We propose a novel flagging system (FS) to make the Neural Network Controller (NNC), which is an extant real-time dynamic buffer tuner, more realistic for real-time applications over the mobile Internet. Even though the extant NNC may prevent buffer overflow, it fails under certain traffic conditions easily. This happens because the intrinsic NNC control intelligence cannot handle the whole spectrum of system dynamics. With the help of this novel FS, which involves real-time elements that include the Convergence Algorithm, Hurst filter and Holder filter, the NNC should be able to tremendously reduce or even eliminate the chance of buffer overflow under its control on the fly.