Introduction to Learning Problems in Cryptography
Emeline Repel (GREYC, Caen)Cryptographic schemes are designed to ensure protecting data in many types of communication (payment, instant messaging, web etc). Their security relies on mathematical problems, known hard to solve with classical computing. Those problems are called primitives. In the context of quantum computing, cryptographic primitives are about to change to ensure security against quantum adversaries in addition to classical adversaries.To handle this problem, the National Institute of Standards and Technology (NIST) began a standardization campaign for post-quantum cryptosystems. From that, new primitives are arising, and particularly lattices problems computationally hard to solve like the Learning With Error (LWE) problem [Regev05].
However, in the precipitation of the standardization process, some aspects of cryptanalysis are not well studied. Especially, physical attacks (using timing, electrical consumption, magnetic field, …), namely Side Channels Analysis, permit to recover additional information on the secret. From those analyses, new problems are arising, in between theoretical security and practical security, to ensure that finding the secret is not that easy with that information.