Multivariate Polynomial Inference in a Cryptographic Setting

In this paper, we generalize to the multivariate setting the current state-of-the-art methods in the literature for the inference of bivariate polynomials constructed recursively, by means of repeated additions and multiplications. We present two main approaches: the first one based on polynomial interpolation and the second one relying on lattice-based techniques for solving modular knapsack-type problems. Both the directions yield natural and practical generalizations, supported by detailed analyses of the underlying mathematical structures. Our methods can be useful for analyzing the security of cryptographic algorithms, given their connection to basic operations serving as building blocks, for example in fully homomorphic encryption schemes.