Secret key watermarking with changing keys

We consider a digital watermarking application where multiple parties can embed additional information using their watermark embedder. These parties are not supposed to influence each other and each watermark detector needs to be able to decode the information embedded by any of the embedder systems. One approach would be to use a single secret key and to assign part of the payload to identify the particular embedder. However, it is generally accepted that for security reasons, each embedder should better have its own secret key. A major drawback of this last approach is related to the detector implementation complexity, which increases linearly with the number of embedders. It is shown that this drawback can be overcome by changing the key in the watermarking system dependant on features of the incoming signal