JPEG image is one of the most popular covers for steganography.The latest report on steganalysis showed that how to design secure JPEG steganographic method is still difficult.This paper presents a novel adaptive steganographic method for JPEG images.The difficulty of adaptive steganography is to design the distortion function that can indicate the embedding impact caused by changing quantized DCT（QDCT） coefficients.Because the magnitude of the absolue values of QDCT coefficients can imply most statistical properties

according to this

we design a new distortion function.Combined with STC（Syndrome Trellis Codes）

we present one-layered embedding method using nsF5’s changing mode and two-layered embedding method using ±1 changing mode respectively.The experimental results demonstrate that the proposed one-layered embedding method achieve stronger security than nsF5 and two-layered embedding method have better adaptability for various steganalyzers than nsF5 and Filler’s Method.