0
$\begingroup$

In my experience, conversations around Homomorphic Encryption tend to focus on the narrative "HE is the holy grail of secure computations, but it's still too computational intensive to be practical".

While in the last years the performance of HE has improved dramatically, one point has remained fixed (and will remain fixed): computing on encrypted data requires an entirely new paradigm to write software. In particular, old software cannot easily be retrofitted to use HE, unless the workflow is a simple sequential computation over the data (no if/else, no actions according to specific values of the data, etc.) and the format of data is exactly known in advance.

Am I missing something? Is there a simpler way to use existing software than rewriting it from scratch?

An example use case could be a Software-as-a-Service solution to do genomics manipulations, such as DNA string matching, in an untrusted cloud instance.

$\endgroup$
4
  • 1
    $\begingroup$ What about motivating the question by giving a plausible use case for Homomorphic Encryption, leaving aside the software rewrite issue? I know none. There is similarity with Quantum Key Distribution: it's a nice achievement that does not seem to solve any common actual problem. $\endgroup$
    – fgrieu
    Apr 16, 2020 at 11:04
  • $\begingroup$ added an example from genomics. Finance is another area with similar issues. $\endgroup$
    – Rexcirus
    Apr 16, 2020 at 13:58
  • $\begingroup$ I see. DNA string matching indeed seems a plausible problem. But devil lies in the details: are interesting enough queries possible with limitations as set by HE? In other words, is there a solvable issue to be solved that is not solved by means other than HE (e.g. mere anonymity, Private Information Retrieval..)? Won't the market be more satisfied with faster/better results at the cost of trusting the data holder? $\endgroup$
    – fgrieu
    Apr 16, 2020 at 14:50
  • $\begingroup$ HE is definitely more secure and privacy-preserving as a solution. Your last question is impossible to answer with confidence. But I'm more interested in the limit (say 5-20 years) in which HE speed is not much of a problem anymore. $\endgroup$
    – Rexcirus
    Apr 16, 2020 at 18:06

0

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.