Digital Watermarking

Errors are inevitable in even the best-designed watermarking systems. In this chapter, we discuss three types of error: false positive errors, false negative errors, and message errors. A false positive error occurs when the detector incorrectly indicates that a watermark is present. Conversely, a false negative error occurs when a detector incorrectly indicates the absence of a watermark. A message error occurs when a watermark detector incorrectly decodes a message.
The designer of a watermarking system must determine what error rates are acceptible during the specification phase of the design. It is therefore necessary to develop models for the errors of interest. The purpose of these models is twofold. First, a model allows us to select a detection threshold to meet the specifications. Second, experimental verification of the model allows us to be confident that the specified error rates will not be exceeded.
The severity of these errors depends on the application. For example, in a broadcast monitoring application to confirm that advertisements are aired, a false negative is very serious, in that it leads to the erroneous conclusion that an advertisement was not broadcast. This, in turn, may have serious adverse repercussions on the relationship between the broadcaster and the advertiser. In contrast, false positives lead to the conclusion that the advertisement is being aired more frequently than expected. This situation is also erroneous, but it does not result in mistrust or litigation between the broadcaster and the advertiser.
Message errors are considered in...