DE19847942A1 - Inserting security fingerprint into electronic documents by allocating point in hyperplane to each marking point in document - Google Patents

Abstract

A point in a projective space of hyperplanes is assigned to one marking point using a secrecy function. At least two hyper planes of the projected space are allocated to one fingerprint exclusively. For each fingerprint, the marking points corresponding to the selected hyperplanes are marked. To detect fingerprint erasure attempts, geometric objects are reconstructed form the marking points and the linear dependence is analyzed to determine in which hyperplanes the objects are contained.

Description

The invention relates to a method in the upper Concept of claim 1 specified type that is described in the main patent DE 198 16 356.8-31.

Because of the rapid growth of the Internet and the result opportunity to digitally distribute documents, there is an increasing need to protect against illegal Distribution of documents and thus to protect a copy rightholders from pirated copies.

Large companies like IBM, NEC and Microsoft, but also smaller companies such as Digimarc (see Funkschau 17/97; P. 21) and research institutes such as the Fraunhofer IGD and GMD Darmstadt Embed so-called digital watermarks in documents. In the process based on this information, that identifies the copyright owner, invisible in the documents to be protected. The type of turned brought digital watermark depends on the respective depending on the type of document (e.g. Postscript, JPEG, MPEG-1).

Digital watermarks allow the copyright owner to his intellectual property on an illegally distributed document evidence. Digital watermarks allow it however, not the culprit of the illegal distribution investigate and prove the illegal distribution.  

Digital fingerprints go one step further. At Using the principle of securing a document about digital fingerprints, is next to the digital water the copyright holder's name and name of the customer who has an electronic copy of the document acquires, invisibly inserted into the document. Distributed this customer now copies his copy contrary to the interests of Copyright holder, he can use his in all electronic copies of illegal copies clearly identified and responsible be. (Dan Boneh and James Shaw, Collusion-Secure Fingerprinting for digital data. Proc. CRYPTO "95" LNCS 963, Springer-Verlag, Berlin 1995, 452-465.)

The principle of securing documents by means of digital fingerprints has so far had a serious weakness, which results from the fact that the documents assigned to the individual customers differ in their bit pattern precisely at the point at which the individual user-related fingerprint of the different customers is arranged. If a customer comes into possession of a second customer's document, or if a group of attackers forms, they are able, by bit-wise comparison of the two documents, to locate the places with the different digital fingerprints and remove them from the document (Clear). With the method according to the main patent, according to FIG. 1, with the help of overlapping finite geometric structures, the intention of such manipulations to obtain a copy of the document from which all user-related fingerprints are removed is partly prevented because an intersection S of Fingerprints A; B and C is the same, is not discovered and is retained for investigation. However, such investigations are difficult for larger groups of attackers and can lead to ambiguous results.

The object of the invention is to remedy this deficiency, to allow more copies of the documents, and also one to recognize a larger group of attackers better.

The method proposed to solve this task steps are in the characterizing part of claim 1 leads.

In the characterizing part of patent claim 2 is an advantageous one Training opportunities for even better determination of Attempted deletion described.

The invention will now be described with reference to embodiments play described in more detail. In the accompanying drawings show the:

Fig. 1, the determination of the intersection S according to the main patent,

Fig. 2 shows a system for generating digital fingerprints in the projective level PG (2.2) and

Fig. 3 is a descriptive representation of the invention in three-dimensional space.

The invention is based on the use of hyperplanes in finite projective spaces PG (d, q). The parameter d specifies the maximum number of attackers that the system can process. The following two requirements are met:

• 1. Exactly one customer is assigned to each copy of the document net.
• 2.Connect and remove k customers with k <d + 1 all fingerprint parts from a document they can determine by (bitwise) comparison, so can these k customers from the manipulated document be clearly identified.

As an example, the invention is first described in the projective level PG (2.2). Fig. 2 gives this level and the different fingerprints A; B and C again.

In Fig. 2 all points and lines of the projective plane PG (2,2) are drawn. Each point on the projective level is assigned a place in the document using a secret function (or one-way function). Customer a then receives a document in which the positions are marked (e.g. by inverting the bits at the relevant positions) that belong to fingerprint A. For the customer to whom fingerprint A is assigned, this would be e.g. B. the positions assigned to points 1 , 5 and 6 .

If two customers compare their fingerprints, they will find all the marked positions, except for the one that is at the intersection of the two fingerprints. Would z. For example, if the customer uses fingerprint A and the customer uses fingerprint B to compare their fingerprints, they could find the locations that correspond to items 2 , 4 , 5, and 7 , but not locations 1 , 3, and 6 .

However, customers can still be identified if a maximum of two customers ally each other, because a large number of intersections can only be overlooked by exactly one pair of customers. In our example, points 1 , 3 , 6 can only be overlooked by customers a and b; Customer b and c would have overlooked points 1 , 2 , 7 , etc.

The system described above can be generalized for higher dimensions d and larger orders q. In the three-dimensional case, a cube (see Fig. 3) provides a vivid picture of the situation: a fingerprint consists of the points that lie on two opposite sides of the cube; Two fingerprints intersect exactly in four straight lines, and three each in exactly eight points. This means that from the intersection of two (or three) fingerprints, one can draw conclusions exactly about the two (or three) customers who tried to remove the fingerprints.

In the general case, the invention is based on the use of hyperplanes in the projective space PG (d, q). Each customer is assigned at least two hyper levels, there may be more. Each hyperplane may only be assigned to a customer once. There are different variants for this assignment:

• 1. The two hyperplanes are selected at random. At this variant is less likely in addition, the circle of suspects is larger than the traitor circle.
• 2. The two hyperplanes are "parallel", i. H. she snows in a selected hyperplane H.

Variant 2 serves to avoid the following problem:
The intersections of fingerprints A and B are discovered in a document. Then it is also conceivable that these intersections were generated by cutting fingerprints A and C, or by fingerprints B and C.

However, this problem can occur for large dimensions (from d = 3) are neglected: The intersection of d fingers pulling is only possible with a negligible probability ambiguity, and in this rare case recognizes the decoding algorithm the ambiguity. The danger that So innocent people are not suspected.  

For decoding, ie for the reconstruction of the hyperplanes from the intersections found (or intersection lines, planes,...) One searches for linearly dependent sets. Lying z. For example, in PG (d, q) only the 2 ^d intersection points of the d hyperplanes are present, one looks for subsets of 2 ^d-1 points each ^that lie in a common hyperplane. If you have found such a hyperplane, a fingerprint has already been identified.

Claims (2)

1. A method for introducing manipulation-proof digital fingerprints into electronic documents, in which each buyer is assigned a copy of a document in which an individual fingerprint assigned to the respective buyer of the copy and not identifiable by the buyer is inserted as a marker and in accordance with DE 198 16 356.8-53 the allocation of the marking positions of the individual digital fingerprints, which are different for each copy, is carried out with the help of finite geometric structures, whereby the copy of each buyer is marked at the locations determined by the geometric substructure assigned to the respective customer and a secret function , in which the points thus determined for each buyer as a digital fingerprint within the geometric structure are defined in such a way that they overlap with the points of the fingerprints of other buyers, and in which a bitwise comparison of the original chen document with a copy ent from the through manipulation of most d buyers parts of fingerprints were removed, pien co involved in the manipulation of the remaining intersections of fingerprints and therefore the active become an aggressor buyers can be determined, characterized that

• 1. each marking point in the document is uniquely assigned a point of the projective space PG (d, q) of hyperplanes by means of a secret function,
• 2. at least two hyperplanes of PG (d, q) are assigned to each fingerprint, each of which is exclusively assigned to one fingerprint, and that
• 3. For each fingerprint, the marking points that correspond to the points of the selected hyperplanes are marked.

2. The method according to claim 1, characterized in that for the determination of deletion attempts of the fingerprints

• 1. the geometric objects are reconstructed from the marked marking locations found in the document,
• 2. by analyzing the linear dependencies, it is determined in which hyperplanes the objects are contained, and
• 3. the fingerprints of the manipulating customers are successively determined by calculating these hyperplanes.