The Soil Moisture and Ocean Salinity (SMOS) satellite mission, operational since 2010, relies on an L-Band microwave interferometric radiometer to generate brightness temperature images along the swath, with global coverage every 3 days. These images are then used to derive sea surface salinity (SSS) with an effective resolution of less than 50 km. However, signal acquisition in some ocean regions is intermittently and significantly disrupted by radio-frequency interferences (RFI) from various terrestrial military or civilian sources worldwide. We develop a new methodology based on principal component and regression analyses to extract the RFI signatures in time and space, thereby enabling the construction of a corrected SSS estimate along the swath. This method successfully filters out many disruptive features characterized by long and wide branches occurring around the RFI sources, hence recovering SSS variability as demonstrated in comparison to in situ reference data. This correction methodology is an alternative to separate filtering procedures that were applied on brightness temperature at Level 1. Independent information indicating the probability of RFI occurrence on land areas or nearby is used to verify the timing of oceanic RFI contamination inferred by the correction process. The methodology performs particularly well in areas where the probability is close to 1 for a significant and contiguous portion of the entire period. Already applied with significant improvement in three selected regions, this correction method is a starting point for expanding and systematizing the methodology to treat as many RFI-polluted regions as possible and to recover SMOS SSS variability.