Spring 2020

Lower Bounds on Lattice Sieving and Information Set Decoding

Tuesday, Jun. 15, 2021 8:00 am8:30 am PDT

Add to Calendar


Elena Kirshanova (I.Kant Baltic Federal University)



The talk will be on the recent result on  new lower bounds on the costs of solving the nearest neighbor search problems appearing in cryptanalytic settings. For the Euclidean metric we show that for random data sets on the sphere, the locality-sensitive filtering approach of [Becker--Ducas--Gama--Laarhoven, SODA 2016] using spherical caps is optimal, and hence within a broad class of lattice sieving algorithms covering almost all approaches to date, their asymptotic time complexity of $2^{0.292d + o(d)}$ is optimal. For the Hamming metric we derive new lower bounds for nearest neighbor searching which almost match the best upper bounds from the literature [May--Ozerov, Eurocrypt 2015]. As a consequence we derive conditional lower bounds on decoding attacks, showing that also here one should search for improvements elsewhere to significantly undermine security estimates from the literature.

This is the joint work with T.Laarhoven that has been initiated at the Spring 2020 program “Lattices: Algorithms, Complexity, and Cryptography.”

PDF icon Slides320.67 KB