This paper is concerned with the decentralized resilient H∞ filtering problem for a class of large-scale discrete-time systems with delayed interactions under communication constraints. A switched time-delay system with randomly occurring uncertainties is employed to model the network-based filtering system where measurement quantization and data missing phenomena are taken into account. The main purpose of this paper is to design decentralized resilient filters with randomly occurring gain variations such that filtering error system is exponentially stable in the mean square and satisfied a prescribed H∞ disturbance attenuation. Sufficient condition are first established in virtue of the switched system method for the existence of the desired decentralized resilient filters. By utilizing the conditions, the filter gains are then designed. Finally, a numerical example is illustrated to show the efficacy of the results proposed.