A number of data items are selected for inclusion in a data set so as to discourage a transmission of the entire set over a limited bandwidth communications path, such as the Internet. Each data item comprises one or more sections, and the totality of sections constitute the complete data set. Each section of the data set contains a watermark that includes an identifier of the section, and an identifier of the data set. In a preferred embodiment, the identifier of the section is the address of the section, and the identifier of the data set is a serial number and an indicator of the total size of the data set. The presence of the data set is confirmed by checking the watermarks of randomly selected sections to verify that the original section that formed the data set is present. If a section is discovered to be missing or altered, subsequent processing of data items of the data set is prevented. In a preferred embodiment, the identifiers are stored as a combination of robust and fragile watermarks.