Digital Watermarking Definition

Digital Watermarking is a technique that hides information and part of the known as steganographic. (steganography is the branch of cryptology dealing concealing messages, to avoid the perception of the existence of the same).

  • Its main objective is to highlight the illicit use of a particular digital service by an unauthorized user.
    Precisely, this technique consists of inserting a message (hidden or not) inside a digital object, such as images, audio, video, text, software, etc.
  • Said message is a group of bits that contains information about the author or intellectual owner of the digital object treated (copyright).

Another technique related to this is “fingerprinting” or “fingerprint,” where it introduces a trademark into the digital object that, in addition to containing information on copyright, includes information on the user who has acquired the right to use that object. In this way, the illegal distribution of digital services can be pursued.

Features

A digital watermarking technique usually requires to be:

  • Imperceptible: invisible to the observer.
    Do not degrade the object.
  • Robust: the removal or reduction of the brand must be difficult or ideally impossible without degrading the quality of the digital object.
  • Likewise, it must support usual transformation processes (compression, filtering, format conversion, geometric distortion…)
  • It must not be ambiguous: the brand must unequivocally identify the intellectual owner so that he can claim his membership.
  • Despite these premises, some watermarks are noticeable and that degrade the object. Such as, for example, watermarks superimposed on images indicating their ownership.

Applications of Digital Watermarking

The first applications of watermarks were those intended for the identification of the author and therefore, the requirements that the marking methods had to meet were the following:

  • It must be invisible to human perception and must not affect the quality of the digital material.
  • The brand recovered must unambiguously identify the owner.
  • Statistical tests should not detect it.
  • It must be challenging to eliminate; that is, the brand must be robust to the different signal processing algorithms such as filters, compressions, rescaling, etc.
  • It must be hard to fake.