In practice, most audio files such as MP3 and AAC are stored and transmitted in the form of compressed files, which can serve as the cover in audio steganography. Currently, the prevailing audio steganography methods are not ideal because of the drawbacks. Some are characterized by low capacity while others are irreversible. In this paper, we propose a method to embed secret messages in MP3 encoding. Our strategy is to hide the information by Huffman table transformation. We extract secret information by analyzing side information. Experimental results show that our method can greatly improve the steganographic capacity with low distortion and high security. Meanwhile, it featured with higher decoding rate and reversible over some state-of-the-art methods.