Neuromorphic Chaotic Dynamics as a Software-Defined Source of Cryptographic Randomness

Abstract

Cryptographic systems fail when their randomness fails. Most software generators are deterministic by construction, while high assurance entropy sources are usually tied to dedicated hardware. This preprint examines a third direction: a neuromorphic,
chaotic, time-axis-driven entropy source implemented as a software-defined dynamical system.

The claim is intentionally behavioral rather than architectural. Internal design details are withheld. What is reported instead is
the observable statistical behavior of the raw stream under familiar validation frames. In the latest recorded evaluation set, the
source passed 186/188 NIST STS subtests, produced a broad dieharder pass profile with a single explicit failure and two weak flags, and reached a latest direct SP 800-90B result of 7.883983 bits/byte under the IID path and 7.322342 bits/byte under the final conservative non-IID path.

These numbers do not establish a finished certification-ready entropy subsystem. They do establish something more important
for a preprint: a neuromorphic temporal process can produce statistically credible cryptographic randomness in software, and it
can do so while operating as a dynamic process rather than as a conventional static pseudorandom routine.