FuLeakage: Breaking FuLeeca by Learning Attacks

Abstract

We show that the code-based signature scheme FuLeeca is vulnerable to lattice-based cryptanalysis. A classical attack using lattice-basis reduction lowers the claimed security levels significantly, whereas learning techniques allow to recover the secret key from the leakage of less than 175.000 signatures in practice. The exploitation of ideal structures and efficient quantum algorithms further yields a full quantum break.