WO2009077944A1

WO2009077944A1 - Method for embedding and detecting a watermark

Info

Publication number: WO2009077944A1
Application number: PCT/IB2008/055222
Authority: WO
Inventors: Mehmet U. Celik; Aweke N. Lemma
Original assignee: Civolution Bv
Priority date: 2007-12-18
Filing date: 2008-12-11
Publication date: 2009-06-25
Also published as: TW200941281A

Abstract

A method of embedding a watermark comprising embedding a plurality of watermark parts in respective parts of an information signal. The method comprises embedding each part of the watermark based upon a value of a previous part of the watermark. This makes the embedded watermark more resilient against collusion attacks. The watermark may be a forensic tracking watermark.

Description

Method for embedding and detecting a watermark

FIELD OF THE INVENTION

The present invention relates to methods and apparatus for embedding a watermark in an information signal, and detecting a watermark embedded in an information signal.

BACKGROUND OF THE INVENTION

Computers are widely used in modern society. Many computers are connected together by way of the Internet, so as to allow data to be passed between computers in a convenient manner. In recent times, the distribution of media files (such as music and video files) over the Internet has become commonplace. While such distribution of media files is convenient for both media owners and consumers, it can be problematic in that illegitimate copies of media files are sometimes distributed in violation of relevant copyright laws. This can deprive a media owner of income to which it is properly entitled.

It is known to embed a watermark in a media file so as to monitor or control the distribution of the media file. For example, some watermarks are used so as to prevent copying of a media file or to limit playback of a media file to particular computers or devices.

Forensic tracking watermarks have been proposed as an effective deterrent to illegal distribution of media files. A unique forensic tracking watermark is embedded in each legitimate copy of a media file identifying a legitimate user of that media file. When an unauthorized copy of the media file is found, the watermark identifies the legitimate user of the media file who allegedly distributed the unauthorized copy. Said detectability of the user responsible for unauthorized distribution acts as a deterrent against unauthorized distribution.

The use of forensic tracking watermarks as described above means that there is a motivation for a user to mask their watermark within a media file, thereby making it difficult or impossible for the user to be identified as the source of unauthorized distribution of the media file.

One known method of attempting to evade the effect of a forensic tracking watermark is a collusion attack. Here a plurality of users (known as a coalition) use personalized copies of a media file to generate an averaged copy of that media file in which it is not possible to detect a watermark associated with any member of the coalition. The larger the number of members in the coalition, the more effective the collusion attack.

A collusion attack of the type described above has two effects on commonly used spread-spectrum forensic tracking watermarks. First, the power of each individual watermark in the averaged media file output from the collusion attack is weakened so as to make detection more difficult and so as to make the watermark less resilient to noise. Second, where a watermark payload is encoded in multiple parts of the media file, the detection of multiple values for each part of a payload causes confusion, possibly to the extent that watermarks of individual members of the coalition cannot be properly reconstructed. That is, even if values of the individual parts of the payload can be properly determined, the detection of multiple values for each part of the payload means that the detector cannot know which values of each part of the payload should be combined to produce complete payload values.

Various methods for providing collusion resilient watermarks have been suggested. In general, these methods are based upon encoding a watermark's payload. While such methods may be effective in offering resilience against collusion attacks involving coalitions having a large number of members they typically require a watermark to be embedded using long bit-strings (often thousands of bits).

Alternatively, it has been suggested that the use of a parity bit within a payload can mitigate the effect of a collusion attack involving a coalition of a small number of users (say two to five). However, even the use of a parity bit increases payload length and is for this reason undesirable.

SUMMARY OF THE INVENTION

It is an object of embodiments of the present invention to obviate or mitigate one or more of the problems set out above. It is a further object of embodiments of the present invention to provide a method for embedding and detecting a watermark in an information signal.

According to the present invention, there is provided a method of embedding a watermark comprising a plurality of parts in an information signal. The method comprises embedding a second part of said watermark into said information signal based upon a value of a first part of said watermark.

A second aspect of the present invention provides a method of detecting a watermark embedded in an information signal, the watermark comprising a plurality of parts. The method comprises detecting a second part of said watermark based upon a value of a first part of said watermark.

According to a further aspect of the present invention, there is provided a computer program for carrying out the method of the first aspect of the present invention. Such a computer program can be carried on an appropriate carrier medium. Such a carrier medium may be a tangible carrier medium such as a floppy disk, hard disc, CD or DVD or alternatively an intangible carrier medium such as a communications signal.

Further aspects of the invention provide apparatus for carrying out the method of the first aspect of the present invention, and apparatus for carrying out the method of the second aspect of the present invention.

The invention is widely applicable to any method in which a watermark comprising a plurality of parts is embedded in an information signal. The invention is particularly applicable to spread-spectrum watermarking, and offers particular benefits in forensic tracking applications. The individual parts of the watermark can be generated in any suitable way.

Where a watermark comprises a plurality of parts, embedding a second part of the watermark based upon a value of the first part of the watermark provides a number of benefits, particularly where the watermark is a forensic tracking watermark. Where the watermark is a forensic tracking watermark the method of the invention provides resilience against a collusion attack by a coalition of users. In any event, the method provides improved error detection capabilities at a detector by allowing erroneous detection of a value for a particular part of a watermark to be more easily detected when a subsequent part of the watermark is detected.

In one particularly preferred embodiment of the invention, a sequence is generated to represent the second part of the watermark. Generating the sequence comprises generating an initial sequence based upon the value of the first part of the watermark, and processing the initial sequence based upon the value of the second part of the watermark so as to generate the sequence. Processing the initial sequence may comprise circularly shifting the initial sequence. The overhead imposed by the invention is negligible in a practical system.

BRIEF DESCRIPTION OF THE DRAWINGS

Embodiments of the present invention will now be described, by way of example, with reference to the accompanying drawings, in which: Fig. 1 is a schematic illustration of a process for embedding a watermark in an information signal;

Fig. 2 is a schematic illustration of a watermark payload; Fig. 3 is a schematic illustration of a process for embedding a watermark comprising a plurality of parts in an information signal;

Fig. 4 is a schematic illustration of a process for embedding a watermark comprising a plurality of parts in an information signal in accordance with an embodiment of the invention;

Fig. 5 is a schematic illustration of a process for detecting a watermark embedded using the process of Fig. 4; and

Fig. 6 is a schematic illustration of an alternative method for generating watermark sequences suitable for use in the process of Figs. 4 and 5.

DESCRIPTION OF EMBODIMENTS Figure 1 shows a process for embedding a watermark W in an information signal S. The information signal can take any suitable form, but will often be a media file such as a music or video file. The watermark W is typically embedded in the information signal so as to prevent or control copying of the information signal, for example to ensure that copyright laws are not violated. In one preferred embodiment of the invention the watermark is a forensic tracking watermark used to identify a legitimate user of a particular copy of a media file.

Figure 1 shows that a seed is provided to a pseudo-random number generator 1 which generates a sequence of predetermined length based upon the input seed. The configuration and use of a pseudo-random number generator will be readily appreciated by those of ordinary skill in the art. The pseudo-random number generator 1 is configured such that a given seed will always generate a given sequence while different seeds will generate different sequences.

A sequence generated by the pseudo-random number generator 1 is input to a shift block 2 where it circularly shifted by an amount determined by a payload PL to generate a watermark sequence W. The watermark sequence W is then embedded into the information signal S by a watermark embedder 3. For example, the watermark sequence W takes the form of a noise pattern that is added to the host signal.

It is known to embed a watermark in an information signal by processing a payload in a plurality of parts so as to generate a watermark comprising a plurality of parts, each of which is embedded in a respective part of the information signal. For example, as shown in Figure 2 a payload 4 has a length of 32 bits and is processed in four parts 5 each having a length of 8 bits. A watermark embedded based upon the payload 4 will have four parts, each of which is based upon one of the parts 5 of the payload 4 and each of which is embedded in the information signal individually. Each part of the watermark may be embedded in a different spatial, time or frequency location of the information signal. Figure 3 shows a process for generating and embedding a watermark comprising a four parts. Four seeds are input to respective pseudo-random number generators Ia, Ib, Ic, Id resulting in the generation of four different sequences (e.g. noise patterns) which are provided to respective shift blocks 2a, 2b, 2c, 2d. The shift blocks 2a, 2b, 2c, 2d each receive as input a respective part 5 of the payload 4 denoted PLo, PLi, PL₂, PL3 respectively, and shift the received sequence based upon the received part of the payload so as to generate a watermark sequence. That is, different payload values are encoded by shifting a particular sequence in a manner determined by the payload. Each watermark sequence represents part of the watermark to be embedded and is provided to a respective one of four watermark embedders 3a, 3b, 3c, 3d. The information signal S is provided to each of the watermark embedders 3a, 3b, 3c, 3d in turn so that each of the watermark sequences is embedded in the information signal S.

Processing the payload 4 in a plurality of parts allows easier detection of a watermark at a detector. Specifically, the use of shorter sequences reduces the computational complexity of the detection process and is therefore preferred. Similarly, the encoding of different payload values by performing different circular shifts of a particular sequence also aids the detection process, and is therefore preferred.

Referring again to Figure 3, it can be seen that in addition to embedding a plurality of watermark sequences representing different parts of a watermark, a sequence generated by a further pseudo -random number generator 6 is embedded in the information signal S by an embedder 7. This sequence is embedded in the information signal S so as to provide a reference point for a watermark detection process.

A process for embedding a watermark in an information signal in accordance with an embodiment of the present invention is shown in Figure 4. The process of Figure 4 is based upon that of Figure 3 but the sequences generated to encode the parts of the payload are based upon a value of the payload of a previous part of the payload. Referring to Figure 4 it can be seen that the payload is embedded in four stages respectively marked A, B, C, D. It can be seen that the part of the watermark embedded in a first stage A is embedded in a manner identical to that described with reference to Figure 3. However, in each of stages B, C and D the respective pseudo-random number generators Ib, Ic, Id receive a seed value from a respective seed generator 8b, 8c, 8d. Each seed generator is arranged to generate a seed value for input to one of the pseudo-random number generators based upon an input seed value and also based upon a value of a payload received from an immediately preceding stage. That is, for example, the seed generator 8b generates a seed value for input to the pseudo-random number generator Ib based upon an input seed value and based upon a value of a payload PLo used in the first stage labeled A.

An arrangement for detecting a watermark embedded using the arrangement of Figure 4 is now described with reference to Figure 5. A detector inputs a seed value to a pseudo random number generator 9 which results in the generation of a sequence. This generated sequence is correlated with a received signal at a correlator 10 to locate the generated sequence within the received signal, thereby acting as a reference point for further processing. The signal containing the watermark is passed to each of four correlators 12a,

12b, 12c, 12d in turn, each correlator being arranged to detect one part of the watermark. Each correlator forms part of a respective one of the four stages of the detection process, labeled A', B', C and D' in Figure 5. referring to the first stage A', it can be seen that a seed is input to a pseudo-random number generator 1 Ia which generates a sequence. It is known that different payload values are encoded by carrying out different shifts of the generated sequence (as described with reference to the shift blocks 2a, 2b, 2c, 2d of Figure 4). The correlator 12a is therefore arranged to correlate shifted versions of the sequence generated by the pseudo-random number generator 11a with the received signal so as to identify a sequence embedded within the received signal. The sequence identified by the correlator is provided to a decoder 13a which generates a value for a first part of the payload.

The second stage B' of the detection process is analogous to the first stage A', but includes a seed generator 14b which generates a seed which is input to a pseudo -random number generator 1 Ib which is part of the second stage B'. The seed generator 14b takes as input an initial seed value and also the value PLo of the payload detected by the preceding stage. That is, while the pseudo-random number generators Ib, Ic, Id of Figure 4 takes as input a seed based upon a payload embedded in a preceding stage, in the detection process of Figure 5 each of the pseudo-random number generators 1 Ib, 1 Ic, 1 Id take as input a seed based upon a payload detected in a preceding stage. It has been described above that embodiments of the invention have particular application in forensic watermarking where a watermark is arranged to identify a legitimate user of an information signal. It has also been described that forensic watermarking schemes are sometimes vulnerable to attacks by a coalition whereby different watermarks are averaged together so as to remove the effect of the watermark.

The detection process described with reference to Figure 5 is resilient to such attacks. Specifically, where an information signal has been subject to an attack by a coalition, the correlation by the correlator 12a may generate two or more sequences, each of which can be decoded by the decoder 13a. Thereafter, by processing each of the decoded payload values in turn subsequent parts of the watermark can be properly processed by generation of an appropriate sequences for correlation by correlators of subsequent stages. This processing results in the detection of a plurality of watermarks, each watermark being associated with a particular member of the coalition responsible for the collusion attack.

The methods described above for embedding and detecting watermarks have been based upon generation of a sequence and appropriately shifting that sequence to encode a payload value. It will be appreciated that sequences representing different parts of a watermark can be generated in any convenient way. An alternative method for generating appropriate sequences is now described with reference to Figure 6.

Figure 6 shows two stages of a process for generating two parts of a watermark. These stages are labeled A and B and correspond to stages A and B of Figure 4. It will be understood that the processing at stages C and D can be analogous to that of stage B.

At stage A a payload value PL₀ is used to look up an appropriate sequence from a set of sequences 15. The sequence obtained from the set of sequences 15 is then used to represent a first part of the watermark. At stage B, a plurality of sets of sequences 16 is used, and one of the sets of sequences 16 is selected based upon the value of the payload PLo encoded in stage A. Having selected one of the sets of sequences 16 in this way, an appropriate sequence from that set of sequences is selected based upon a value of the payload PLi which is to be encoded in stage B.

Where a watermark is encoded using the embodiment of the invention shown in Figure 6, a watermark is detected by initially correlating all sequences of the set of sequences 15 with a provided signal. The results of this correlation is used to select one or more of the sets of sequences 16 which should be correlated to detect the second part of the watermark. It will be appreciated that the selection of one of the sets of sequences 16 (for both embedding and detection) based upon a payload value can be carried out in any convenient way. For example, a set of sequences may be selected based solely on the payload value embedded in a preceding stage. Alternatively a set of sequences may be selected by pre-processing a payload value embedded in a preceding stage so as to generate an appropriate value which is used to select a set of sequences.

Although particular methods for generating watermarks have been described above, it will be appreciated that any method arranged to generate a particular part of a watermark based upon a value of a preceding part of the watermark can be suitably used. Although preferred embodiments of the present invention have been described above, it will be appreciated that various modifications can be made to the described embodiments without departing from the scope of the appended claims.

In summary, a method of embedding a watermark comprising embedding a plurality of watermark parts in respective parts of an information signal is described. The method comprises embedding each part of the watermark based upon a value of a previous part of the watermark. This makes the embedded watermark more resilient against collusion attacks. The watermark may be a forensic tracking watermark.

Claims

CLAIMS:

1. A method of embedding a watermark comprising a plurality of parts in an information signal, the method comprising embedding a second part of said watermark into said information signal based upon a value of a first part of said watermark.

2. A method according to claim 1, wherein embedding said second part of said watermark comprises: generating a sequence representing said second part of said watermark based upon the value of the first part of said watermark; and embedding said sequence in said information signal.

3. A method according to claim 2, wherein generating said sequence representing said second part of said watermark comprises: generating an initial sequence; processing said initial sequence based upon a value of said second part of said watermark to generate said sequence.

4. A method according to claim 3, wherein processing said initial sequence comprises shifting said initial sequence.

5. A method according to claim 4, wherein generating said initial sequence comprises generating said initial sequence based upon the value of the first part of said watermark.

6. A method according to claim 5, wherein generating said initial sequence comprises providing an input to a generator and said input is based upon said value of said first part of said watermark.

7. A method according to claim 6, wherein said generator is a pseudo-random number generator, and said input is a seed value.

8. A method according to any preceding claim, wherein said watermark comprises an ordered plurality of parts, and each part is embedded in said information signal based upon a value of an immediately preceding part.

9. A method of detecting a watermark embedded in an information signal, the watermark comprising a plurality of parts, the method comprising detecting a second part of said watermark based upon a value of a first part of said watermark.

10. A method according to claim 9, wherein detecting said second part of said watermark comprises: generating a sequence; and processing said sequence to detect said second part of said watermark.

11. A method according to claim 10, wherein said processing said sequence comprises correlating said sequence with the information signal.

12. A method according to claim 10, wherein generating said sequence comprises generating said sequence based upon the value of the first part of said watermark.

13. A method according to claim 12, wherein generating said sequence comprises providing an input to a generator and said input is based upon said first part of said watermark.

14. A method according to claim 13, wherein said generator is a pseudo-random number generator, and said input is a seed value.

15. A method according to any one of claims 9 to 14, wherein said watermark comprises an ordered plurality of parts, and each part is detected in said information signal based upon a value of an immediately preceding part.

16. A computer program configured to control a computer to carry out a method according to any preceding claim.

17. A carrier medium carrying a computer program according to claim 16.

18. A computer apparatus comprising: a memory storing processor readable instructions; and - a processor configured to read and execute instructions stored in said memory; wherein said processor readable instructions comprise instructions controlling the computer to carry out a method according to any one of claims 1 to 15.

19. Apparatus for embedding a watermark comprising a plurality of parts in an information signal, the apparatus comprising a watermark embedder arranged to embed a second part of said watermark into said information signal based upon a value of a first part of said watermark.

20. Apparatus for detecting a watermark embedded in an information signal, the watermark comprising a plurality of parts, the apparatus comprising a watermark detector arranged to detect a second part of said watermark based upon a value of a first part of said watermark.