention has the potential to pave the way for more secure cloud computing services – without having to decrypt or reveal original data,” said Craig Gentry, IBM Researcher and co-inventor on the patent. “Fully homomorphic encryption will enable companies to confidently share data and more easily and quickly overcome challenges or take advantage of emerging opportunities.”
Following initial revelation of the homomorphic encryption breakthrough in 2009 Gentry and co-inventor Shai Halevi began testing, refining and pursuing a working implementation of the invention. In 2011, the scientists reported a number of optimizations that advanced their goal of implementing of the scheme. The researchers continue to investigate homomorphic encryption and test its practical applicability.
IBM invests more than $6 billion annually in R&D and consistently explores new approaches to cloud computing that will deliver a competitive advantage to the company and its clients.
For 20 consecutive years, IBM has topped the list of U.S. patent recipients. The company’s invention and patent leadership is illustrated at http://ibm.co/11k6fRn.
IBM has a tradition of making major cryptography breakthroughs, such as the design of the Data Encryption Standard (DES); Hash Message Authentication Code (HMAC); the first lattice-based encryption with a rigorous proof-of-security; and numerous other solutions that have helped advance data security.