Adaptive networked control of misinformation epidemics: safety and usability trade-offs with resilient estimation

We propose a resilient framework for the mitigation of misinformation epidemics within dynamic information ecosystems under operational latencies and adversarial telemetry corruption. A dual-layer control architecture that balances platform-level usability with hard safety constraints is designed. The framework utilizes a polyhedral backward induction scheme to synthesize a verified controlled invariant cover. This geometric formulation guarantees that node-level infodemic penetration levels remain strictly bounded within a designated safe target set. To counter coordinated false data injection (FDI) attacks on state reporting channels, we integrate an online state observer, utilizing a private physical watermarking sequence, $$Delta w(t)$$. This mechanism creates an asymmetric information structure that exposes stealthy evasion tactics through a Chi-Squared ($$chi ^2$$) tracking residual monitor. Parametric sensitivity profiling maps the operational boundaries of the network, isolating the primary destabilizing role of virality ($$beta $$) alongside the primary stabilizing lever of intervention effectiveness ($$kappa $$). Empirical validation conducted on synthetic networks and the ESOC COVID-19 Misinformation Dataset demonstrates that the self-triggered adaptive control law consistently outperforms baseline implementations, yielding a platform usability cost reduction between $$38.5%$$and $$53.8%$$while maintaining absolute safety integrity. These results establish the framework as a robust tool for securing critical information infrastructure against sophisticated, coordinated manipulation.