The theory of domination in graphs has traditionally operated within the binary framework of classical graph theory, wherein a vertex either dominates a neighbour or does not. This paper introduces a comprehensive generalization through fuzzy domination, where vertex membership values σ: V → [0, 1] and edge membership functions μ: E → [0, 1] replace the crisp adjacency relation with graded influence. We define the fuzzy dominating set, fuzzy domination number γf(G), and prove tight upper and lower bounds relating γf(G) to the classical domination number γ(G) and the order n of the graph. The central contribution of this paper is a cryptographic key generation scheme that exploits the minimum fuzzy dominating set as a combinatorial entropy source. By encoding the dominating vertex set and its weighted cover structure into a seed-extraction function composed with a Merkle–Damgard hash chain, we construct a deterministic key generation algorithm whose security reduces to the hardness of computing minimum fuzzy dominating sets in sparse random graphs, a problem we show to be NP-hard via reduction from weighted set cover. Experimental evaluation on Erdos–Renyi random graphs with n ∈ {10, …, 100} vertices demonstrates that the proposed scheme achieves Shannon entropy of 7.9994 bits/byte, avalanche effect of 50.6 ± 0.28%, and a key space of 2384, outperforming three recent graph-theoretic key generation baselines. The paper additionally proves two new theorems on fuzzy domination in complete bipartite graphs and cycle graphs that are of independent mathematical interest.