Stealth cyber intrusions in networked DC microgrid clusters can modify converter measurements without triggering standard alarms and, as a result, can degrade stability in ways that are not immediately visible at the supervisory level. In droop-controlled DCMGs with hierarchical control, this issue is particularly acute because the same measurement channels are used for both voltage restoration and power sharing. This work analyzes the small-signal stability of a parallel-connected DCMG cluster subjected to step-type false-data-injection attacks on converter voltage sensors. The attack is modeled as a nonlinear, discontinuous perturbation of the voltage feedback, and a quasi-linearized state-space model is derived around the attack-dependent equilibrium. From this model, operating-point dependent loop transfer functions are obtained and evaluated using Nyquist plots, gain and phase margins, and eigenvalue migration in the s-plane. The results show that increasing attack magnitude progressively reduces classical stability margins, drives lightly damped poles toward the imaginary axis, and ultimately produces right-half-plane eigenvalues for sufficiently large FDI levels.
