Localizing interictal epileptic activities is a difficult problem in clinical practice. We report a novel noninvasive technique, resting functional magnetic resonance imaging (fMRI) with spatio-temporal independent component analysis (ICA), for localizing interictal epileptic activities. First, the fMRI data is separated into independent spatial patterns by spatial-ICA, and the patterns with Z-values larger than a threshold are selected as the potential spatial patterns of the epileptic activities. Second, the temporal series of the active points in the selected patterns are separated by temporal-ICA, and the component with the biggest Gaussian deviation (kurtosis) is selected as the representative of the epileptic discharge activity in a sub-region. Finally, those spatial sub-regions, which have distinct epileptic discharge activities confirmed by temporal-ICA are considered as the epileptic foci. This method was applied to fMRI data of six epileptic patients, and the results are consistent with the clinical assessment. Though more studies are required to validate this technique, the above preliminary results demonstrate the potential of using the resting fMRI with spatio-temporal ICA to detect and localize latent epileptic activities.