The default mode network (DMN) is a network of brain regions that are active during rest and suppressed during a cognitively demanding task. Previous studies have shown that the DMN can be altered by development, aging, disorder, cognitive tasks and offline training. However, it's unclear whether activity in the DMN can be altered by real-time training. Recently, real-time functional magnetic resonance imaging (rtfMRI), as a novel neurofeedback technique, has been applied to train subjects to voluntarily control activities in specific brain regions. In the current study, it was found that by using rtfMRI to guide training, subjects were able to learn to decrease activity in the posterior cingulate cortex (PCC), which is a "key hub" in the DMN, using motor imagery strategy. After the real-time training, activity in the medial prefrontral cortex/ anterior cingulate cortex (MPFC/ACC) of the resting state DMN was decreased. By contrast, the control group without neurofeedback produced increased activity in the MPFC/ACC of the DMN during the post-training resting state. These findings suggest that this rtfMRI technique has great potential to be used in the regulation of the DMN and may be a novel approach for studying functional plasticity of the cortex.