Flood disasters are one of the most destructive hazards faced across the globe, and thus flood hazard analysis is an essential factor for flood management. Due to the randomness and fuzziness of flood hazard, a series of possibilities is required for a certain flood level rather than just one exact probability. As a method for fuzzy hazard analysis derived from the information diffusion theory, the Interior-Outer-Set Model (IOSM) has the potential to reflect the randomness and fuzziness of flood hazard, however, the controlling intervals might have no sample points particularly when there are extraordinarily large flood peak flows or the flood samples concentrate. Based on this, the current study proposes a new framework for flood hazard analysis. First, flood samples are extracted from daily observed peak flow data. Second, the traditional IOSM is improved using design peak flows from flood frequency analysis (FFA) as the controlling points. From this, probability and floods hazard values are estimated via the FFA based IOSM (FFA-IOSM).The proposed framework is applied using data from the Dongjiang River, South China. Results demonstrate that the estimated flood probability was able to more effectively reflect the randomness and fuzziness of flood hazard compared to the traditional IOSM. This study provides a basis for reasonable flood engineering practices and supports the government with effective guidance on flood risk management, particularly under the increasing frequencies of the extreme precipitation events.