In a microgrid with hierarchical control, a typical cyber–physical power system (CPPS), cyber failures may induce physical system damage in a cross-layer fashion. Particularly during extreme conditions, the reliability of communication facilities may be jeopardized, inducing a degraded quality of service (QoS) and subsequently disrupting frequency and voltage regulation in a cross-layer fashion. This article proposes a co-optimization technique on QoS, frequency regulation and voltage regulation that comprises: 1) a specialized CPPS model for quantitatively analyzing the cross-layer impact of resource allocation to physical states, specifically frequency and voltage; 2) a multiobjective formulation aimed at enhancing QoS while concurrently minimizing the impact on physical states regulation; and 3) a weight adjusted model reference adaptive search algorithm to narrow the search space by leveraging the characteristics of our CPPS model. Compared with state-of-the-art techniques that focus solely on optimizing QoS, our proposed technique achieves a reduction of frequency (13.74%) and voltage (4.57%) deviations, albeit with a minor compromise in QoS (0.34%).