Many applications require watermarks to be detected in Works that may have been altered after embedding. Watermarks designed to survive legitimate and everyday usage of content are referred to as robust watermarks. In this chapter, we present several general methods for making watermarks robust and discuss specific methods for handling some of the most common types of processing.

We draw a distinction between robust watermarks and secure watermarks. ^[1] Whereas robust watermarks are designed to survive normal processing, secure watermarks are designed to resist any attempt by an adversary to thwart their intended purpose. Because in most applications a watermark cannot perform its function if it is rendered undetectable, robustness is a necessary property if a watermark is to be secure. In other words, if a watermark can be removed by application of normal processes, it cannot be considered secure. However, robustness is not sufficient for security, because secure watermarks must also be capable of surviving novel processes that are specifically designed to remove them. Thus, the designer of a secure watermark must consider the range of all possible attacks. The designer of a robust watermark can limit his attention to the range of probable processing. Secure watermarks are discussed in Chapter 9.

In designing a robust watermark it is important to identify the specific processes that are likely to occur between embedding and detection. Examples of processes a watermark might need to survive include lossy compression, digital-to-analog-to-digital conversion, analog recording (such as...