A method and system for embedding an authentication signature in an audio-visual signal such that only a part of an entire frame of the audio-visual signal is stored in a memory while the signature bits are calculated and the watermark is embedded. A signature is formed based on a first portion of said audio-visual signal, whereby said first portion is a pattern of horizontal lines of said audio-visual signal and has fewer lines than the number of lines of the entire audio-visual signal. Thereafter the signature generated is embedded in said audio-visual signal in the first portion and/or in another portion of the frame to be authenticated, whereby the other portion also is a pattern of horizontal lines. Thus only memory for some lines instead for all lines of the audio-visual signal is needed. In the case of an interlaced audio-visual signal, the first portion is preferably the first field of a frame of said audio-visual signal and the second portion is the second field of said audio-visual signal. In the case of a non-interlaced audio-visual signal, slices of said lines are preferably used for said portions.