The problem of single domain embedding and encoder dependence exists in almost all adaptive MP3 steganographic schemes, leading to relatively concentrated modification, limited embedding capacity, long execution time, and re-compression. Therefore, a fast and secure MP3 steganographic scheme with multi-domain is proposed in this paper. First, to enlarge the embedding capacity and spread the modifications, we extract the linbits and sign bits of quantized modified discrete cosine transform (QMDCT) coefficients to combine multi-domain. Second, a distortion function is designed based on the statistical properties of the inter and intra frames for better security. The message is embedded into the multi-domains by incorporating the syndrome trellis coding (STC) with the distortion function. Finally, the embedding method is able to get rid of the MP3 encoder and further avoid the problems of re-compression. Experiments illustrate that our scheme achieves a significant improvement in imperceptibility and statistical undetectability. In addition, the execution time is decreased by 63.6% $\sim$ 76.7%, and the embedding capacity is increased by at least 47.4%.

几乎所有的自适应 MP3 隐写方案都存在单域嵌入和编码器依赖的问题，导致修改相对集中，嵌入容量有限，执行时间长，需要重新压缩。因此，本文提出了一种快速、安全的多域 MP3 隐写方案。首先，为了扩大嵌入容量并传播修改，我们提取量化修改离散余弦变换 (QMDCT) 系数的 linbits 和符号位以组合多域。其次，为了更好的安全性，基于帧间和帧内的统计特性设计了失真函数。通过将校正子网格编码 (STC) 与失真函数相结合，将消息嵌入到多域中。最后，嵌入方法能够摆脱MP3编码器，进一步避免重新压缩的问题。实验表明，我们的方案在不可察觉性和统计不可检测性方面取得了显着改善。此外，执行时间减少了63.6%$～$ 76.7%，嵌入容量至少提高了47.4%。