The audio watermarking scheme recently proposed by Fallahpour in [1] is one of the schemes with highest payload published to date. In this letter a key-based security improvement is proposed for that scheme. It is achieved by adding a Pseudo-Random Number Sequence (PRNS) in the frequency domain to the data samples, before applying the insertion algorithm. Experimental results show that the proposed enhancement keeps the perceptual transparency and the robustness to all attacks originally reported by Fallahpour’s scheme and the payload is not significantly affected.