A TUTORIAL ON DIGITAL WATERMARKING Fernando P´ erez-Gonz´ alez and Juan R. Hern´ andez Dept. Tecnolog´ıas de las Comunicaciones, ETSI Telecom., Universidad de Vigo, 36200 Vigo, Spain email: fperez@tsc.uvigo.es, jhernan@tsc.uvigo.es ABSTRACT This paper gives a tutorial on the techniques and reference models used in digital watermark- ing. Distorsions, attacks and applications are de- scribed in some detail. Finally, the need for bench- marking is discussed. 1. INTRODUCTION In his wonderful book The Codebreakers [1], D. Kahn recounts one of the stories in the Histories of Herodotus in which Histiaeus tatooed a message in the shaven head of a slave and waited for the new hair to grow before sending him to Aristago- ras at Miletus with instructions to shave –again– the slave’s head. Obviously, bandwidth was not a concern at those times, but methods haven’t changed so much when compared with the state of the art in digital watermarking. Since the publication of a seminal work by Tanaka et al. in 1990 [2], we have witnessed an extraordinary growth of techniques for copyright protection of different types of data, especially mul- timedia information. This interest is not surpris- ing in view of the simplicity of digital copying and dissemination: digital copies can be made identi- cal to the original and later reused or even ma- nipulated. Cryptography is an effective solution to the distribution problem, but in most instances has to be tied to specialized –and costly– hard- ware to create tamper-proof devices that avoid di- rect access to data in digital format (even so, there exist software/hardware tools that allow to resam- ple the analog output of the device with decent results). Moreover, most cryptographic protocols Work partially funded by CICYT under project TIC-96- 0500-C10-10 are concerned with secured communications in- stead of ulterior copyright infringements. For in- stance, access control in set-top-boxes used for digital television demodulation and decoding suc- ceed in avoiding unathorized access to programs that are being broadcast in scrambled form [3] but fail in precluding further storage and illegal dis- semination actions. There is then an increasing need for software (or in the worst case, hardware) that allows for protection of ownership rights, and it is in this context where watermarking techniques come to our help. Perceptible marks of ownership or au- thenticity have been around for centuries in the form of stamps, seals, signatures or classical wa- termarks, nevertheless, given current data manip- ulation technologies, imperceptible digital water- marks are mandatory in most applications. A digi- tal watermark is a distinguishing piece of informa- tion that is adhered to the data that it is intended to protect, this meaning that it should be very dif- ficult to extract or remove the watermark from the watermarked object. Since watermarking can be applied to various types of data, the imperceptibil- ity constraint will take different forms, depending on the properties of the recipient (i.e., the human senses in most practical cases). In addition to imperceptibility there are some desirable characteristics that a watermark should possess, which are somewhat related to the so- called robustness issue. First, the watermark should be resilient to standard manipulations of uninten- tional as well as intentional nature. Second, it should be statistically unremovable, that is, a sta- tistical analysis should not produce any advantage from the attacking point of view. Finally, the wa- termark should withstand multiple watermarking to facilitate traitor tracing, as discussed in Section