Monday, December 10, 2012

Indocrypt 2012

Indocrypt 2012 is currently on in Calcutta, India. I am there as an invited speaker, along with Vinod Vaikuntanathan and Orr Dunkelman. I expect my invite has more to do with my good friend Steven Galbraith being programme co-chair than anyone wanting to hear me speak.  In any case the hospitality and organization shown by our hosts has been amazing. Unlike the major conferences, conferences like Indocrypt and Africacrypt always have a great deal of speaker/audience interaction outside of the main talks and it is always great to see new faces joining the subject.

The programme is slightly biased towards more symmetric cryptography style areas; an area I am not really expert in. However, one highlight in this area for me was a tutorial on the RC4 cipher and various cryptanalytic results given by Subhamoy Maitra. This complimented two tutorials given the day before the conference started by Steven Galbraith on lattices in cryptography and by Francisco Rodriguez-Henriquez on hardware design for cryptography.  Francisco's talk gave a gentle introduction to the different issues one faces when moving from a mindset of software implementation to hardware implementation; as well as an interesting subsection on implementation of a length preserving mode of operation using a tweakable block cipher.

The first morning was however much more "up my street". The first session being particularly interesting, with four very good talks. Focused around the loose theme of protocols, the four talks considered completeness theorems for MPC, aspects of the Fiat-Shamir transform from Sigma protocols to NIZKs, a look at the BB84 QKD protocol and leakage resilient MPC protocols.

The BB84 talk, by Arpita Maitra, was particularly appealing as it showed how a simple observation could be overlooked for a long time. The basic idea was that the original BB84 protocol sends four qubits per agreed key bit (essentially), a problem related to the resulting security gaurantee was found in this a while back and a modified protocol was presented which sends six qubits per key bit. However, what the Indocrypt talk showed was that if one repeated the two protocols and looked at the bit-security per qubit sent (which is an obvious metric used in other areas of cryptography) then the original four qubit BB84 protocol turns out to be better.

Vinod Vaikuntanathan gave a nice invited talk on recent advances in Fully Homomorphic Encryption. Such talks are now becoming very accessible, although perhaps I am not the best person to comment on this. In the past explaining the basic idea behind the original scheme of Gentry was challenging in a one hour slot. But now with the advent of the LWE based schemes, and a better understanding of how the different ideas fit together, such talks are I think more accessible to a general crypto audience. Of course this talk was aided in this respect by Vinod's standard brilliant delivery, which makes even the most complex concept look effortlessly simple.

No comments:

Post a Comment