The main objective of this study is to estimate nuclear data uncertainties on the effective multiplication factor (Keff) related to elastic, inelastic, capture and fission cross sections and the correlations between them. Different rapid and thermal cases of the different IHECSBE benchmarks have been studied by using nuclear data evaluation JENDL4 to calculate the sensitivity vectors for 235U and 238U isotopes and four cases used to validate our sensitivity vectors. These sensitivity vectors are calculated by using the adjoint-weighted perturbation method based on the Kpert card of the Monte Carlo code MCNP6. Thus, the uncertainties induced by nuclear data have been calculated by combining the sensitivity vectors with the covariance matrices that are generated by the ERRORJ module of the recently updated of the nuclear data processing system NJOY99. In this study, we found the four cross sections (elastic, inelastic, capture and fission) of the 235U and their covariance matrices Lack the adjustement especially in the rapid energies. Against, the cross sections and covariance matrices of the 238U not lack the adjustement especially in the thermal energies.