In this paper, we present noise-domain non-orthogonal multiple access (ND-NOMA), an innovative communication scheme that utilizes the modulation of artificial noise mean and variance to convey information. Distinct from traditional methods such as power-domain non-orthogonal multiple access (PD-NOMA) that heavily rely on successive interference cancellation (SIC), ND-NOMA utilizes the noise domain, considerably reducing power consumption and system complexity. Inspired by noise modulation, ND-NOMA provides lower bit error probability (BEP), making it highly suitable for next-generation Internet-of-things (IoT) networks. Our theoretical analyses and computer simulations reveal that ND-NOMA can achieve exceptionally low bit error rates in both uplink and downlink scenarios, in the presence of Rician fading channels. The proposed multi-user system is supported by a minimum distance detector for mean detection and a threshold-based detector for variance detection, ensuring robust communication in low-power environments. By leveraging the inherent properties of noise, ND-NOMA offers a promising platform for long-term deployments of low-cost and low-complexity devices.