One-way multilinear functions of the second order with linear shifts

Abstract

We introduce and analyze a novel class of binary operations on finite-dimensional vector spaces over a field K, defined by second-order multilinear expressions with linear shifts. These operations generate polynomials whose degree increases linearly with each iterated application, while the number of distinct monomials grows combinatorially. We demonstrate that, despite being non-associative and non-commutative in general, these operations exhibit power associativity and internal commutativity when iterated on a single vector. This ensures that exponentiation a^n is well-defined and unambiguous. Crucially, the absence of a closed-form expression for a^n suggests a one-way property: computing a^n from a and n is efficient, while recovering n from a^n (the Discrete Iteration Problem) appears computationally hard. We propose a Diffie-Hellman-like key exchange protocol based on this principle, introducing the Algebraic Diffie-Hellman Problem (ADHP) as an underlying assumption of security. In addition to the algebraic foundations, we empirically investigate the orbit structure of these operations over finite fields, observing frequent emergence of long cycles and highly regular behavior across parameter sets. Motivated by these dynamics, we further propose a pseudorandom number generation (PRNG) strategy based on multi-element multiplication patterns. This approach empirically achieves near-maximal cycle lengths and excellent statistical uniformity, highlighting the potential of these operations for cryptographic and combinatorial applications.