With the development of the Internet Era, the network attacks are on the rise. To some extent, the conditional independence assumption of Naive Bayes (NB) algorithm sacrifices the accuracy of classification, especially in dealing with complex network intrusion data. Aiming to solve this problem, this paper proposes a feature weighted JRNB intrusion detection algorithm. First, in order to remed for the deficiency of the equal analysis of all feature terms in Naive Bayesian algorithm, JS divergence method is introduced to measure the weight of each feature term to highlight the difference between different feature terms; Then, take into consideration of the impact of class frequency on sample classification, the reverse class frequency (RCF) is proposed to improve the calculation of feature weight and further reducing the impact of conditional independence. Compared with the traditional Naive Bayes algorithm and other popular classification algorithms, this algorithm in this paper has some improvement in detection performance.