This paper proposes an estimator-based iterative deviation-free residual generator (IDRG) for linear discrete-time stochastic systems. The communication between intelligent nodes and the remote estimator is realized by adopting a shared network. With the purpose of random access protocol (RAP) is utilized to schedule the data transmission process, according to which only one node at one instant is selected to get access to the communication network. The aim of the addressed problem is to detect the occurrence and disappearance of the faulty signal via utilizing the designed estimator. By solving an optimization problem subject to the deviation-free constraint, a novel batch estimator is firstly designed with finite response. Then, the IDRG is developed by extracting the residual signal out of the proposed estimator. By adopting the norm of the residual as the evaluation function, the Chebyshev inequalities are used to obtain the lower and upper stochastic thresholds. Finally, an illustrative example is given to demonstrate that the designed IDRG can detect the occurrence and disappearance of the faulty signals in a short transient stage, and is capable of reducing the false alarm rate in comparison to the benchmark Kalman filter residual generator (KFRG).