This paper addresses the centralized fusion estimation problem in networked systems with stochastic uncertainties characterized by random parameter matrices together with multiplicative and additive noises. To reflect real-world engineering situations, it is further assumed that the network transmissions are simultaneously subject to random packet dropouts and deception attacks which randomly alter real measurements by replacing them with noises. A novel approach is proposed that avoids the need for a specific state equation, relying instead only on the mean and covariance functions of the processes involved. The additive noises in the sensor measurements are considered to be time-correlated and packet dropouts are managed through a zero-order hold compensation strategy that attenuates the effect of data loss on the estimation process. On the basis of the available measurement information, recursive fusion filtering and smoothing algorithms are developed using an innovation-based methodology. The proposed approach is validated by numerical simulations, demonstrating its feasibility and correctness. Comparative results show the superior performance of the proposed fusion estimation scheme over existing filters in the literature, highlighting its effectiveness in mitigating the impact of deception attacks and packet dropouts in networked systems.