Nowadays, watermarking technology is widely applied to protect multimedia information, ensuring the security and integrity of digital content. Watermarked contents span a variety of media types, including images, audio, video, graphics, text, and other kinds of media. The process involves embedding watermarks into the media signals, which can be accomplished by adding them to specific locations either in the sample domain or within a suitably transformed domain. In this paper, we propose an innovative invisible audio watermarking technology that leverages algebraic transforms. This method ensures that the watermark is imperceptible to human senses while maintaining robustness against various signal transformations and compression. The core of our approach is a deep neural network architecture, which is a modified version of the Resdep and ResIndep deep architectures. These modifications are tailored to enhance the watermarking process. The key feature of the proposed architecture is its unique approach to handling audio signals: it completely bypasses the reliance on spectral characteristics and direct transformations in the time domain. Instead, the architecture focuses on embedding the watermark in a manner that maximizes resilience against audio signal processing techniques like compression, noise addition, and filtering. The proposed method’s advantage lies in its ability to maintain the integrity of the watermark under various conditions that typically degrade the quality of the audio signal, such as MP3 compression. By rejecting traditional spectral analysis methods and emphasizing algebraic transformations within the time domain, this technology sets a new standard for robust, invisible audio watermarking. This approach not only enhances security but also ensures that the quality of the original audio content remains uncompromised.