In this paper, a new blind steganalytic scheme is proposed to effectively detect JPEG steganography. In DCT domain, the co-occurrence matrices are used to capture the intra-block and inter-block correlations among the quantized coefficients. In spatial domain, we estimate the invariant components of the decompressed raw pixels between the cover and stego images by wiping selected intermediate frequency AC coefficients, and collect eighteen statistics in total. Combining the DCT and spatial features leads to be more effective for classification. The experimental results have demonstrated that the proposed method is stable, and outperforms the recently reported steganalysis in attacking the advanced JPEG steganography.