Researchers from Microsoft Research have been able to speed up calculations in a homomorphic cryptographic system by several orders of magnitude.

Homomorphic encryption is a cryptographic system that allows you to perform certain mathematical operations with clear text by performing (in general, other) operations with encrypted text.

The first fully homomorphic cryptosystem was created by Craig Gentry in 2009. However, the computation speed was so low that it was impossible to talk about any practical application.

A completely homomorphic cryptographic system with an acceptable speed of operation is a kind of “Holy Grail” of cryptography. It would allow searching, statistical calculations and any other operations with encrypted data, which would eliminate the need to find a compromise between privacy and convenience. This would give a powerful impetus to cloud platforms, solve the issue of leakage of personal data of users of many Internet services, greatly simplify the life of organizations working with sensitive medical or financial data.
')
Christine Loter and her colleagues Vinod Vaikuntanathan and Mikael Nerig managed to create a prototype of such a system capable of doing 100 addition operations of 128-bit numbers in 20 milliseconds on a regular laptop. Previously, such calculations took minutes and hours.

For the sake of such acceleration, we had to make some compromises, so, strictly speaking, the system cannot be called completely homomorphic. But even with a limited set of mathematical operations, it is already possible to perform some statistical calculations. For example, on the basis of medical indicators from the database to calculate the probability of a heart attack in a patient.

It is the work with medical information that Christine Lother cites as an example of an area where practical use of homomorphic cryptography can begin in the near future. Still, the turtle computing speed does not allow using such systems widely, but where confidentiality may be much more important than performance — in banking, in medicine or in government databases — it makes sense to look at these experiments right now.

Details:

A Cloud that Can't Leak
Detailed commentary on the news on Slashdot
Habratopik about homomorphic encryption