To better understand the effects of short-term computer-based cognitive rehabilitation (cCR) on cognitive performances and default mode network (DMN) intrinsic functional connectivity (FC) in cognitively impaired relapsing remitting (RR) multiple sclerosis (MS) patients. Eighteen cognitively impaired RRMS patients underwent neuropsychological evaluation by the Rao's brief repeatable battery and resting-state functional magnetic resonance imaging to evaluate FC of the DMN before and after a short-term (8 weeks, twice a week) cCR. A control group of 14 cognitively impaired RRMS patients was assigned to an aspecific cognitive training (aCT), and underwent the same study protocol. Correlations between DMN and cognitive performances were also tested. After cCR, there was a significant improvement of the following tests: SDMT (p < 0.01), PASAT 3″ (p < 0.00), PASAT 2″ (p < 0.03), SRT-D (p < 0.02), and 10/36 SPART-D (p < 0.04); as well as a significant increase of the FC of the DMN in the posterior cingulate cortex (PCC) and bilateral inferior parietal cortex (IPC). After cCR, a significant negative correlation between Stroop Color-Word Interference Test and FC in the PCC emerged. After aCT, the control group did not show any significant effect either on FC or neuropsychological tests. No significant differences were found in brain volumes and lesion load in both groups when comparing data acquired at baseline and after cCR or aCT. In cognitively impaired RRMS patients, cCR improves cognitive performances (i.e., processing speed and visual and verbal sustained memory), and increases FC in the PCC and IPC of the DMN. This exploratory study suggests that cCR may induce adaptive cortical reorganization favoring better cognitive performances, thus strengthening the value of cognitive exercise in the general perspective of building either cognitive or brain reserve.