Microsoft throws crypto foes an untouchable elliptic curveball (open source encryption code!)

The Register ^ | Sep 15, 2015 | Richard Chirgwin

[dayglored]

While Washington mulls ways to make crypto less effective, the industry, thank heavens, continues to push in the other direction. Microsoft Research has just published an elliptic curve library it reckons is considerably faster than what's currently available.

Outlined in this International Association for Cryptologic Research (IACR) paper, the implementation, the FourQLib, comes from noted Redmondian researchers Craig Costello and Patrick Longa.

The aim with FourQ sis to update today's elliptic curve cryptography (ECC) – implementations like the National Institute of Science and Technology's (NIST's) P-256 and the non-NIST-influenced Curve25519 – since one inevitability of crypto tech is that it will become obsolete.

That be-prepared approach also means a new ECC library needs to be fast, so that crypto doesn't become too great a burden on the processors handling it. Here's what the authors say about FourQ's performance:

“On Intel’s Haswell, Ivy Bridge and Sandy Bridge architectures, our software computes a variable-base scalar multiplication in 59,000, 71,000 cycles and 74,000 cycles, respectively; and, on the same platforms, our software computes a Diffie-Hellman shared secret in 92,000, 110,000 cycles and 116,000 cycles, respectively."

...

And importantly, the code has been made public – it's available for download here – for others to audit.

[dayglored]

Very good!

[PreciousLiberty]

To add a few points to your post:

ECC is considered more secure than RSA for the same size keys.
ECC requires less compute resources than RSA.
There is a known quantum computer algorithm to quickly crack RSA encryption, given enough entangled bits (a working quantum computer). There is no such known algorithm for ECC.

Since this is open source, it will be audited by numerous crypto experts for potential weaknesses. I’m quite sure Microsoft had some very capable crytpo folk design this library.

There is a sticky point in that there are coefficients involved in the ECC algorithm, and wide distrust of those recommended by the NSA. I don’t know where things are at regarding arriving at widely trusted coefficients.

I’m no big fan of Microsoft in general, but in this case it is to be applauded.