Ponencia presentada en la Séptima Reunión Española sobre Criptología y Seguridad de la Información, celebrada en septiembre de 2002 en Oviedo, España.
Resumen: We consider private watermarking systems with additive noise attack in which watermarks (WM) are embedded as additive component into host data. Our model allows to upper bound the probabilities of WM-missing and false alarm of WM detection. Three main cases are investigated: when host data is known by the WM decoder but it is not used by the WM encoder, CM is unknown by the decoder and it is not used by the WM encoder and CM is unknown by the decoder but it is used by the WM encoder. We can find out how many WM bits are necessary to embed reliable WM for different conditions.