In some recent advanced protocols, like Zero-Knowledge Proofs (allowing a prover to convince a verifier that he knows a secret without revealing it), or Multi Party Computation, algorithms do not process bit-oriented information, but rather elements of big finite fields

We present a faster dual lattice attack on the Learning with Errors (LWE) problem, based on ideas from coding theory. Basically, it consists in revisiting the most classical dual attack by replacing modulus switching by a decoding algorithm. This replacement