Traditionally, kernel learning methods requires positive definitiveness on the kernel, which is too strict and excludes many sophisticated similarities, that are indefinite, in multimedia area. To utilize those indefinite kernels, indefinite learning methods are of great interests. This paper aims at the extension of the logistic regression from positive semi-definite kernels to indefinite kernels. The model, called indefinite kernel logistic regression (IKLR), keeps consistency to the regular KLR in formulation but it essentially becomes non-convex. Thanks to the positive decomposition of an indefinite matrix, IKLR can be transformed into a difference of two convex models, which follows the use of concave-convex procedure. Moreover, we employ an inexact solving scheme to speed up the sub-problem and develop a concave-inexact-convex procedure (CCICP) algorithm with theoretical convergence analysis. Systematical experiments on multi-modal datasets demonstrate the superiority of the proposed IKLR method over kernel logistic regression with positive definite kernels and other state-of-the-art indefinite learning based algorithms.