This paper addresses the distributed fusion estimation problem of stochastic signals from quantized measurements with random parameter matrices and time-correlated additive noises. These measurements are assumed to be exposed to mixed network attacks, including both random deception attacks and denial-of-service (DoS) attacks and the stochastic nature of these attacks is aptly modeled by Bernoulli random variables. Using a covariance-based methodology and a prediction compensation strategy to counteract the random loss of information caused by DoS attacks, recursive algorithms are designed for the distributed fusion filtering and fixed-point smoothing problems.