No abstract uploaded!

No abstract uploaded!

Key exchange protocols from the learning with errors (LWE) problem share many similarities with the Diffie–Hellman–Merkle (DHM) protocol, which plays a central role in securing our Internet. Therefore, there has been a long time effort in designing authenticated key exchange directly from LWE to mirror the advantages of DHM-based protocols. In this paper, we revisit signal leakage attacks and show that the severity of these attacks against LWE-based (authenticated) key exchange is still underestimated. In particular, by converting the problem of launching a signal leakage attack into a coding problem, we can significantly reduce the needed number of queries to reveal the secret key. Specifically, for DXL-KE we reduce the queries from 1,266 to only 29, while for DBS-KE, we need only 748 queries, a great improvement over the previous 1,074,434 queries. Moreover, our new view of signals as binary codes enables recognizing vulnerable schemes more easily. As such we completely recover the secret key of a password-based authenticated key exchange scheme by Dabra et al. with only 757 queries and partially reveal the secret used in a two-factor authentication by Wang et al. with only one query. The experimental evaluation supports our theoretical analysis and demonstrates the efficiency and effectiveness of our attacks. Our results caution against underestimating the power of signal leakage attacks as they are applicable even in settings with a very restricted number of interactions between adversary and victim.

We develop the affine sieve in the context of orbits of congruence subgroups of semisimple groups acting linearly on affine space. In particular, we give effective bounds for the saturation numbers for points on such orbits at which the value of a given polynomial has few prime factors. In many cases these bounds are of the same quality as what is known in the classical case of a polynomial in one variable where the orbit consists of the integers. When the orbit is the set of integral matrices of a fixed determinant, we obtain a sharp result for the saturation number, and thus establish the Zariski density of matrices all of whose entries are prime numbers. Among the key tools used are explicit approximations to the generalized Ramanujan conjectures for such groups, and sharp and uniform counting of points on such orbits when ordered by various norms.

No abstract uploaded!