RU2458395C2 - Methods and apparatus for ensuring integrity and authenticity of documents - Google Patents

Abstract

FIELD: information technology.

SUBSTANCE: apparatus and method determine physical characteristics of cells of the same configuration in accordance with given printing conditions, wherein the ratio of cells printed with printing errors, arising exclusively due to unforeseen variables, is greater than a given first value and less than a given second value; an information item is provided by changing the external appearance of the cells representing said physical characteristics; said configuration is printed using said printing conditions, wherein said configuration is made while allowing recognition of a copy which modifies the external appearance of the set of said cells.

EFFECT: design of an information matrix for recognising copies or forged documents.

31 cl, 30 dwg

Description

The present invention relates to a method and apparatus for the protection and authentication of documents, in particular to the recognition of copies of documents, packaging, finished products, molded products and cards, such as identification or banking. The term "document" means any medium.

The visual representation of surface information that can be read by a machine is a barcode. Initially, barcodes represented information in the form of the width of parallel lines and the width of the gaps between the lines, which limited the amount of information per unit area. Therefore, these barcodes are called one-dimensional, or “1D,” barcodes. To increase the amount of information, barcodes evolved and turned into patterns of concentric circles or dots.

Barcodes are widely used to quickly and reliably automatically identify the captured image through automatic processing.

Barcodes can be read by portable optical readers or scanners equipped with appropriate software.

Two-dimensional matrix barcodes, called 2D barcodes, are information carriers, which for the most part consist of square elements located inside a given perimeter, with each element or each cell taking one of two predetermined colors, for example black or white in accordance with the value of the binary character described in this cell. In addition, a two-dimensional barcode allows you to present on the same area a much larger amount of information than a one-dimensional one.

Thus, a two-dimensional barcode is preferable to a one-dimensional one, although reading systems for it are more complex and expensive, and in most cases they provide less flexible reading relative to the relative position of the reader and the barcode.

Two-dimensional barcodes are widely used for storing or transmitting information on passive objects, such as paper, ID cards, stickers, metal, glass or plastic. The system that generates two-dimensional barcodes, as input, usually receives a sequence of characters in a given alphabet, for example, a 128-character ASCII alphabet, a 36-character alphanumeric alphabet, or a binary alphabet.

At the output, the system produces a digital image, which is then printed onto an object, in accordance with the present invention, called a “document”. To read barcodes and reconstruct the information contained in two-dimensional barcodes, a system is usually used that captures an image and is connected to a processor.

The barcode, one-dimensional or two-dimensional, is used to transmit information from the sender to the recipient. In many applications, this method of transmitting information must be performed in a safe manner, which entails ensuring the following protection properties:

1) maintaining the confidentiality of the transmitted message (it is undesirable to read it by third parties), 2) the authenticity of the message (confidence in its origin), 3) the integrity of the message can be checked (confidence that the message has not been modified or tampered with), 4) the sender cannot refuse from the message (it is desirable to avoid a situation in which the author of the transmitted information denies sending it). These various degrees of information security can be achieved by encrypting or encoding the message with an encryption key known only to authorized persons or organizations. To obtain several of the above information protection properties, encryption methods with a public key or a private key are usually combined.

When encrypting a message, a 2D barcode provides a physical document with security features that were originally designed to protect digital information and digital documents. Thus, a 2D barcode helps to avoid falsification of documents or detect it. For example, text information printed on a document in an unencrypted form can be easily changed, for example, the expiration date or expiration date of a document, or personal identification card data, while the same data encoded in a 2D barcode cannot be changed easily and in accordance with the change of text information. Thus, a 2D barcode provides the ability to detect changes in text information.

In addition, a 2D barcode can be used to track documents. The origin, purpose and / or route of distribution of the document can be encrypted in a 2D barcode printed on this document, which allows you to make sure that the document is in a legal place according to the route of its distribution. Encryption of the specified information in these cases is necessary, because otherwise the information may be falsified or may not even have any relation to the true one.

Due to the use of barcodes, digital cryptographic methods can be used to protect analog documents (representing real objects) and passive documents (not able to respond to a signal). Thus, the security properties equivalent to the security properties of digital information or digital documents can be given to said documents.

However, 2D barcodes do not provide protection against accurate, so-called blind copying. Typically, each cell of a 2D barcode can be identified and read with great accuracy, as a result of which an exact copy of each barcode can be easily made. Therefore, a two-dimensional barcode does not provide a complete solution to the main problem of recognizing the authenticity of the source (origin) of a document, since an encrypted 2D barcode does not allow to answer the question whether the document containing it is an original or its reproduction.

In addition, intellectual property owners, especially trademarks, and organizations issuing official documents that have implemented encrypted 2D barcodes or other storage media, such as RFID electronic labels (short for Radio Frequency Identification, to solve problems related to fakes) radio frequency identification), however, to avoid or detect blind copying, they are forced to use completely different authentication methods (“authenticators”), for example, holograms that protect colors, microtexts or the so-called "guilloche" patterns (thin curved lines interfering with digital playback systems, for example due to the watermark effect).

However, these tools have their own limitations, which become more noticeable with the acceleration of the spread of technology that allows manufacturers of fakes to better and faster copy the specified authenticators. Thus, holograms are being copied by manufacturers of fakes better and better, and end users have neither the ability nor the motivation to check them. Protective paints, the so-called “guilloche” patterns and microtexts are unprofitable, difficult to integrate into the production processes of companies or information channels, and also do not provide the level of protection required in most cases. In addition, they can be difficult to identify and do not provide real guarantees of protection from manufacturers of fakes.

To determine the authenticity of the document, the read information is used, if possible, in combination with a database. Thus, you can, for example, indirectly detect a fake if another document containing exactly the same information was found earlier or in another place. It should be noted that in this case it is assumed that a separate document contains unique information, which is not always feasible by all means of producing a document, especially by offset printing. However, the implementation of such solutions is expensive, and quick access to the database may not be possible, especially if the reader is portable. Finally, even access to the database does not solve the problem of recognizing which of two seemingly identical documents is a fake.

Patterns for copy recognition are a type of visible authentication patterns that usually form a kind of background and which are created using the key in a pseudo-random way. These patterns are mainly used to recognize original printed documents and printed documents copied from them, for example by photocopying or using a scanner and printer. For this, the captured analog image, i.e. a real, pattern for recognizing a copy with the original digital representation of this pattern to measure the degree of discrepancy between them. Due to image distortion during copying, a higher degree of divergence of such a captured image of a pattern that was not obtained using the original analogue pattern.

In order for the pseudo-random image to contain information, it is cut into blocks, and the pixel colors of each block representing one of the binary values are inverted, leaving the pixels of each block representing the other binary value unchanged. In addition, other encoding of a binary value of a block may be used. In practice, the blocks should be large enough so that the specified binary value is reliably read, so that the amount of information that the image carries is limited.

However, this technology has disadvantages. In particular, it is optimal for copy recognition, but does not allow carrying a large amount of information on a given area, while in many areas of application documents carry a large amount of protected information and at the same time have great limitations (aesthetics, available area, brand image, etc. .d.), limiting the area available for recognition of copies. Using this technology requires comparing two images and scaling the captured pattern, which requires a large amount. In addition, this scaling can lead to distortion of the modified image, which under certain circumstances can lead to a limitation of the ability to recognize copies. In addition, the reader must reproduce and store in memory a pattern for recognizing a copy at the stage of image comparison, which is simultaneously an expensive and potentially dangerous operation, since an attacker can “read” the memory, which may allow him to accurately reproduce the pattern for recognizing a copy.

The aim of the present invention is to eliminate both the disadvantages of 2D barcodes and the disadvantages of patterns for recognizing copies. In particular, an object of the present invention is to provide means and a method for creating an information matrix providing recognition of copies or forged documents.

The first aspect of the present invention is the creation of a method of protecting a document containing:

- the stage of determining the printing conditions of the specified document,

- the step of determining the physical characteristics of the cells of at least one configuration in accordance with the specified printing conditions, for example, the percentage of cells printed with a printing error that occurs solely due to unforeseen variables is greater than the predetermined first value and less than the predetermined second value,

- a step of presenting a unit of information by changing the appearance of cells representing the indicated physical characteristics, and

- the stage of printing the specified configuration using the specified printing conditions, and the specified configuration is made with the possibility of recognizing a copy that modifies the appearance of many of these cells.

In the present description, “printing error” is called a change in the appearance of a cell that modifies the presentation of information that this cell carries when analyzing to avoid reading or capturing errors, for example, under a microscope. It should be noted that if initially the cells often have binary values, then the captured images often contain shades of gray, and therefore, a value is associated with the cell that is not binary and that can be interpreted, for example, as the probability that the cell will have the original binary value.

The authors of the present invention have found that if the proportion of printing errors exceeds a predetermined value, copying a configuration made using the same printing means as the original print, or similar means, necessarily causes the appearance of an additional fraction of errors that makes it possible to recognize the specified copy.

In addition, the authors of the present invention found that, depending on the specified restrictions (for example, limiting the number or physical size of cells of the protected information matrix), there is a proportion of printing errors that is optimal to enable copy recognition. The indicated optimum percentage of printing errors corresponds to the specified cell size or print resolution and depends on the printing media.

Thus, contrary to popular belief, the highest print resolution is not necessary, and even rarely, is the resolution that provides the best result to enable copy recognition.

In this case, it is necessary to distinguish between the native resolution of the printing means and the resolution of printing cells, each of which usually consists of many dots of printing ink, each of which corresponds to its own resolution of printing. Obviously, the print resolution of the protected information matrix cannot vary. In fact, most print media print in binary code (a printing ink point is present or absent) with a fixed resolution, gray levels or other colors are reproduced using various screening methods. In the case of offset printing, the indicated “native" resolution is determined by the resolution of the printing form, which is, for example, 2400 dpi (2400 dpi). Thus, a grayscale image for printing at a resolution of 300 pixels per inch (300 ppi) is, in practice, printed in binary code at a resolution of 2400 dpi (2400 dpi), with each pixel representing approximately 8 × 8 dots.

Despite the fact that, as a rule, the print resolution cannot be changed, on the other hand, you can change the cell size of the protected information matrix in pixels so that one cell is represented by several printing points. Thus, you can, for example, imagine a cell as a square block of 1 × 1, 2 × 2, 3 × 3, 4 × 4 or 5 × 5 pixels in size (non-square shape blocks are also possible), corresponding to a resolution of 2400, 1200, 800, 600 and 480 cells per inch, respectively.

According to the distinguishing features of the invention, at the step of determining the physical characteristics of the cells, the sizes of the cells to be printed are determined.

According to the distinguishing features of the invention, at the stage of determining the physical characteristics of the cells, a part of the cell is determined that has a constant shape and a changing color to represent different values of a unit of information, said part being certainly smaller than the specified cell.

According to the distinguishing features of the invention, the first predetermined value is more than 5%.

According to the distinguishing features of the invention, the first predetermined value is more than 10%.

According to the distinguishing features of the invention, the first predetermined value is more than 15%.

According to the distinguishing features of the invention, the first predetermined value is more than 20%.

According to the features of the invention, the second predetermined value is less than 25%.

According to the features of the invention, the second predetermined value is less than 30%.

According to the distinguishing features of the invention, at the printing stage, the native resolution of the printing means performing said printing is used.

According to the distinguishing features of the invention, a document security method, a brief description of which is given above, further comprises the step of creating a configuration in a digital information matrix representing a message containing redundancy.

In fact, the authors of the present invention found that any copy or any print of a unit of a printed information matrix having a small size reproduces the number of errors increasing with increasing print quality, and the insertion of redundant information, for example, error correction codes, into the matrix makes it possible to determine the document is a copy or original: inserting redundant information allows you to read the message on a noisy channel and / or measure the number of errors of the encrypted message, thus By providing an opportunity to determine whether this document is a copy or an original.

It should be noted that the distortions that occur during printing or copying depend on many factors, for example, print quality, resolution of the substrate and the image used during image capture or at the marking step for the purpose of making a copy.

According to the distinguishing features of the invention, at the stage of creating the configuration, there is a sufficient proportion of redundancy that allows you to adjust the proportion of errors greater than the specified first set value.

According to the distinguishing features of the invention, at the stage of creating the configuration, said redundancy contains error correction codes.

Thanks to these tools, the contents of the label makes it possible to correct errors at the stage of labeling and reproduce the original message.

According to the distinguishing features of the invention, at the stage of creating the configuration, said redundancy contains error correction codes.

Thanks to each of these tools, the number of errors that distort the label can be determined and used as a basis for recognizing a copy of the specified label.

According to the distinguishing features of the invention, at the stage of generating the information matrix, said information matrix at the level of each unit cell and independently of neighboring unit cells represents a message containing redundancy.

Thus, the amount of information that the label carries increases in relation to the representation of values by blocks of points.

According to the distinguishing features of the invention, at least five percent of unrelated errors are made at the time of labeling, and the use of redundancy allows them to be calculated.

The authors of the present invention found that it is easier to use a high error rate made at the stage of applying labels to distinguish the copy from the label, and the error rate of the copy depends on the error rate of the original label.

According to the distinguishing features of the invention, at the stage of generating the information matrix, redundancy is created to enable recognition in the label of unrelated errors made at the stage of labeling.

According to the distinguishing features of the invention, at the stage of labeling, an additional reliable label carrying the message is added to the label of the information matrix.

Thanks to these means, a message that carries an additional mark is more resistant to distortions caused by copying, and therefore will be read even if the indicated distortions are significant, for example after several consecutive copies. According to the distinguishing features of the invention, at the stage of generating the information matrix, the presentation of the specified message is encrypted using an encryption key.

According to the distinguishing features of the invention, at the stage of generating the information matrix, the presentation of the specified message is encoded to create the specified redundancy.

According to the distinguishing features of the invention, at the stage of generating the information matrix, the representation of the specified message is reproduced in order to create several identical copies.

Thus, in a very simple way, redundancy is created that provides error detection during the label reading process.

According to the distinguishing features of the invention, at the stage of generating the information matrix, the positions of the presentation elements of the specified message are interchanged in accordance with the secret key.

According to the distinguishing features of the invention, at the stage of generating the information matrix, the positions of the presentation elements of the specified message are partially interchanged in accordance with a secret key that differs from the key of the first permutation.

According to the distinguishing features of the invention, at the stage of generating the information matrix, a value substitution function depending, on the one hand, on the value of the element and, on the other hand, on the value of the element of the secret key, is applied to at least one part of the presentation elements of the specified message.

According to the distinguishing features of the invention, at the stage of generating the information matrix, the function of partial replacement of values, which depends, on the one hand, on the value of the element and, on the other hand, on the value of the element of the secret key, which differs from the secret key of the first replacement function, is applied at least to one part of the presentation elements of the specified message.

According to the distinguishing features of the invention, said substitution function replaces the values associated with neighboring cells in the specified configuration in pairs.

Thanks to each of these means, the specified message is provided by reading protection properties by unauthorized third parties.

According to the distinguishing features of the invention, at the stage of generating the information matrix, at least one key is used, and this is not the associated key that is necessary for reproducing the information, but another key.

Thus, the key to determine the authenticity of a document or product containing a label representing the specified information matrix cannot be used to create another information matrix containing another message.

According to the distinguishing features of the invention, at the stage of generating the information matrix, a digital information matrix is generated representing at least two messages equipped with various security features.

Thanks to these means, different people or computer systems can have different powers and means of reading, for example, to distinguish between authentication functions and determine the origin of fakes.

According to the distinguishing features of the invention, one of these messages is the information required when reading the information matrix to determine another message and / or detect errors in it.

According to the features of the invention, one of these messages is at least one key required to read another message.

According to the distinguishing features of the invention, at the stage of generating the information matrix, its hash value is added to the representation of the specified message.

A second aspect of the present invention provides a device for securing a document, comprising:

- means for determining the printing conditions of the specified document,

- means for determining the physical characteristics of the cells of at least one configuration in accordance with the specified printing conditions, for example, the percentage of cells printed with a printing error that occurs solely due to unforeseen variables is greater than the predetermined first value and less than the predetermined second value,

- means for representing a unit of information by changing the appearance of cells representing the indicated physical characteristics, and

- means of printing the specified configuration using the specified printing conditions, and the specified configuration is made with the possibility of recognizing a copy that modifies the appearance of many of these cells.

Since the advantages, objectives and features of the specified device, the creation of which is proposed by the second aspect of the present invention, are the same as for the method, the creation of which is proposed by the first aspect of the present invention, they are not repeated here. A third aspect of the present invention is the creation of a computer program containing instructions that can be read by a computer, and the implementation of the method, a brief description of which is given above.

A fourth aspect of the present invention is the creation of a computer readable medium containing computer readable instructions and the use of the method, a brief description of which is given above.

In addition, the present invention relates to a method and apparatus for protecting documents and products based on improved secure information matrices, in particular to the identification and authentication of documents and products. The invention is used, in particular, for unambiguous identification, authentication of originals and recognition of copies of documents, packaging, industrial products, cast products and cards, identification or banking.

There are many ways to protect documents using either expensive means (holograms, security inks, etc.), since they require consumables, or digital means, which in most cases are more economical. Digital tools provide an additional advantage, as they are well adapted to digital data processing. Thus, not very expensive detectors can be used, containing mainly a processor connected to an device for capturing images or signals (a scanner, etc.) and an interface with an operator.

To protect a document with digital means, you can turn to the use of digital authentication codes. For example, you can print on it a protected information matrix or pattern for copy recognition. In addition, digital authentication codes provide the ability to contain encrypted information and, thus, track the movement of documents or products.

A digital authentication code is a digital image that, firstly, is printed on a document, and secondly, provides the ability to track it, and at the same time recognize a copy of the document. Unlike a two-dimensional barcode, which is simply a storage medium that can be copied identically, any copy of digital authentication codes entails its distortion. The specified distortion can be measured by computer means from the captured image and allows the reader to determine whether the digital authentication code is the original or a copy. Moreover, the information contained in the digital authentication code is typically encrypted and / or scrambled.

Digital authentication codes can be invisible or at least visually difficult to see, such as, for example, a copy-sensitive digital watermark integrally with the image, or a pattern of dots with pseudo-random distribution, also known as “AMSM”. This type of digital authentication code is usually located over a large area and has a low information density. However, digital authentication codes can have a high information density and can be concentrated on a small area, for example, protected information matrices and patterns for copy recognition. Protected information matrices and patterns for copy recognition are often included in a digital file of a document or product and printed at the same time.

Patterns for copy recognition are background patterns generated in a pseudo-random manner using a cryptographic key. Copies are recognized by comparing and measuring the similarity between the original digital image and the captured image. In addition, the pattern for recognizing copies may contain a small amount of information.

Protected information matrices are information matrices designed to contain a large amount of information in encrypted form, and when reading this information is resistant to a high error rate. Copies are recognized by measuring the error rate of the message.

Protected information matrices and patterns for copy recognition often consist of half “black” (or color) pixels and half of “white”, or non-print, pixels. However, for some types of printing, paper types, or printing machine settings, the printed information matrix may be overloaded with ink. Excessive ink filling in a protected information matrix can significantly reduce its readability and even the ability to be distinguished from one of its copies. Therefore, it is extremely important to avoid an excessive level of ink, however, in practice this is not always achievable, since the ink level is rarely a value completely controlled by the printer, and in some cases it is a value set by the printer to the client. Thus, it is more preferable to create secure information matrices whose properties are less sensitive to the level of ink applied to the paper.

It turns out that protected information matrices are usually more sensitive to an increased level of paint than to a lower one. In fact, if the ink level is lowered, black or color-containing cells are usually printed, and therefore have little effect on the reading of the matrix. If the ink level is increased, the ink tends to overfill the substrate, and white areas are “flooded” to some extent by paint from neighboring black areas. A similar effect is observed when applying marks by contact printing, laser engraving, etc.

Theoretically, protected information matrices create, in accordance with a given print resolution, for example, 600 dots per inch (600 ppi). However, it turns out that depending on the printing conditions, the optimal print resolution, or the resolution that provides the best recognition of originals and copies, varies: the higher the print quality, the higher the print resolution of the matrix or the size of its cells should be reduced accordingly.

The objective of the fifth and sixth aspects of the present invention is to remedy these disadvantages.

To this end, according to a fifth aspect of the present invention, there is provided a method for protecting documents, according to which a configuration is printed comprising cells that represent a unit of information and the appearance of each of which changes in accordance with the information that it represents, said configuration being capable of recognizing a copy , modifying the appearance of these cells, characterized in that it contains:

a step of determining a part of cells that has a constant shape and a changing color to represent different values of a unit of information and which is certainly less than the specified cell, and

- the stage of presenting a unit of information in the specified configuration by changing the appearance of the parts of the cells.

Thanks to the indicated means, even if an increased level of ink is present during printing, since only a limited part of the cells is filled with ink, the risk of flooding another cell and changing its appearance is reduced, and the ability to recognize a copy is increased.

Thus, in order that the protected information matrices can guarantee the recognition of copies under any printing conditions, use such protected information matrices in which at least one part corresponds to the printing conditions under which the ink filling level is increased. Thus, the properties of the protected information matrix, which provide copy protection, become insensitive to the level of ink filling used in printing. It should be noted that the choice of part in each cell to be printed is preferably combined with the choice of cell size, which is described elsewhere, in order to obtain the proportion of errors that best contributes to the recognition of copies.

According to the features of the invention, the method, a brief description of which is presented above, comprises the step of determining several configurations not superimposed on one another, the cell sizes of at least two different configurations being different.

Thanks to these means, the same secure information matrix can be printed on different media with various printing means that do not have the same resolution, and, nevertheless, retain its copy recognition properties.

According to the features of the invention, the method, a brief description of which is presented above, comprises the step of determining several configurations that are not superimposed on one another, and at the stage of determining the part, the specified part differs for at least two different configurations.

Thanks to these means, protected information matrices are obtained that are resistant to a wide range of ink filling levels, since several parts of the specified protected information matrix corresponding to the above configurations are adapted to various ink filling levels. Thus, a protected information matrix may contain several sections in which cell density, i.e. the ratio of the area of the part to the area of the cells varies so that at least one of the values of the density of the cells corresponds to the level of ink filling used in printing. In this case, the reading can be performed using areas having the most suitable level of filling with paint.

According to the features of the invention, each cell is square, and said part of the cell is also square.

For example, if the cell size is 4 × 4 pixels, you can choose to print a square portion of 3 × 3 pixels or 2 × 2 pixels. Consequently, ink filling is reduced in the ratio of 9/16 and 1/4, respectively (it should be noted that white cells are not affected). In another example, if the cell is 3 × 3 pixels, you can select to print a square portion of 2 × 2 pixels or 1 × 1 pixels.

According to the distinguishing features of the invention, said part has the shape of a cross. For example, the indicated cross is formed of five out of nine printed pixels.

According to the features of the invention, the method that is the subject of the present invention, a brief description of which is given above, comprises the step of determining the cell sizes of at least one configuration to be printed in accordance with the printing conditions, so that the proportion of cells printed with a printing error , arising solely due to unforeseen variables, was greater than the given first value, and less than the given second value.

Since the features of the method proposed by the first aspect of the present invention are also features of the method proposed by the fifth aspect of the present invention, they are not repeated here.

A sixth aspect of the present invention is the creation of a printed configuration containing cells that represent a unit of information and the appearance of each of which varies in accordance with the information represented by the specified cell, and this configuration is made with the possibility of recognizing a copy that modifies the appearance of these cells, characterized in that the cells contain a part that has a constant shape and a changing color, representing different unit values of information, and otorrhea certainly less than the specified cell, the appearance of portions of cells is the specified information.

A seventh aspect of the present invention is the creation of a device for protecting documents, containing a device for printing a configuration formed by cells that represent a unit of information and the appearance of each of which changes in accordance with the information represented by the specified cell, and this configuration is made with the possibility of recognition of the copy, modifying the appearance of these cells, characterized in that it contains:

- means for determining a part of cells having a constant shape and changing color to represent different values of a unit of information, said part being certainly smaller than a specified cell, and

- means of representing a unit of information by changing the appearance of parts of cells.

Since the advantages, objectives, and features of this printed configuration, the creation of which is proposed by the sixth aspect of the present invention, and the device, the creation of which is proposed by the seventh aspect of the present invention, are the same as in the fifth aspect of the present invention, they are not repeated here.

To decide on the authenticity of the document, depending on the errors that the configuration cells carry, the message contained in the configuration can be decoded, or the image of the specified configuration can be restored. However, in the second case, it is necessary to provide means for restoring the original digital configuration in the copy recognition device, which is a very weak point from the point of view of information security, since the fake manufacturer who takes possession of this device can generate original configurations without errors. If in the first case marking leads to a significant distortion of the message, especially copies, or if the message carries a large amount of information, it may be unreadable, in this case, the error rate cannot be measured. In addition, reading the message carried by the configuration using the copy recognition device is another weak point in terms of information security, since a fake manufacturer who has taken possession of this device can take advantage of this message.

In addition, the determination of the authenticity of the configuration entails the intensive use of memory resources, data processing and / or means of communication with a remote authentication server.

The eighth aspect of the present invention aims to address these drawbacks.

To this end, an eighth aspect of the present invention provides a method for determining the authenticity of a configuration printed on a document containing:

- the stage of determining the sets of cells of the specified printed configuration, and the cells of each set of cells correspond to the same unit of information,

- the step of capturing an image of the specified configuration,

- the step of determining the proportion of cells of each set of cells of the specified configuration, which represent an information value different from the information value in other cells of the specified set, and

- the step of determining the authenticity of the specified configuration in accordance with the specified fraction of the cells of at least the specified set of cells.

Thus, by using the method according to the eighth aspect of the present invention, neither restoration of the original reproduced message is required, nor even decoding of the message, and it is not required that the message mean anything, because Information may be random. In fact, the number of message errors is measured using some properties of the message itself at the time the encoded message is evaluated.

However, it should be noted that groupings of cells representing an informational value, usually binary, should be known. According to the distinguishing features of the invention, at the stage of determining the fraction of cells, the average information value that is carried by different cells of the same set of cells is determined.

According to the distinguishing features of the invention, at the stage of determining the proportion of cells, the indicated average value is determined by weighing the information value carried by each cell in accordance with the appearance of the specified cell.

Thus, a weight or coefficient is associated with each cell, indicating the probability that the evaluation of each bit of the encoded message is correct. The indicated weight is used to weight the contributions of each cell in accordance with the probability with which the associated bit is correctly estimated. The simplicity of the implementation of this method consists in the fact that the values read in each cell from a plurality of cells are not converted to binary form.

According to the distinguishing features of the invention, the method, a brief description of which is given above, comprises the step of determining the average value across the configuration of the values represented by the cells, and the step of compensating for the difference between the specified average value and the expected average value.

It should be noted that the more noisy the message, the higher the risk that the estimated bit of encoded information is erroneous. This causes a systematic error, which consists in the fact that when measuring the number of errors, the real number of errors is underestimated. The indicated systematic error is estimated statistically and corrected in the process of measuring the number of errors.

According to the distinguishing features of the invention, at the stage of determining the fraction of cells of the specified set of cells representing the information value that is different from the information value in other cells of the specified set, in order to obtain the information value in the specified cell, a cryptographic key is used that changes the information value represented by at least one cell image configuration.

According to the distinguishing features of the invention, at the stage of determining the fraction of cells of said plurality of cells representing an information value different from the information value in other cells of said set, the probability of having an image point value for at least one configuration image point is used.

Reading the DAC requires accurate determination of its position in the captured image so that the value in each of its constituent cells is restored with the greatest possible accuracy, taking into account signal distortions resulting from printing and, possibly, capture. However, captured images often contain symbols that may interfere with the positioning step.

Establishing the position of a protected information matrix can be difficult due to capture conditions (poor lighting, blurring of borders, etc.), as well as an arbitrary position angle in the range of 360 degrees.

Unlike other two-dimensional barcode characters, which vary relatively little depending on various types of printing, DAC characteristics, such as image texture, vary greatly. Thus, known methods, for example disclosed in US 6775409, are unacceptable. In fact, this method is based on the directionality of the brightness gradient, i.e. on its variation in accordance with the direction of its position for detecting codes. However, the brightness gradient of the protected information matrix does not have a special orientation.

Some DAC positioning methods may use the fact that they are square or rectangular, which causes a noticeable contrast with continuous segments that can be detected and processed by standard image processing methods. However, in some cases, these methods are unsuccessful, and on the other hand, it is desirable to be able to use DACs, which are not necessarily square or rectangle (or are not necessarily inscribed in them).

Generally speaking, the area on which the DAC is printed has an increased ink density. However, although the use of ink density measurements is useful, it cannot be the only criterion. In fact, Datamatrixes (registered trademark) or other barcodes that may be adjacent to the DAC have an even higher ink density. Therefore, it is not sufficient to use only the specified criterion.

In the patent EP 1801692 in order to determine areas of the image belonging to the patterns for recognition of copies, it is proposed to use increased entropy of such patterns. However, although these patterns do have high entropy before printing, the latter can change dramatically during printing, capture, or measurement. For example, a simple measurement of entropy, based on a histogram of the scatter of pixel values of each section, can sometimes lead to increased values in areas that are not very full in content, which theoretically should have low entropy, for example, this may occur as a result of artifacts during compression in JPEG format , due to the texture of the paper appearing in the captured image, or as a result of reflection effects from the substrate. Thus, it is clear that the criterion of entropy is also insufficient.

Generally speaking, methods for measuring or characterizing textures seem more appropriate in order to characterize at the same time the intensity or spatial relationships specific to DAC textures. For example, Haralick's Statistical and Structural Approaches to Texture describes many dimensions of texture characteristics that can be combined so that they uniquely describe a large number of textures.

However, DACs can have textures that vary greatly depending on the type of print or capture, and in most cases it is impossible or at least impractical to provide texture characteristics for the way the DAC is placed, mainly because they must be adjusted to the specific effects of the media. capture when measuring texture.

The objective of the ninth aspect of the present invention is to remedy these disadvantages.

To this end, a ninth aspect of the present invention is the creation of a method for determining the position of a configuration, comprising:

- the step of dividing the configuration image into sections so that the configuration area corresponds to the number of sections greater than a predetermined value,

- the step of measuring the texture index of each section,

- the step of determining the detection threshold of part of the configuration,

- the stage of determining the sites belonging to the specified configuration by comparing the indicator of the texture of the site with the corresponding detection threshold,

- the stage of determining continuous groups of sites belonging to the specified configuration,

- the step of determining the contour of at least one group of sections and

- the step of matching the contour of at least one group with the contour of the specified configuration.

Thus, a plurality of criteria are used in the present invention to reliably determine the configuration position.

According to the features of the invention, a texture index characterizes the density of the ink for printing the configuration.

According to the distinguishing features of the invention, the texture index characterizes the local dynamics. It should be noted that the local dynamics may include various physical parameters, for example, the frequency or proportion of local changes, or, for example, the sum of the gradients.

According to the distinguishing features of the invention, at the stage of determining the detection threshold, the specified threshold varies depending on the position of the plot in the image.

According to the distinguishing features of the invention, at the stage of determining the areas belonging to the specified configuration, at least one operation of stretching and / or blurring the image is used.

According to the distinguishing features of the invention, this configuration is rectangular, and at the stage of comparison, two pairs of points are formed, formed by the points farthest from each other, and it is determined whether the ratio of the lengths of the line segments formed by these pairs of points lies in a given range of values.

According to the distinguishing features of the invention, this configuration is rectangular, and at the stage of comparison, two pairs of points are formed, formed by the points farthest from each other, and it is determined whether the angle between the straight lines formed by these pairs of points lies in a given range of values.

According to the distinguishing features of the invention, said configuration is rectangular, and the Hough transform is used in the matching step.

A tenth aspect of the present invention provides a device for determining a configuration position, comprising:

- means for dividing the configuration image into sections so that the configuration area corresponds to the number of sections greater than a predetermined value,

- a means of measuring the texture index of each section,

- means for determining the detection threshold of part of the configuration,

- means for determining areas belonging to the specified configuration, by comparing the indicator of the texture of the site with the corresponding detection threshold,

- a means of determining continuous groups of sites belonging to the specified configuration,

- means for determining the contour of at least one group of sections and

- means for matching the contour of at least one group in accordance with the contour of the specified configuration.

Since the advantages, objectives, and features of this device, the creation of which is proposed by the tenth aspect of the present invention, are the same as in the ninth aspect of the present invention, they are not repeated here.

An eleventh aspect of the present invention provides a method for generating a configuration for copy protection, comprising:

- the step of determining at least one print parameter of the specified configuration,

- the stage of inclusion in the composition of the specified configuration messages representing the specified print option,

- the step of printing the specified configuration using the specified print option.

The inventors of the present invention have found that if the print parameters, including the printing device, the substrate used and other printing parameters (including the size of the raster for offset printing) are known, they can be useful when using a configuration designed to protect against copying, especially for its authentication.

According to the features of the invention, at least one of said printing parameters is a characteristic of the type of substrate on which said configuration is printed.

For example, it is specifically indicated what the substrate is: paper, cardboard, aluminum, PVC, glass, etc.

According to the features of the invention, at least one of said printing parameters is a characteristic of the device used for printing.

For example, it is specifically indicated whether the print media are offset, letterpress, screen, gravure, etc.

According to the features of the invention, at least one of these printing parameters is a characteristic of the density level of the ink used in the printing process.

According to the distinguishing features of the invention, at the stage of determining at least one print parameter, the image of the pattern printed using the printing means used in the printing step is captured and the value of the specified parameter is automatically determined by processing the specified image.

A twelfth aspect of the present invention provides a method for determining the authenticity of a printed configuration for copy protection, comprising:

- a step of capturing an image of a printed configuration for copy protection,

- the step of reading from the specified image a unit of information representing at least one print parameter of the specified configuration, and

- the step of determining the authenticity of the printed configuration, intended to protect against copying, using the specified information representing at least one print parameter of the specified configuration.

A thirteenth aspect of the present invention provides a device for generating a configuration for copy protection, comprising:

- means for determining at least one print parameter of the specified configuration,

- means for including in the composition of the specified configuration messages representing the specified print option, and

- print media of the specified configuration using the specified print option.

A fourteenth aspect of the present invention provides a device for determining the authenticity of a printed configuration for copy protection, comprising:

- means for capturing images of a printed configuration intended to protect against copying,

- means of reading from the specified image a unit of information representing at least one print parameter of the specified configuration, and

- means for determining the authenticity of the printed configuration intended to be protected against copying, using the specified information representing at least one print parameter of the specified configuration.

Since the advantages, objectives, and features of this method, the creation of which is proposed by the twelfth aspect of the present invention, and the devices, the creation of which are proposed by the thirteenth and fourteenth aspects of the present invention, are the same as in the eleventh aspect of the present invention, they are not repeated here.

The essential or distinguishing features of each aspect of the present invention are the hallmarks of other aspects of the present invention with the aim of creating a document security system that benefits all aspects of the present invention.

BRIEF DESCRIPTION OF THE DRAWINGS

The following is a detailed description in the form of an example, which does not limit the scope of the invention and in which, for a better understanding of the other advantages, objects, and features of the present invention, reference is made to the accompanying drawings:

figure 1 schematically in the form of a logical diagram illustrates the steps of detecting, printing and capturing information for the original and a copy of the specified original,

figure 2 schematically in the form of a logical diagram illustrates the steps used to apply labels to documents or products in order to enable further authentication,

figure 3 schematically in the form of a logical diagram illustrates the steps used to authenticate documents or products with a label applied using the steps illustrated in figure 2,

on figa and 4B shows information matrices intended for marking an object,

on figa and 5B shows a captured image of a genuine label information matrix and the copied label of the specified information matrix, respectively,

6 shows an information matrix printed with an ink fill level too high,

7 shows an information matrix containing in the central part a matrix of points with variable characteristics,

on Fig shows an information matrix surrounded by a matrix of points with variable characteristics,

figure 9 shows the information matrix containing a completely filled with paint plot,

figure 10 shows the information matrix containing the paint-filled adjacent area,

figure 11 shows, on the one hand, at the top of the information matrix and, on the other hand, at the bottom the same information matrix, modulated cell by cell, using the reproduced message,

12 shows various information matrices in which only a limited portion of the cells has a variable appearance and is black or white,

13 shows information matrices in which various parts of cells with a variable appearance are used, and at the end, a matrix (13c) coated (or paved) with a second (13b),

on Fig shows an information matrix captured at an angle of about 30 ° and with a resolution of about 2000 dpi (2000 dpi),

on Fig presents the measurement result of a combined indicator of the texture (106 × 85) of the image shown in Fig,

in Fig.16 shows the image shown in Fig.15, after comparing with a threshold value,

in Fig.17 shows the image shown in Fig.16, after applying at least one stretching operation and one blurring operation,

on Fig shows the contour of the information matrix, determined by processing the image shown in Fig,

in Fig.19 shows the angles of the contour shown in Fig.18, determined by processing the image shown in Fig.18, and

- Fig.20 presents graphs showing the proportion of errors in accordance with the cell size of the information matrix.

DETAILED DESCRIPTION OF EMBODIMENTS OF THE INVENTION

Before a detailed description of some embodiments of the present invention, the following are definitions of terms used.

An “information matrix” is a machine-readable physical representation of a message, usually applied to a hard surface (as opposed to watermarks or steganographic images, which usually modify the pixel values of a printable image). The definition of the information matrix covers, for example, two-dimensional barcodes, one-dimensional barcodes and other less intrusive means of presenting information, for example Dataglyphs (information marking).

A "cell" is an element of an information matrix representing a unit of information.

A “document” is any object (material) that contains an information matrix.

“Marking” or “printing” refers to any process by which a transition is made from a digital image (including an information matrix, document, etc.) to its representation in the real world, and this representation is usually performed on the surface, for example, inkjet, laser, offset, thermal printing, as well as stamping, laser engraving, the formation of holograms. In addition, this includes more complex processes, such as casting, in which a digital image is first engraved on the mold and then formed on each object (it should be noted that the “cast” image can be considered three-dimensional in the physical world, even if its digital representation is two-dimensional) . It should also be noted that some of the mentioned methods include several transformations, for example, standard offset printing (as opposed to offset printing using computer-printing form technology) includes creating a film, making a printing form using the specified film and printing using the specified printing form. Other methods also allow you to print a unit of information in the invisible region or using frequencies outside the visible spectrum or by putting information inside the surface, etc.

“Capture" refers to any process by which a digital representation of the real world is obtained, including a digital representation of a physical document containing an information matrix.

Throughout the following description, square configurations are used. However, the present invention is not limited to this type of configuration, but, on the contrary, applies to all configurations that can be printed. For example, it is possible to use configurations consisting of secure information matrices having different resolutions and different levels of ink filling according to the present invention, the advantage of which is, in particular, that at least one of these matrices corresponds to the optimal resolution and optimum ink density.

Throughout the present description, filling in a printed configuration is used, which may be represented by a matrix of cells. However, the present invention is not limited to this type of configuration, but, on the contrary, applies to all types of filling cells with the same or different shapes or sizes.

As an introduction to the description of embodiments of the method and device according to the present invention, it should be noted that the result of distortion of the information matrix is that some cells cannot be correctly decoded.

At each stage of creating the information matrix, the goal is to ensure that the original message is read without errors, even if the original reading of the information matrix is violated by errors. In particular, one of the goals of creating this information matrix is to use the number or error rate of encoded, reproduced, rearranged, or scrambled messages in order to determine the authenticity of the label of the information matrix and, therefore, the document containing it.

In fact, the specified distortion coefficient can be selected in accordance with the printing parameters, since making a copy leads to additional errors, as a result of which the error rate is higher on average when reading a copy than when reading the original.

To understand why measuring the error rate in a message may be sufficient to determine if the document is the original and a copy, an analogy with communication systems may be useful. In fact, the transition of an encoded scrambled message into an information matrix is nothing more than a modulation of a message, which is defined as the process by which a message is converted from its original form to a form suitable for transmission over communication channels. The specified communication channel, namely the means of transmitting a message connecting the source to the recipient and ensuring the transmission of the message, differs depending on whether the captured information matrix is original or copied. The communication channel may vary: the communication channel of the “original” and the communication channel of the “copy” are different. This discrepancy can be measured based on the signal-to-noise ratio, which is smaller for the captured copied information matrix.

The encoded message extracted from the captured copy of the information matrix will contain more errors than the encoded message extracted from the captured original information matrix. In accordance with some features of the present invention, the number or coefficient of errors detected is used to distinguish a copy from the original.

The communication channel of the original and the communication channel of the copy are mainly described on the basis of the subchannels contained in them, and these subchannels are partially different in the two indicated cases. As described below, each subchannel of the signal transmission channel, i.e. information matrix, is an analog-to-digital or digital-to-analog conversion.

1 shows communication channels for a captured original information matrix and a captured copy of the information matrix. The first channel contains a subchannel 105, which converts the digitally generated information matrix into the actual information matrix, i.e. into the analogue label on the document whose protection you want to ensure, i.e. on the original document, and the second subchannel 110, corresponding to the reading of the specified label. In the case of a copy, in addition to the indicated first two channels, a third subchannel 115 is used, which is used to reproduce a tag from a read tag in the real world, and a fourth subchannel 120, is used to read the specified fingerprint for authentication.

It should be noted that, as an option, it is possible to produce a second copy on the basis of the first purely analogous way (for example, using analog photocopying or analog photography), however, the fifth analog-to-analog subchannel 125 is usually characterized by a large signal distortion exceeding the distortion due to reading using a sensor high resolution images.

The third, fourth, and / or fifth subchannels impose additional message distortion, allowing one to distinguish the original, for example, the image 505 shown in FIG. 5A, from the copy, for example, the image 510 corresponding to the same information matrix as the image 505 shown in FIG. 5B. As can be seen from the comparison of images 505 and 510, the copy has less detail, the distortion of these images corresponds to errors reproduced by the label of the original information matrix.

Because counterfeit manufacturers seek to minimize production costs, the subchannels used to make the copy, and in particular the subchannels leading to an analog print, in this case the third and fifth channels, are sometimes performed with poor quality marking or printing. Therefore, messages contained in copies made in this way have a significantly lower signal to noise ratio, which makes it possible to recognize these copies in an even simpler way. However, it should be noted that in cases where the manufacturer of fakes uses print media that are similar or even superior in quality to the print media used in the manufacture of original documents, usually there are no particular problems. In fact, the manufacturer of fakes can not completely avoid the imposition of noise, which when printing a copy leads to additional errors in the process of demodulation of the information matrix. Thus, this operation will reduce the signal-to-noise ratio. In most cases, the signal-to-noise ratio difference will be sufficient to distinguish captured original information matrices from captured copies of information matrices.

Preferably, the information matrix and, in particular, the clarity of its details, is such that a print whose characteristics are known in advance is such that the printed information matrix is distorted. In addition, the encoded message contains errors when reading in a significant, but not excessive amount. Thus, when printing a copy, the fake manufacturer will not be able to avoid additional distortion. As indicated, the distortion in the printing process of the original should be natural and random, i.e. be caused by a physical phenomenon of a locally unpredictable nature, the dispersion of ink in paper, the natural instability of the printing press, etc., i.e. one that cannot be traced. Since the loss of information will be essentially irreversible, this distortion is such that the manufacturer of fakes will not be able to either correct errors or avoid additional errors, and the same physical phenomenon will occur when printing a copy.

In order to increase protection against counterfeiting, the creation of an information matrix is dependent on at least one parameter that is kept secret, called a secret key or secret keys. Therefore, if the previous key was disclosed by third parties, you only need to change the secret key in order to return the original degree of protection. To simplify the description, we will usually talk about one secret key, it is understood that the specified key may itself consist of several secret keys.

The secret key is used to encrypt the original message before encoding it. Since this type of encryption can use an avalanche effect, and in this case errors that occur during demodulation or matrix reading, in most cases are eliminated by an error correction code, two information matrices generated from the same key that contain messages that differ by only one bit , i.e. the smallest possible discrepancy between the two messages will seem completely different.

The same is true for two information matrices containing identical messages, but generated from different keys. The first of these properties is especially advantageous, since the manufacturer of fakes cannot recognize a periodically repeating pattern that can be used to make fakes by analyzing information matrices created using the same key but containing different messages. It should be noted that it is also possible to add a random number to the message so that the two information matrices generated using the same key and the same message, but containing different random numbers added to the message, also seem completely different.

The information matrix can be considered as the result of modulating a message represented by symbols of the alphabet, for example, binary. In some embodiments, synchronization, alignment, or positioning symbols are added at the message level, or patterns are inserted at the level of the information matrix to determine location.

Figure 2 in the form of a logical diagram illustrates the steps of generating an information matrix and labeling a document in accordance with some features of the method according to the present invention.

After the start of the process (step 185), at step 190, at least one marking or printing parameter is obtained or measured, for example, the type of printing, the type of medium, the type of ink used. Then, at step 195, it is determined whether the area of the secure information matrix area or the number of its cells is fixed for a given application or for a given client. At step 200, the ink density is determined corresponding to certain marking / printing parameters, for example, by reading in a database or in a density look-up table corresponding to printing parameters. At step 205, the cell size of the protected information matrix is determined, for example, by reading in the database or in the look-up table, the cell sizes corresponding to the printing parameters. It should be noted that the correspondences stored in databases or in look-up tables are determined as described below, in particular with respect to FIG. The indicated correspondences are aimed at achieving good print quality and a fraction of printing errors between the first and second preset values, for example 5%, 10%, 15% or 20% for the first preset value and 25% or 30% for the second preset value.

Then, at step 210, a message is received intended for transmission using a document and mainly depending on the identifier of the document, and at step 215, at least one secret encryption and / or scrambling secret is obtained.

The original message is, for example, the name of the document, one or more copyright holders of the attached intellectual property rights, the manufacturing order number, the addressee of the document, and the manufacturer’s data. The specified message is made using known methods. The original message is represented by a given alphabet, for example, alphanumeric.

To create an encrypted message at step 215, the message is encrypted using a symmetric key or preferably an asymmetric key, for example, a PKI key pair (short for “Public key infrastructure” - public key infrastructure) .Thus, in order to increase the degree of security of the message, it is encrypted with so that a change in one unit of binary information of the message at the encryption input causes a change in a large amount of binary information at the output of the encryption process.

Encryption is typically performed on blocks of bits of a fixed length, for example 64 bits or 128 bits. The following encryption algorithms can be used: DES (short for data encryption standard) with a 56-bit key and a 64-bit message block, DES triple encryption with a 168-bit key and a 64-bit message block, and AES (short for "advanced encryption standard") with 128, 192 and 256 bit keys and a 128-bit message block, as they are widespread and recognized as resistant to attacks. However, it is also possible to use many other encryption algorithms, block or sequential. It should be noted that theoretically, when using block encryption algorithms, encrypted messages of the same length are obtained as the original message, in so far as it is a multiple of the length of the block.

It is believed that the AES standard has the highest level of protection, but it should be noted that it works with message blocks of a minimum length of 128 bits. If the length of the transmitted message is a multiple of 64 bits, an algorithm similar to DES triple encryption should be used. Finally, it is possible to create a new encryption algorithm, especially if you are limited to a very small message size, for example 32 bits. However, it should be noted that the security of these algorithms will be limited due to the small number of different encrypted messages.

However, it should also be noted that theoretically cryptographic attacks aimed at searching for a key cannot be used, at least in their standard form, known in cryptography. In fact, the manufacturer of fakes theoretically has access only to the captured image of the printed original information matrix, and to start a cryptographic attack, he needs at least access to the decrypted message. However, a message can only be decrypted if it is previously descrambled; for this, a search for the scrambling key is required.

These encryption methods are called "symmetric", i.e. when decrypting, the same key is used. The transfer of keys to the recognition module and their storage must be carried out in a very well-protected manner, since the fake manufacturer who has taken possession of the specified key will be able to create encrypted messages that will not look fake. However, these risks can be mitigated by using asymmetric encryption methods in which the decryption key is different from the encryption key. In fact, since the decryption key does not allow encrypting messages, the fake manufacturer who received the specified key will not be able to create either new valid messages, or, as a result, information matrices carrying another message.

At step 220, the encrypted message is encoded to obtain an encoded encrypted message. Preferably, coding uses convolutional coding, which is very fast, and decoding also occurs quickly, for example using the well-known method developed by Viterbi. If the convolutional encoding uses a polynomial generator of the ninth degree, and the code rate is two bits per output per bit at the input, we obtain a code increase of 7 dB relative to the same message subjected to simple reproduction, as a result, the risk of error is significantly reduced during decoding. If the message intended for encoding contains 128 bits, using the above convolutional code, a 272-bit encoded message is received (two bits are obtained for each of the 128 bits of the code, eight more bits are allocated by the encoder memory for the polynomial generator of the ninth degree). However, it should be noted that it is possible to use many other types of coding (arithmetic coding, turbo codes, etc.) using the same principle.

Preferably, hereinafter said encoded encrypted message is written in the binary alphabet, i.e. it contains “0” and “1”.

At step 225, an encoded encrypted message is reproduced and inserted into the list of available cells in an information matrix, inaccessible sections of which contain synchronization, alignment or positioning symbols or location patterns, which, according to embodiments of the invention, are determined using a secret key. Patterns for alignment are, for example, matrices of 9 × 9 pixels, periodically distributed in the information matrix. Thus, the encoded encrypted message is reproduced or repeated so that each unit of binary information is presented several times in accordance with the number of available cells of the information matrix. The specified playback, depending on the coding of repetition or redundancy, can significantly reduce the error rate of the encoded message, which will be input to the decoding algorithm of the convolutional code. Errors that are not corrected by repetition are in most cases corrected by a convolutional code.

At steps 235 and 240, the reproduced encoded encrypted message is encrypted by permutation (the so-called “scrambling”) to obtain a scrambled encoded encrypted message.

Preferably, the scrambling function of the reproduced encoded encrypted message consists in sequentially using binary permutation of the message in step 235 and substitution in step 240, both steps using a second secret key, possibly identical to the first secret key. Preferably, the substitution is performed using an exclusive OR function and a pseudo-random sequence.

Thus, the scrambling of the encoded encrypted message is performed in a non-trivial way using a secret key, which may be identical to or different from the key used to encrypt the message. It should be noted that if, according to the distinguishing features of the invention, another key is used, it can be calculated from the function of the key used for encryption.

The use of a secret key both for encrypting a message and also for scrambling a coded message provides an increased level of protection against fakes. For comparison, it should be mentioned, since the existing methods for creating 2D barcodes in which the encoded message is not scrambled. The fake manufacturer can easily reproduce the original information matrix after decoding the message of the captured information matrix. Even if the decoded message is encrypted, it does not need to decrypt this message to accurately reproduce the information matrix.

Preferably, the scrambling in this case consists of a combination of permutation at step 235 and the use of the exclusive-OR function, or the XOR function, at step 240, the table of which has the following form:

BUT AT S = A XOR B 0 0 0 0 one one one 0 one one one 0

In fact, this type of scrambling avoids an increase in the number of errors (there is no so-called "avalanche effect" here: an error in one element of the scrambled message leads to one and only one error in the descrambled message). The avalanche effect is undesirable because it makes it difficult to read the information matrix when a single error is present in the scrambled message. However, errors have been shown to play an important role in the use of the present invention.

The permutation (step 235) is carried out in accordance with the permutation algorithm using a key that ensures that all permutations are performed in a pseudo-random way. At step 240, the exclusive OR function is applied to a sequence containing permutations (the length of which corresponds to the number of available cells) and to a binary sequence of the same length also generated from the key. It should be noted that if the message is not presented in binary form (the cells are capable of representing more than two possible values), the permutation can be performed in the same way, and the function "exclusive OR" can be replaced by a function that performs modulo summation of the number of possible values of the message with a pseudo-randomly generated sequence containing the same number of possible values as the scrambled message.

Many permutations depend on the existing secret key. A simple algorithm is to loop through the increasing subscript i = from 0 to the message length N-1 and for each index i generate a pseudo-random integer j = 0 to N-1 with the subsequent permutation of the message values in the position of the indices i and j.

Pseudorandom numbers can be generated using chain encryption algorithms, such as the above, or a hash algorithm, such as SHA-1 (the second version of the "Secure Hash Algorithm", which is part of the US government standard). To initialize the algorithm, a key is used, which is re-initialized at each new stage from the numbers generated in the previous stage.

After the binary data of the message is permuted in step 235, the bit values in step 240 are passed through an exclusive-or-xor filter, along with a sequence of pseudo-random bit values having the same length as the message. According to an embodiment of the invention, step 240 is performed before step 235 of the permutation.

Thus, each of the binary data of the scrambled reproduced encoded encrypted message is modulated in the cell of the information matrix by assigning one of two colors, for example black and white, to the binary value "0" and the second of them to the binary value "1", and the correspondence can vary in image surface.

Depending on the printing method, only one of two colors can be printed in step 245, the second color corresponding to the original color of the substrate or pre-printed as a “background”. For printing methods that create a physical relief, for example for embossing or laser engraving, choose one of two colors associated with a specific binary value, for example randomly.

Generally speaking, the image size in pixels is determined by the area of the available surface of the document and the print resolution. If, for example, the available surface area is 5 mm × 5 mm, and the print resolution of the matrix is 600 pixels per inch (this value is often expressed in the British measurement system), a specialist can calculate that the available surface area in pixels is 118 × 118 pixels. If we assume that a black border of 4 pixels is added on each side of the matrix, the matrix size in pixels will be 110 × 110 pixels, 12,100 pixels in total. Assuming that the size of each cell is one pixel, the information matrix will contain 12100 pixels.

Alignment blocks having a known or determined by the detector value may be included in the matrix. These blocks, for example, 10 × 10 pixels in size, can be inserted at regular intervals from the upper left corner of the matrix, for example, every 25 pixels. Therefore, it should be noted that the matrix will consist of 5 × 5 = 25 alignment blocks, each with a size of 100 pixels, a total of 25 × 100 = 2500 alignment pixels, or 2500 message cells. Thus, the number of cells available for playing back the encoded message is 12100-2500 = 9600. According to the above, the encoded message contains 272 bits, and can be played completely 35 times and partially 36th time (80 first bits of the encoded message). It should be noted that these 35 repetitions can increase the signal-to-noise ratio of the encoded signal by more than 15 dB, which provides a very small risk of error when reading a message.

Two examples of representations of the information matrix obtained according to the document security method illustrated in FIG. 2 are shown in FIG. 4A, a matrix 405 is shown without visible alignment blocks, and FIG. 4B is a matrix 410 with visible alignment blocks 415. The last Due to the regular arrangement, alignment blocks 415 formed by black crosses on a white background are clearly visible. According to other variants of the invention, as shown in FIG. 4A, said blocks have exactly the same appearance as the rest of the image. Finally, as shown in FIGS. 4A and 4B, a black border 420 may be added on all sides of the message and possible alignment blocks.

It should be noted that, in addition to edging and alignment blocks, which can be pseudo-random, the binary values “0” and “1” are preferably equally probable.

According to one embodiment of the invention, the edging of the information matrix is made up of cells having a larger size than the cells of the rest of the labeling region in order to present a more stable message in them. For example, in order to construct a square border cell, four peripheral cells of the information matrix are connected and in this border a scrambled encoded encrypted message is presented. In this way, the contents of the border will be very resistant to subsequent distortion of the mark, in particular to capturing its image or copying it to another document.

According to other variants of the invention, in addition to the message that the information matrix carries, the message also carries a document, for example, on an electronic label or a two-dimensional barcode. As indicated below, this additional message may be an initial message or a message used in the process of authentication of a document, for example, keys used to generate the information matrix, data associated with these keys and stored in remote memory, the threshold for the number of errors used to make a decision about the authenticity of the document.

According to embodiments of the invention, after steps 235 and 240, a partial scrambling of the reproduced encoded encrypted message is performed using the third secret key. Therefore, the scrambled reproduced encoded encrypted message can itself be partially scrambled using a key different from the key or keys used in the previous steps. For example, the specified additional partial scrambling affects approximately 10 to 20% of the cells (usually a fixed amount). Cells subjected to said additional scrambling are selected in a pseudo-random manner using the additional scrambling key. The values in the selected cells can be systematically modified, for example, for binary values by replacing "1" with "0" and "0" with "1". In one embodiment of the invention, the selected cells may be passed through an exclusive-OR filter generated from an additional scrambling key and therefore have a 50% chance of modification.

The purpose of this additional scrambling is that a detector that is not equipped with an additional scrambling key, nevertheless, could correctly retrieve the message and recognize the copies. However, such a detector, if it falls into the hands of unauthorized persons, does not contain all the information or all the keys necessary for reproducing the original. In fact, without an additional scrambling key, an attacker will be able to generate and print an information matrix, which will be recognized by a detector equipped with an additional scrambling key as a copy. Generally speaking, detectors considered less reliable will not be equipped with an additional scrambling key.

Below we will consider other options implemented using this principle, consisting in the failure to provide all the keys or parameters used to create the information matrix.

At step 245, the document is marked with an information matrix, for example by printing or engraving, with a label resolution such that the presentation of the information matrix contains errors that occur due to the indicated step of labeling in such a way that any reading of the specified information matrix shows a non-zero error rate. At the stage of labeling, a label is formed containing, due to physical conditions of labeling, local errors that are at least partially random or unforeseen, i.e. errors that also affect the representations of the cells of the information matrix.

The physical conditions for marking include, in particular, the physical tolerances of the marking means, the carrier (image) and, in particular, the state of its surface and material, for example, possibly applied paint. The term "unforeseen" means that until the physical implementation of labeling a document, it is impossible to determine which cells of the information matrix will be correctly represented using the label and which will be erroneous.

For each of the used secret keys, if the key of the previous level turned out to be disclosed by third parties, it is enough to change the secret key to return the original level of protection.

It should be noted that encoding and, possibly, reproduction make it possible, on the one hand, to significantly increase the message's stability to distortion, and on the other hand, the ability to authenticate a document by evaluating or measuring the number or error rate detected when reading the label of the information matrix.

It should be noted that the steps of encoding, encrypting, scrambling, further scrambling and reproducing are reversible, provided that the secret key or secret keys are known.

When checking the original information matrices captured and printed with a resolution of 1200 dpi and containing cells of 8 × 8, 4 × 4, 2 × 2 and 1 × 1 pixels, it was noted that reading at high resolution the binary value represented by each cell:

- with a cell size of 8 × 8 pixels, it practically does not lead to errors,

- with a cell size of 4 × 4 pixels leads to a small number of errors,

- with a cell size of 2 × 2 pixels reproduces many errors and

- with a cell size of 1 × 1 pixel, the error coefficient approaches a maximum of 50%, which error correction will probably be ineffective, and the quality deterioration due to copying becomes invisible, since it is impossible to change the error coefficient.

The optimum is between the extreme cell sizes, and in the limited selection presented here, one of the cases when the cell size is 4 × 4 or 2 × 2 pixels is optimal. Below is a method for determining the specified optimum.

As shown in FIG. 3, according to the features of the invention, a document authentication method after an initialization step 305 comprises:

- step 310 of obtaining at least one secret key,

- step 315 of capturing the image label information matrix on the specified document, preferably using an array of image sensors, such as cameras,

- step 320 determine the location of the label information matrix,

- step 325 of sorting patterns to align and locate the cells of the information matrix in the specified label,

- steps 330 and 335 of descrambling message elements using a secret key to obtain a reproduced encoded encrypted message to perform the substitution, step 330, and permutation, step 335,

- step 340 of accumulating repetitions of the reproduced encoded encrypted message to obtain an encoded encrypted message,

- step 345 decoding the encoded encrypted message to obtain an encrypted message,

an optional step 350 of decrypting the encrypted message using the secret key,

- step 355 of determining the number of errors affecting the encrypted message using the redundancy associated with the message at the encoding stage and

- step 360 of deciding whether the document is marked with a label information matrix, a copy or the original.

Each secret key is preferably random or pseudo-random.

Thus, the authentication method optionally comprises the step of decrypting the original message using a symmetric or asymmetric encryption key. Depending on the type of encryption, i.e. using symmetric or asymmetric keys, the decryption key is identical to or different from the encryption key. Identical keys are used for symmetric encryption, while different keys are used for asymmetric. An important advantage of asymmetric encryption is that the decryption key does not allow the generation of a valid encrypted message. Thus, third parties who have access to the detector and managed to get the decryption key will not be able to use it to generate a new valid message.

In order to reproduce a mark formed on a document, firstly, it is captured by an image sensor, usually an array of camera image sensors, for example monochrome. The digitized captured image has a format, for example, a matrix of points (a format known as "bitmap" is a bitmap).

5 shows an example of a digitized captured image of 640x480 pixels containing eight binary data per pixel (i.e., a scale of 256 gray levels). Below is a detailed description of the sequentially used functions when reading. Firstly, the enumeration function is applied to the resulting image for each of the 25 patterns for alignment. The output of the specified function contains 50 integers of vertical and horizontal values of 25 patterns for alignment. The specified function is performed in two stages:

- the stage of determining the position of the information matrix as a whole in the digitized captured image and the stage at which local enumeration (enumeration into parts of the image) of each of the patterns is performed for alignment in order to determine their positions.

To perform the first step, a specialist can use a known method, for example, disclosed in US patent No. 5296690. Another simple and fast algorithm is to delimit the area of the digitized captured image containing the information matrix by carefully studying sharp transitions of shades of gray line by line and column by column, or after performing, firstly, summing all the lines and, secondly, summing all columns to create one row and one column in which to search. For example, grayscale levels with the highest absolute values correspond to the edges of the information matrix, and thus a rough estimate of the position of its four corners can be performed.

For the second stage, the estimated positions of the four corners of the information matrix are used to assess the position of the patterns for alignment by known geometric methods.

To determine the shift, scaling and rotation angle of the information matrix in the image captured by the image sensor, standard geometric methods can be used. Similarly, the indicated shift, scale and rotation angle can be used to determine the positions of the corners of the information matrix. Thus, a sequential approximation can be performed by repeating these two steps.

Generally speaking, the expected position of each alignment pattern is obtained with an accuracy of approximately X pixels along the horizontal and vertical axis. The indicated value of X depends on the application conditions, in particular, on the relationship between the capture resolution and print resolution, the maximum allowable reading angle, and the accuracy of the expected positions of the four corners of the information matrix. An acceptable X value is 10 pixels, in which case the search area has a size of 21 × 21 pixels. Collapse between the alignment pattern and the alignment block, the alignment block is possibly scaled if the ratio of the capture resolution to the print resolution is different from unity. The position of the resulting convolution matrix with the maximum value corresponds to the initial position of the alignment block.

The positions of 25 alignment patterns are stored in memory. The specified data set is used at the demodulation stage to determine the position of each cell of the information matrix in the captured image with maximum accuracy.

For each binary value, the closest alignment pattern of the corresponding cell is used as the starting point for the expected cell position in pixel units of the captured image. Using the expected rotation and scaling and the known relative position of the cell in the digital information matrix, the center position of the cell in the captured information matrix is calculated using known geometric methods.

The use of the descrambling function, the inverse scrambling function used to create the original information matrix, allows you to restore the original reproduced message damaged by errors. If the indicators are stored, they receive real or integer numbers that can be positive or negative and directly to which the exclusive OR function cannot be applied. Thus, to obtain the expected descrambled message from the indicators, it is necessary to multiply the indicator by -1 if the value of the exclusive OR filter is 0, and by +1 if its value is 1. It should be noted that the permutation is performed in the same way for different types of indicators (binary, integer and real).

The next step is to evaluate the value of each bit of the encoded message in accordance with the analysis of the captured cell values of the descrambled information matrix. To do this, the next step contains the definition of the binary value indicator assigned to the cell, assuming that black has a binary value of “0” and white “1” (or vice versa). The indicated indicator may be, for example, the average brightness (or the average value on a gray scale) of a small adjacent area surrounding the center of the cell (and most of all the corresponding cell area), or the highest value of the indicated small surrounding area, or the smallest brightness value in the specified adjacent area. An approach may be useful in identifying two adjacent regions, a small region surrounding the center of the cell, and a larger region surrounding and excluding the indicated smaller adjacent region. Thus, the indicator can be based on a comparison of the brightness values in the outer adjacent region and in a smaller adjacent region called the inner region. Measurement by comparison may be the difference between the average brightness in the inner adjacent region and the average brightness in the larger adjacent region.

After the binary value indicator has been determined for each cell of the information matrix, it is useful to perform additional processing of these indicators. Depending on the changes that the information matrix is experiencing, from the digital information matrix to the captured information matrix, indicators may represent a deviation. A simple additional processing in order to reduce this deviation consists in subtracting the average value or the value of the indicator occupying a middle position, and, possibly, in normalizing these indicators in the range of values from -1 to +1. The indicated normalized indicators can be used to determine the most probable binary values, by comparing them with a threshold value, for example, “0”, higher values are assigned the value “1”, and lower values are assigned “0”, which leads to the same number binary values "0" and "1".

According to a preferred embodiment of the invention, for each binary value sought, the indicator values are summed over all representations, and then compared with a threshold value. The specified processing, with more intensive use of resources, in essence gives greater reliability at this stage.

Returning to the above example, the information matrix contains 35 times encoded message with a length of 272 binary values, 80 of which are reproduced 36 times. Therefore, for each value of the encoded message, there are 35 or 36 indicators. Concentration or accumulation leads to the preservation of only one final value (binary, real or integer), depending on the specified set of representations of the same initial binary value. For example, averaging of 35 or 36 indicators is performed, while a positive average value is decoded as "1", and a negative average as "0". Thus, the average values of indicators can be compared with a zero threshold value.

According to variants of the invention, more complex methods of statistical processing are applied to indicators, which in some cases requires a training step. For example, in order to calculate the probability of the corresponding initial binary value “0” or “1”, respectively, non-linear operations can be performed on the average values of indicators. In fact, the estimated probability can provide a refinement of the result of the decoder.

At the end of the accumulation step, an encoded message is received containing redundancy, designed to enable correction or at least error detection.

The decoder, the principle of which in the case of a convolutional code, is preferably based on the Viterbi method, outputs an encrypted message, the length of which in the present example is 128 bits. The decoded encrypted message is then decrypted using the algorithm used for encryption, preferably AES for 128-bit blocks, in reverse order.

It should be recalled that part of the message can be reserved for storing the mathematical function of the rest of the message, for example a hash function. In the above example, 16 bits are reserved for storing the mathematical function of the remaining 112 bits of the message. Insignificant bits are added to the remaining 112 bits of the message and a hash value or digest of type SHA-1 of the message is added, to which insignificant bits are added, and the same secret key was used to create it. If the first 16 bits of the received hash value correspond to 16 reserved bits, the accuracy of the information is confirmed, and the reading process can proceed to the next step. Otherwise, a 112-bit message is considered invalid. The reasons for this invalidity may be different: incorrect reading, message generated in the wrong way, etc. The exact cause of the problem can be determined by a more detailed analysis, possibly with human intervention.

The decrypted 112-bit message is interpreted so as to provide the user with relevant information at the output. The specified message in itself can provide the user with important information about the nature of the document or medium containing the information matrix: about the period of use of the product, the distribution route or coordination with other information from the same document, etc. In addition, this information can be used to send a request to the database, which can provide new information, confirm validity, check the origin of the document or recognize a copy, etc.

However, as indicated above, reading and analyzing the transmitted message does not give an exact answer to the question, "is the document in question the original or a copy." In fact, a high-quality copy of the original document will contain a readable message containing information that is, in principle, valid. Even if the information extracted from the copy is considered invalid, for example, if the copy of the document is transmitted through the distribution network that does not match the information extracted from the information matrix, is it important to know the exact reason for the falsification and to recognize if the genuine or fake product passed through an unauthorized channel? The following is a description of the various methods for determining the source of a document (original or copy).

There are many decoders that provide measurement of the error rate in an encoded message. For example, for a convolutional code, the Viterbi detector calculates, based on a given metric, in the state space of the decoder, the shortest route leading to the encoded message in question.

The choice of metric depends on the representation of the encoded data entering the decoder. If the specified data is binary, the metric is based on the Hamming distance, namely, the number of positions or different bit values between the code received at the decoder input and the code corresponding to the shortest path in the state space. If the data is not binary, but has a more accurate measurement or is an integer or real number, use the appropriate metric.

Regardless of which particular metric is used to measure the error rate in a message, the specified error rate will theoretically be higher for a captured copy of the information matrix than for a captured original information matrix. To determine the type of information matrix (original or copy), a decision threshold is required. For example, you can use the following method to calculate the specified threshold:

- generate a typical sample application, for example, 100 different original information matrices, each of which is captured three times in the application to obtain a total of 300 captured images,

- for each of the 300 captured images measure the error rate,

- calculate the measured average value and the spread of the error coefficients of the sample, for example the arithmetic mean and standard deviation of the sample,

- in accordance with the measurements of the average value and the spread, the threshold for deciding on the error rate is determined, above which the information matrix will be considered as a copy. The specified decision threshold may be, for example, the average value +4 * standard deviation,

- in order to detect possible deviations when printing original information matrices, you can set the lower decision threshold equal to, for example, the average value - 3 * standard deviation, below which the user receives a warning about a particularly low error rate in the sample.

If the conditions for capturing information matrices are so inadequate that due to poor capture conditions this leads to an excessively high error rate, you can also specify an interval in which it is not possible to determine the source of the information matrix with confidence, then a request for re-image capture is performed. The indicated interval may lie, for example, between the mean value +2 * standard deviation and the decision threshold (in this example, equal to the mean value +4 * standard deviation).

The measurement of the error rate obtained at the decoding stage is theoretically calculated directly at the decoding stage, so its use is very practical. As indicated above, the indicated error rate of the encoded message is based on the accumulation, in our example, 35 or 36 indicators for each bit of encoded information. However, in some cases, it is desirable to conduct a deeper analysis of the error coefficient based directly on the values of the indicators, and not on the accumulated values of these indicators. In fact, a deeper analysis of the error rate may provide the ability to better recognize copies of information matrices.

For this, it is necessary to determine the position of errors for each of the indicated indicators. To do this, first of all, determine the original encoded message, which can be provided by the decoder. Otherwise, it can be computed by encoding the decoded message. It should be noted that this encoding step is particularly advantageous when using a convolutional code. Then, the encoded message is reproduced to obtain the original reproduced message. The original reproduced message can be compared with the previously received original reproduced message, damaged by errors, and measurement of the error rate can be calculated in a suitable metric. If a reproduced message damaged by errors is represented in binary values, the number of errors (equivalent to the Hamming distance) can be directly calculated and normalized by dividing it by the length of the reproduced message. If the indicator values are saved, the reproduced message damaged by errors is presented in whole or real numbers. In this case, the reproduced message can be likened to a vector, and a metric can be selected that provides the ability to calculate the distance between these vectors. For example, a linear correlation coefficient between two vectors, ranging from -1 to +1, is widely used to measure the similarity between vectors. It should be noted that the distance between the vectors can be calculated by simply changing the sign of the similarity measurement, in this case, changing the sign of the indicated linear correlation coefficient.

Obviously, many other methods of measuring distance can be used, while maintaining the essence of the method. Measuring the distance of the reproduced message provides a deeper analysis, at the level of the elementary units of the message, represented by the cells of the matrix. It may be necessary to analyze at an additional level of accuracy, separately considering the various geographical areas of the matrix. For example, you may need to analyze and determine the error rate in a certain area, for example, in the upper left corner of the matrix. This feature is of particular interest in the case, for example, if the information matrix was locally damaged (scratch, bend, abrasion, stain, etc.) or if it was unevenly captured (too dark or too light areas, or not focused).

In fact, it is required to avoid the indicated distortions, which can damage the original information matrices, resulting in a high error rate for these matrices. Therefore, the analysis of local elements may provide an opportunity to neglect or give an underestimated weight value to damaged areas having an increased error rate.

The indicated problem can be solved by considering a message with rearranged bits or a scrambled message instead of a reproduced one. In fact, the information matrix is generated in a fixed way (regardless of the key) from the scrambled message and alignment blocks, therefore, it is easy to extract portions of the message with the rearranged bits or the scrambled message corresponding to the exact geographical areas. It should be noted that if a message with rearranged bits is used instead of a scrambled message, this avoids the step of imposing an exclusive-OR filter to obtain a scrambled original message.

For an arbitrary geographical area, the above measurements of the distance between the original message with the rearranged bits or the scrambled original message and the message with the rearranged bits or the scrambled message damaged by errors can be applied. In all cases, alignment blocks can be included in the analysis.

You can use many algorithms that regulate the use of different geographical areas. In addition, in some cases, the participation of a human operator is possible, which may be able to determine the cause of the distortion (accidental, intentional, systematic, etc.) However, often the analysis should be performed automatically and generate a special result: original, copy, error reading etc Thus, the general approach includes dividing the information matrix into special sections of the same size, for example, 25 sections of 22 × 22 pixels for the 110 × 110 pixels matrix from the above example. Therefore, 25 distance values between the original message and the message damaged by errors corresponding to the specified individual geographical areas are calculated. Then, the eight highest distance values corresponding to the eight least damaged areas are retrieved. Finally, the average error rate for the indicated eight geographic areas is calculated. Thus, preference is given to sections of the information matrix, the probability of which is correctly read is the highest.

It should be noted that, since several error rate indicators can be calculated in accordance with an encoded, reproduced, rearranged bit or scrambled message, and in accordance with different geographical areas, it is possible to group different measured error rates in order to make a global measurement of the error rate .

Based on the binary values of the message with a length of 255 binary values, the decoder determines the decoded message and the number or error rate. In the case of using decoders that do not determine the number or error rate, the decoded message is re-encoded and compared with the message received from the captured information matrix.

From the binary values, according to the number of errors detected, it is determined whether the captured analog information matrix is an original or a copy.

If the message can be decoded and error positions can be determined, at the output of the decoding step, a list of 255 binary values equal to "1" for errors and "0" if there is no error in the corresponding binary value of the decoded message is obtained.

It should be noted that the number of errors that can be decoded is limited until a decoded message is determined, it is known that the number of errors is greater than the specified detection limit. When a message is decoded, it is decrypted using a secret key. It should be noted that the use of asymmetric keys makes it possible to increase the security of this stage.

According to the experience of the inventors, print parameters as a result of physical tolerances of the marking means used, the surface condition of the document and possibly the coating, generating at least 5 percent, and preferably from 10 to 35 percent, and even more preferably from 20 to 25 percent of incorrectly printed characters provide a good level of copy recognition. To achieve the indicated error rate, print parameters are varied that affect the quality of the printed message.

The following is a more detailed description of how the concept of the secure information matrix is optimized for the printing environment.

Further, the optimization of creating secure information matrices depending on printing conditions is described in more detail.

It should be noted that the not yet printed secure information matrix in digital form does not contain errors. In fact, there is no random, intentional, or “artificial” error generation. In addition, these cases are not printing errors according to the present invention, because “printing error” refers to a change in the appearance of a cell modifying the presentation of the information that the cell carries when analyzing to avoid reading or capturing errors, for example, under a microscope. It should be noted that if initially the cells often have binary values, then the captured images often contain shades of gray, and therefore, a value is associated with the cell that is not binary and that can be interpreted, for example, as the probability that the cell will have the original binary value.

Thus, the error contains the printed version of the secure information matrix. Considered errors used in the present invention are not caused artificially, they occur naturally. In fact, the errors in question arise randomly and naturally at the stage of marking when printing a protected information matrix with a fairly high resolution.

These errors are necessary, even if they form a complex combination. In fact, if the information matrix is applied without errors (or with a very low error rate), then a copy of the same information matrix, made under comparable printing conditions, will not contain more errors. Thus, it is obvious that the information matrix printed “almost completely” can be accurately copied using a similar labeling tool. On the contrary, if the information matrix is plotted with as many errors as can be expected, only a smaller part of the cells can be copied with additional errors. Therefore, it is necessary to avoid too high a resolution when marking, as this reduces the ability to distinguish originals from copies.

Obviously, the print resolution of the protected information matrix cannot vary. In fact, most print media are printed in binary code (a dot of printing ink is present or absent) with a fixed resolution, shades of gray or other colors are copied using various screening methods. In the case of offset printing, the indicated “native" resolution is determined by the resolution of the printing form, which is, for example, 2400 dpi (2400 dpi). Thus, an image in grayscale intended for printing with a resolution of 300 pixels per inch (300 ppi) is, in practice, printed in binary code with a resolution of 2400 dots per inch (2400 dpi), with each pixel corresponding to approximately 8 × 8 dots.

Despite the fact that in general the print resolution cannot vary, on the other hand, it is possible to change the cell size of the protected information matrix in pixels so that one cell is represented by several printing points and, according to embodiments of the invention, a part of each cell, the appearance of which is in binary information matrices can vary, i.e. be printed in white or black. Thus, it is possible, for example, to represent a cell as a square block of 1 × 1, 2 × 2, 3 × 3, 4 × 4 or 5 × 5 pixels in size (blocks of non-square shape are also possible), corresponding to a resolution of 2400, 1200, 800, 600 and 480 cells per inch respectively.

In accordance with some features of the present invention, the number of pixels in a cell is determined, which leads to a natural deterioration in print quality, which allows the discrepancy between originals and copies to be maximized.

The following model allows you to get an answer according to the specified definition, even if it is a consequence of the simplification of the process used. Suppose that a digital secure information matrix consists of n binary cells and that there is a probability p that each cell is printed with an error (for example, “1” will be read as “0” and vice versa).

Further, suppose that the copy will be made using equivalent printing means, which is expressed with the same probability p of error in the cells during copying. It should be noted that the probability p of an error greater than 0.5 in the context of this model does not make sense, since in this case the correlation between the printed information matrix and the digital information matrix is zero (therefore, a value of 0.5 corresponds to the maximum distortion).

Based on the captured image, the detector calculates the number of errors (the number of cells that do not correspond to the initial binary value) and, based on the specified number of errors, makes a decision about the origin of the protected information matrix (original or copy). It has been established that in practice the captured image in most cases consists of shades of gray scale, thus, to obtain binary values, it is necessary to compare the values of cells with a threshold value.

In order to prevent information loss at the stage of comparison with the threshold value, the values presented in shades on the gray scale can be interpreted as probabilities of binary values. However, it is further discussed how binary values of the cells of the protected information matrix are set from the obtained image.

To measure the reliability of recognition of copies in accordance with the probability p of error in each cell, indicator I is used, which is equal to the difference between the average number of errors for copies and for originals, normalized to the standard deviation of the number of errors in the original. Thus, we obtain the following formula: I = (Ec-Eo) / So, where

Eo - the average number of errors in the originals,

Ec - the average number of errors in copies,

So is the standard deviation of the number of errors in the originals.

It should be noted that due to the simplicity of the model, the standard deviation for the copies is neglected. Since in the indicated model there is a probability p that each cell is printed with an error, one can apply formulas for the mean and standard deviation of the binomial distribution. Therefore, we find the values of Eo, Ec and So in accordance with p and n:

E _o = n · p

E _c = 2np (1-p)

Therefore, the value of indicator I is equal to:

, или 19,1% несвязанных ошибок печати.In Fig.20, the solid line 700 shows the value of the indicator I in accordance with p for p from 0 to 0.5, normalized in the range from 0 to 1. Thus, the following can be noted: for p = 0 and p = 0.5, those. for the minimum and maximum error rates, the indicator value is zero, which results in the absence of discrepancies between the originals and copies. In fact, in the absence of any distortion of the cells during printing, it is not possible to distinguish the originals from the copies, and, conversely, if the distortion is very strong (i.e. close to 0.5), there are practically no cells available for distortion, and as a result, the ability to distinguish originals from copies is reduced. Therefore, it is normal that the indicator passes through an optimum that corresponds to the value

, or 19.1% of unrelated print errors.

The inventors have discovered the optimum distortion, independent of the number n of available cells. In addition, it was found that the indicator I increases in accordance with n: therefore, it is necessary that n be as large as possible. However, it is relatively common that a printable secure information matrix is a fixed area, for example 0.5 × 0.5 cm. Thus, a matrix of 50 × 50 cells of 8 × 8 pixels each occupies the same area as the matrix of 100 × 100 cells sized 4 × 4 pixels. In the latter case, there are four times as many cells, but it is very possible that the probability of p errors will be higher. Thus, the determination of the optimal value of p should be carried out taking into account the fact that for a higher resolution, more cells are used. If we make the assumption that the probability p is inversely proportional to the area available for the cell, then p = α · n, where α is a constant, since the total area is divided by the number n of cells. Therefore, indicator I is expressed as follows:

, или 27,1% несвязанных ошибок.As shown in FIG. 20 on a curve with a dashed line 705, taking into account the change in p depending on n, the indicator passes through a maximum at a p value equal to

, or 27.1% of unrelated errors.

Thus, the use of an error coefficient lying in the range from 20 to 25% is preferable since this interval is between the above optima equal to 19.1% and 27.1%. An optimum of 19.1% corresponds to a case in which the number of cells is fixed, for example, when the reading technique allows only information matrices with a fixed number of cells to be read, while an optimum of 27.1% corresponds to a case in which the number of cells is not limited, while the physical the size of the information matrix is limited.

The following are options or improvements using some of the features of the present invention.

1) Use of non-binary information matrices. The implementation of information matrices is not limited to binary information matrices. At all stages, to move from the initial message to the information matrix, message elements can have more than two different values. Consider the case where the cells of the information matrix can take 256 different values, which corresponds to printing images in shades of gray, taking a value between 0 and 255. The scrambled and encoded message will also take 256 values. To extract the scrambled and encoded message from the encoded information, the permutation function can remain the same, while the exclusive-OR function can be replaced by summing 255 modulo, and the pseudo-random sequence used for the specified summing 256 modulo also contains values lying between 0 and 255.

The original message and part of the encoding corresponding to the application of the error correction code can also be represented (but not necessarily represented) by binary values. However, in the playback sub-step, it is necessary to convert the binary encoded message to the reproduced one, containing values, for example, from 0 to 255 (8 bits). To do this, in one of the options group the binary encoded message into units of 8 consecutive bits with the subsequent presentation of these units on a scale from 0 to 255.

2) Definition of copies based on the result of decoding, without calculating the error rate. According to embodiments of the invention, with reference to the accompanying drawings, an error rate is used to determine the origin of the captured information matrix (original or copy). In addition, as indicated above, the error rate can only be measured if the captured encoded message of the information matrix can be decoded. The description discloses the necessary steps to enable decoding of a message, in most cases up to the desired error rate. Thus, you can verify that in most cases the error rate can be measured for copies, provided that they contain a sufficient number of errors.

In some cases, they do not rely solely on the error rate in order to determine if the information matrix is a copy. This happens, in particular, in the case when the amount of information inserted into the information matrix is too large compared to the available area or the available number of pixels, as a result of which the encoded message cannot be reproduced repeatedly (in our example, the encoded message is reproduced 35 or 36 time). Thus, they strive to make sure that the original information matrices are read correctly and, conversely, copies of information matrices are in most cases read incorrectly. Correct reading allows you to make sure that the information matrix is the original, on the contrary, incorrect reading is not necessarily a guarantee that the information matrix is a copy.

The amount of information is high if the message is encrypted with an asymmetric key, for example, using an RSA public key encryption algorithm with an encrypted message length of 1024 bits, or in the case of symmetric encryption of the photo of the owner of the identification card (from 2000 to 5000 bits). If the size of the information matrix is limited (for example, less than 1 sq. Cm), multiple reproduction of the encoded message is not possible, and in this case, depending on the print quality, the copy message will probably be unreadable.

3) Use of information matrices containing several messages. It is possible to create information matrices containing several messages using different keys in a recursive manner. This is particularly useful for applications in which different levels of authority are assigned to different verification tools or different users. In addition, it is useful to provide several levels of protection: if the public key is disclosed by third parties, only part of the information matrix can be faked.

To simplify the selection of these embodiments of the invention, consider two messages (messages 1 and 2). Messages 1 and 2 can be grouped at several levels, for example:

- consistently connect messages 1 and 2, separately encrypted with keys 1 and 2. Key 1 (or group of keys 1) is used at the stages of permutation, scrambling, etc. Message 2 can only be decrypted by readers that are equipped with a key 2. Authentication to determine whether the information matrix is an original or a copy can be performed entirely from the matrix using key 1. This approach provides benefits if the image is captured by a portable device, connected to a remote server equipped with a key 2, and the connection is expensive, time consuming and difficult to establish a connection: as a result, the amount of data transmitted x is not very important,

- the scrambled message 1 and the scrambled message 2 are connected in series, and the information matrix is formed from the serially connected scrambled messages 1 and 2. It should be noted that in the information matrix the two indicated messages are physically separated.

- the scrambled message 1 and the scrambled message 2 are connected in series, and the series-connected message is rearranged and scrambled using the key 1. It should be noted that the positions of the scrambled message 2 depend on both keys 1 and 2, as a result of which both keys are required to read message 2.

The use of several secure messages with different keys makes it possible to control different levels of authority of different users of the verification module. For example, some departments are authorized to read and authenticate the first message, while others can only authenticate the first message. A stand-alone verification module that does not have access to the server for verification is generally not capable of reading or authenticating the second message. Obviously, many other options are also possible. It should be noted that the above considerations can be extended to information matrices containing more than two messages.

4) Insert code to detect errors or fakes. The higher the error rate, the higher the risk that the message is incorrectly decoded. It is desirable to have a mechanism for recognizing incorrectly decoded messages. Sometimes this can be done at the application level: an incorrectly decoded message is not consistent. However, it is not possible to use the value of the decoded message to verify its validity. Another approach is to assess the risk that the message is incorrectly decoded using a signal-to-noise ratio measurement of the encoded message, relative to the type of code used and decoding. There are graphs, especially in the book “Interference Protection Coding”, Second Edition, Lin and Costello (“Error Control Coding”, Second Edition, Lin and Costello). For example, on page 555 of this book it is shown that for a convolutional code with a memory of 8 and a factor ^of _1/2 , with software decoding with continuous data input, the error rate per encoded bit is 10 ^-5 for a signal-to-noise ratio of 6 dB.

Another approach, which may supplement the previous one, is to add the hash value of this message to the encrypted message. For example, you can use the SHA-1 hash function to calculate the number of hash bits in an encrypted message. The specified hash bits are added to the end of the encrypted message. Upon detection, the hash bits of the encrypted message are compared with the attached hash bits. If they match, we can conclude with a high probability that the message is correctly decoded. It should be noted that if there are 16 hash bits, there is one chance out of 2 ¹⁶ that the error was not detected. It is possible to increase the number of hash bits, but due to the number of cells available to play the encrypted message.

5) Hashing can be used to add a layer of security. In fact, suppose that symmetric encryption is used and that third parties manage to take possession of the encryption key. The specified attacker can generate an unlimited number of valid information matrices. However, it should be noted that if the additional scrambling key was originally used, and it did not become the property of third parties, the information matrices generated by third parties will be recognized as copies by a detector equipped with the specified additional scrambling key. However, the hash value of the raw or encrypted message, depending on the key theoretically not stored in detectors potentially accessible to third parties, can be sequentially attached to the encrypted message. Verification of the hash value, possibly on secure readers, makes it possible to verify that a valid message has been generated. Thus, third parties equipped with an encryption key, but without a hash key, are not able to calculate the valid hash value of the message. In addition, the specified valid hash value makes it possible to generally ensure the integrity of the information contained in the message.

6) The use of information matrices in the server system. Processing is completely carried out by the server, remote from the means of image capture, or in the case of authentication by a stand-alone reader, possibly containing a set of secret keys.

According to a preferred embodiment of the invention, the server reads the message and the portable reader provides copy recognition.

Preferably, part of the steps 320-350 of the algorithm for reproducing the original information is performed by a reader at the image capture site of the information matrix, and another part of the playback step is performed by a computer system, for example, a server remote from the image capture site of the information matrix. Thus, data related to the creation and reading of information matrices (keys and associated parameters) can be stored in one place or on the same server with a strict protection mode. Authorized users can connect to the server after authentication in order to order a certain number of information matrices that will be applied to documents that must be protected and / or monitored. These information matrices are generated by the server, the keys used are stored on the server. They are transmitted to the user or directly to the printing press in a secure manner (for example, using encryption).

In order to carry out technical control directly on the production line, capture modules (sensor + processing software + information transfer means) allow the operator to capture images of printed information matrices, which are automatically transmitted to the server. The server determines the keys and corresponding parameters, reads and authenticates the captured information matrices and returns the result to the operator. It should be noted that this method, in addition, can be automated using industrial cameras, automatically capturing the image of each printed information matrix, skipped on the production line.

If portable field capture devices can connect to the server, a similar method can be used for reading and / or authentication. However, the specified connection is not always desirable or possible, in these cases some of the keys must be stored in the authentication device. Thus, it is especially advantageous to use a partial scrambling key, because if it is not stored in a portable reader, then said reader does not have enough information to create the original information matrix. Similarly, if encryption is performed in an asymmetric manner, the decryption key stored in the portable reader is not capable of encryption and, therefore, of generating an information matrix containing a different message that would be valid.

In some areas of application, the server for distribution and certification of information matrices must cope with a large number of different "profiles" (a profile is a unique key-parameter pair). In particular, this occurs when the system is used by various companies or institutions wishing to protect their documents, products, etc. You can see the advantage in that different users have different keys: usually the information contained in the information matrix is confidential. Therefore, the system may have a large number of keys for management. In addition, as is known in cryptography, it is preferable to update keys at regular intervals. Key multiplication should be clearly considered from the point of view of verification: in reality, if the verification module does not know in advance which key was used to generate the matrix, there is no other way out than checking all the keys capable of this one by one. In this example of the use of the invention, it is preferable to include two messages in the information matrix, using different keys for each. Therefore, you can use the fixed key for the first message so that the verification module can directly read and / or authenticate the first message. To read the second message, the first message contains, for example, an indicator that allows the verification module to request a secure database that can provide it with keys for reading and / or authentication of the second message. In general, the first message contains general information, while the second message contains confidential information that can be personalized.

7) Detection Threshold / Print Settings. In order to facilitate the autonomous authentication of the information matrix, the decision threshold or decision thresholds or other parameters related to printing can be stored in the message or messages included in the information matrix. Thus, it is not required to query the specified parameters in the database and store them in stand-alone verification modules. In addition, this makes it possible to control applications or information matrices that are of the same nature in terms of scope, but printed in various ways. For example, information matrices printed on the same type of document, but printed on different machines, could use the same key or the same keys. They may contain print settings stored in the corresponding messages.

8) The above permutation of a reproduced message can be expensive. In fact, for permutation, the maximum number of pseudo-random numbers must be generated. In addition, in the recognition process in some applications, a plurality of scrambled messages can be calculated from the captured image, so that the smallest measured error rate is calculated on this set of scrambled messages. In addition, each of these scrambled messages must be rearranged, which will cost even more, since you have to deal with a large number of scrambled messages.

The cost of this permutation can be reduced by grouping a certain number of neighboring units of the reproduced message, followed by the permutation of these grouped units. For example, if the reproduced message contains binary values and consists of 10,000 elements and if the indicated units are grouped in pairs, we get 5,000 groups, each of which is able to take four different values (in the quaternary representation). 5000 groups are permuted, then the quaternary values are represented in two bits before using the exclusive-OR function and / or modulation. According to embodiments of the invention, the exclusive-OR function is replaced with a modulo summation (as disclosed in MIS 1 patent), then the values are re-translated into a bit representation.

For coded messages with a length multiple of two, the number of grouped units can be an odd number, for example 3, in order to avoid the fact that two adjacent bits of the encoded message were always adjacent in the protected information matrix. This enhances message security.

During the reading process, the groups of values or these values accumulated on a single object are reversed, so as to ensure their sequential separation.

The following describes a method for optimizing the printing parameters of digital watermarks. As an example, spatial digital watermarks are considered.

Digital watermarks use masking models to predict the number of possible modifications in the image that will be invisible or at least acceptable from an application point of view. Thus, these modifications will be set in accordance with the content of the image and, therefore, typically will be larger in the textured or bright areas, since the human eye “masks” the differences to a greater extent in these areas. It should be noted that digital images intended for printing can be modified so that modifications appear visible and distort the digital image, but become invisible or less distorting the image after printing. Therefore, suppose that for a digital image in grayscale or in a color consisting of N pixels, the masking model makes it possible to calculate the amount by which the grayscale or color in each pixel can be changed in a way that is acceptable for a given application. It is noted that the frequency masking model by a specialist can be easily adapted to the derived spatial masking values. In addition, suppose that a spatial model of a digital watermark in which an image is divided into blocks of pixels of the same size and a message element is inserted into it, for example, one bit of a watermark in each block by increasing or decreasing the gray or color value in each pixel to allowable maximum or minimum, increase or decrease are created in accordance with the inserted bit. It should be noted that the watermark bits may, for example, be equivalent to the scrambled message of the secure information matrix.

In order to determine whether the captured image is an original or a copy, based on the error rate in the message, measured by the number of incorrectly detected message elements. It should be noted that this requires that the message be read correctly, which implies the insertion of a sufficient amount of redundancy.

Many methods are known for measuring the value of a bit stored in an image block, which use, for example, a low-pass filter or a band-pass filter, normalizing values across the entire image surface or area. Usually get a non-binary value, continuous, positive or even negative. The indicated value can be compared with a threshold to determine the most probable bit, and the error rate can be measured by comparison with the inserted bit. In addition, you can save the values and measure the correlation coefficient, from which, as indicated above, the error coefficient is derived.

It should also be noted that the error rate of the message can be measured indirectly, using the method of recognition of copies without reading the message, as indicated elsewhere.

For a specialist it is obvious that the larger the block size in pixels, the lower the error rate in the message. On the other hand, message redundancy will be lower. Depending on the quality and resolution of the print, the specialist will determine the block size, which allows to obtain the best compromise between the error rate of the message and redundancy, in order to maximize the likelihood that the message will be correctly decoded. On the other hand, the prior art has not solved the problem of selecting the size of cells in order to optimize the recognition of copies. This problem is solved due to some features of the present invention.

Here, the theoretical model used above to determine the optimal error rate of the digital authentication code can be used. In fact, each block can be considered as a cell with the probability p to be deteriorated, and it is possible to search for the optimum p in the case when there is a fixed physical cell size (in fact, the image intended for printing has a fixed pixel size and resolution). Moreover, it can be approximately assumed that the probability p is inversely proportional to the area available for the cell. Moreover, it was found that indicator I is maximum at p = 27%. Other models are possible that lead to a different optimum.

To determine the block size that is optimal for copy recognition, you can use the following steps:

- get at least one image characterizing the image used in this application,

- using a masking model, calculate for each pixel of each image the maximum discrepancy that can be made,

- for various sizes of the tested blocks, for example, 1 × 1, 2 × 2, ... up to 16 × 16 pixels, generate at least one message whose size corresponds to the number of image blocks,

- in each image insert each of the messages corresponding to each block size, to obtain images with a mark,

- at least once each of the images is printed with a mark printed in the printing conditions of the application,

- at least once capture each of the images with the mark,

- read the watermark and determine the error rate for each of the captured images, group the measured error rates by block sizes and calculate the average error rate for each block size and

- determine the size of the block, the average error rate of which is closest to the target value, for example 27%.

The following describes how to optimize AMSM print settings.

AMSMs consist of points distributed in a pseudo-random manner with some density that is low enough to make it difficult to detect the exact location, for example with a density of 1%. The index corresponding to the maximum cross-correlation between the original and captured AMSM corresponds to the energy level of the signal and theoretically will be lower for copies. It should be clarified that if the copy is “slavish,” for example a photocopy, it is likely that a large number of dots, already weakened by the first printout, will completely disappear during the printing process, therefore, in this case it will be very easy to recognize the copy whose energy level is much weaker. On the other hand, if before printing copies are applied a programmable image designed to identify points and restore them to their original energy level, such a copy will have a significantly higher energy level and indicator. To reduce this risk and ensure the maximum discrepancy in the indicator between copies and originals, they should be printed with a resolution or dot size that maximizes the difference in energy levels. However, solutions to this problem are unknown from the prior art, and AMSMs are often created in a less optimal way with respect to copy recognition.

A simple argument allows us to conclude that ideally, the AMSM dots should be sized so that approximately 50% of these dots “disappear” during the initial printing process. The term “disappear”, as used in the present description, means that an algorithm trying to locate and restore points can correctly detect only 50% of the source points.

In fact, suppose that on average, p percent of the dots disappear when printing the original. If the copy is made under the same printing conditions, p percent of the remaining points will also disappear, as a result of which the percentage of points that have disappeared will be p + p * (1-p).

Using the above criterion, according to which they try to maximize the discrepancy between the originals and copies, normalized to the standard deviation of the originals, equal to p * (1-p), the following criterion C is brought to a maximum depending on p:

, где N - фиксированное количество точек в AMSM.

where N is a fixed number of points in AMSM.

It was found that the maximum value of C corresponds to p = 0.5.

The above model is used in cases where the number of points is fixed. On the other hand, if you want the pixel density to be fixed, for example 1% of the pixels marked, and if the dots contain several pixels, you can use a larger number of N points per preset density.

If we determine the density d and the number of pixels per point m, we obtain the ratio:

Having made the assumption that the probability that the point will disappear can be approximately expressed as inversely proportional to the size of the point in pixels, we obtain:

, где а - постоянная.

where a is a constant.

Depending on p, d, a and m C is expressed as follows:

It was found that if d and a for a given application are constant, the maximum value of C corresponds to p = 2/3, or 66.6%.

To implement the invention, you can use the following steps:

- at a fixed density (black pixels) print AMSM patterns containing dots of various sizes, for example 1 × 1, 1 × 2, 2 × 2, etc.,

- capture at least one image of each of the various AMSM patterns,

- for each AMSM determine the number of correctly identified points, and measure the error rate and

- choose the parameters corresponding to the AMSM pattern having an error rate that is closest to the optimal for the selected criterion, for example 50% or 66%.

It should be noted that if AMSM contains a message, error correction codes should be adjusted to reflect the indicated maximum error rate. In addition, it should be noted that if the detector takes the maximum energy level as the basis, the copy index can be artificially inflated by printing correctly defined points in such a way that they make the maximum contribution to measuring the energy level of the signal.

Finally, other criteria for determining the optimum are possible, taking into account, for example, the density of points, the number of pixels having deviations of position, shape or size, the number of correctly colored pixels in each cell, etc.

It should be noted that VCDP processing (short for Variable Characteristic Dot Pattern - a pattern of dots with variable characteristics) can be performed in a similar way, it being understood that cells damaged by printing or copying errors do not necessarily change the appearance from presence to absence, but these errors may modify their positions, sizes or shapes, which are modified in accordance with the information provided.

VCDP (a pattern of points with variable characteristics) is produced by generating such a distribution of points in which:

- at least half of the points of the specified distribution do not come in contact with four other points of the specified distribution of points and

- at least one size of at least one part of the points of the specified distribution of points has the same amplitude order as the average value for the absolute value of the specified random deviation.

Thus, this makes it possible to use the individual geometric characteristics of the plotted points and measure the deviations of the characteristics of the indicated points, so as to integrate them into the metric (i.e., to determine whether they satisfy at least one criterion used in the measurement), making it possible to distinguish originals from copies or fakes.

Preferably, more than half of the points in the distribution of points do not touch any other point of the specified distribution. Thus, in contrast to protected information matrices and patterns for copy recognition and similarly to AMSM and digital watermarks, patterns of dots with variable characteristics provide the insertion of invisible or subtle labels. In addition, these tags are easier to integrate than digital watermarks or AMSM. They provide a more reliable way to recognize copies than digital watermarks, and they can be individually characterized in the process of static printing, which provides the ability to uniquely identify each document.

According to embodiments of the invention, points are created, at least one of the characteristics of which is variable, and the geometric amplitude of the generated deviation has the same bit category as the average size of at least a part of the points. Therefore, this makes it possible to generate and optimally use images of patterns from dots with variable characteristics, also called "VCDPs", designed to make copying more difficult and even impossible by accurate reproduction.

According to embodiments of the invention, the generated deviation corresponds to:

- deviation of the positions of the points in at least one direction with respect to the position in which the centers of the points are aligned in parallel lines perpendicular to the specified direction, and separated from each other in the specified direction by a distance equal to at least one size of these points. Thus, this makes it possible to use the exact characteristics of the position of the points and measure very small changes in the exact position of the points, so that they are included in the metric, making it possible to distinguish originals from copies,

- deviation of at least one size of the points in at least one direction relative to the average size of these points in this direction,

- deviation of the shape of the points in relation to the average shape of the points in the specified direction.

The distribution of points may be a representation of a coded unit of information, thereby allowing storage or transmission of information in a distribution of points with variable characteristics. With the same amount of information contained, point distributions can occupy a substantially smaller area than AMSM, for example, several square millimeters, which allows them to be captured by portable devices with high resolution and, therefore, high reading accuracy is ensured.

The following describes how, by measuring the number of errors in a message, you can decide on the authenticity of a document in accordance with the specified number of errors. To do this, it is theoretically necessary to decode the specified message, because if it is unreadable, it is impossible to determine the errors that damage it. However, if the labeling has significantly impaired the quality of the message, especially copies, or if a large amount of information is transmitted, it is possible that the message will be unreadable, in which case it is impossible to measure the error rate. It would be desirable to be able to measure the number of errors without decoding the specified message.

Secondly, at the stage of decoding information, algorithms are used that can turn out to be expensive. If only authentication of the message is required, without reading it, the decoding operation is carried out only with the aim of measuring the error rate, it would be preferable to exclude this step. In addition, if a deeper analysis of the error rate is required, it is necessary to restore the reproduced message. Recovering a reproduced original message can be costly; it would be preferable to avoid this.

However, according to one of the features of the present invention, it was found that to measure the number of errors is not required, and it is paradoxical, neither the restoration of the reproduced original message, nor even its decoding. In fact, the number of errors in a message can be measured using some of the properties of the message itself when evaluating the encoded message.

Consider the case of a binary message. An encoded message consists of a sequence of bits that are duplicated, then scrambled, and the scrambled message is used to create a secure information matrix. Scrambling typically includes a permutation and, optionally, the use of an exclusive OR function, and generally depends on at least one key. Thus, each bit of the message can be represented several times in the matrix. As shown in the example of FIG. from 1 to 5V, one bit is repeated 35 or 36 times. At the stage of accumulation of the encrypted message, all indicators of the value of each bit or message element are accumulated.

Thanks to this operation, the statistical error of the bit value is usually significantly reduced. Therefore, this estimate, regarded as the true value of the bit, can be used to measure the number of errors. In fact, if the applied matrix contains relatively few errors, all these errors are mainly corrected at the summation stage, and therefore, the recovery of the encoded message is not required, since an error-free version is already available. In addition, if some bits of the encoded message were incorrectly estimated, they usually have a reduced effect on measuring the number of errors.

The following is an algorithm for measuring the number of errors for binary data without decoding the message:

- for each bit of the encoded message summarize the values of the indicators,

- by comparing with the threshold value, the most probable bit value ("1" or "0") is determined, the most probable estimate of the encoded message is obtained, and

- calculate the number of indicators corresponding to the bit estimate of the corresponding encrypted message (for a single cell, this may be a density or a normalized brightness value). Thus, it is possible to measure an integer number of errors or the ratio or proportion of bits damaged by errors.

As an alternative to the last indicated step, it is possible to store the value of the indicator in memory and measure the overall similarity coefficient between the values of the indicators and the corresponding evaluation bits of the encoded message. The similarity coefficient can be, for example, the correlation coefficient.

According to one embodiment of the invention, it is possible to establish a correspondence with a weight coefficient indicating the probability that each estimated bit of the encoded message is correctly estimated. The specified weighting factor is used to weight the contributions of individual indicators in accordance with the probability with which the associated bit is correctly estimated. The simplicity of this approach is that the accumulated values corresponding to each bit of the encoded message are not compared with a threshold value.

It should be noted that the more noisy the message, the higher the risk that the estimated bit of the encrypted message is erroneous. This leads to the appearance of a systematic error, which consists in the fact that when measuring the number of errors, the true number of errors is underestimated. The indicated systematic error can be estimated by statistical methods and corrected when the number of errors is measured.

It is interesting to note that using this new approach to measuring the number of errors, the protected information matrix can be authenticated without the need for recognizing, directly or indirectly, the messages necessary for its creation. Simply, groups of cells that share common properties should be known.

According to embodiments of the invention, several sets of indicators are obtained that correspond to various pre-processing methods applied to the image, for example, transforming a histogram, or reading a protected information matrix at different positions, for each set of indicators, the number of errors is calculated and the lowest error rate is stored in memory. To speed up the calculations, the evaluation of the encoded message can be performed only once (a low probability that this estimate changes for each set of indicators).

We can assume that images (or matrices) are generated, parts of which have common properties. In the simplest case, parts of cells or pixels have the same meaning, and they are distributed in the image in a pseudo-random manner in accordance with the key. The property in question does not have to be known. When reading, it is not necessary to know this property, since it can be evaluated. Thus, the measurement of the indicator, which serves as a sign of authenticity, does not require either reference to the original image or definition of the message. Therefore, according to embodiments of the invention, the following steps are used to authenticate a document:

- obtaining a set of subgroups of image elements, for example pixel values, each subgroup of image elements using the same characteristic, these characteristics are not necessarily known,

- image capture,

- measuring the characteristics of each image element,

- assessment of the characteristics common to each subgroup of image elements,

- measuring the correspondence between the specified estimates of the characteristics common to each subgroup and the specified measured characteristics of each of the image elements, and

- making a decision on the authenticity in accordance with the specified measurement of conformity.

According to other variants of the invention, the description of which is presented below, for authentication of a digital authentication code, there is no need to either know or reproduce the original image, or to decode the message that it carries. In fact, when creating, it’s enough to create an image containing subgroups of pixels having the same value. In recognition, it is simply required to know the positions of pixels belonging to each of the subgroups. A property, for example, the value of pixels belonging to the same subgroup, should not be known: it can be found in the process of reading without the need for decoding the message. Even if the specified property has not been correctly defined, digital authentication codes can still be authenticated. This new type of digital authentication code is hereinafter referred to as a “random authentication pattern” (“RAP”). The word "random" means that inside the RAP can take any of a given set of possible values, and after creating the image this value is not saved.

For example, suppose there is a digital authentication code consisting of 12,100 pixels, i.e. a square of 110 × 110 pixels. The indicated 12100 pixels can be divided into 110 subgroups of 110 pixels each, so that each pixel is strictly in one subgroup. The pixels are divided into subgroups in a pseudo-random manner, preferably with a cryptographic key, so that without the key it would be impossible to find out the positions of the various pixels belonging to the subgroup.

After 110 subgroups are determined, the pixels of each subgroup are assigned a random or pseudo-random value. For example, for binary pixel values, you can assign the value "1" or "0", a total of 110 values. In the case of randomly determining values, 110 bits are generated using a random number generator, which can then be saved or not saved. It should be noted that for this division into subgroups, there are ^2,110 possible random authentication patterns (RAPs). In the case of pseudo-random value generation, a pseudo-random number generator equipped with a cryptographic key is used, which is then usually stored. It should be noted that for a similar generator based on the SHA-1 hash function, the key length is 160 bits, while in this example only 110 bits need to be generated. Thus, said generator may have limited use.

Knowing the value of each pixel, then you can collect the image, in this case, the size of 110 × 110 pixels. The image can be an ordinary square, with the addition of a black border to facilitate its detection, or have an arbitrary shape, contain microtext, etc. It is also possible to use groups of pixels with known values that serve to precisely align the image.

The image is applied in such a way as to optimize the degree of its distortion, in accordance with the quality of the marking, which, in turn, depends on the quality of the substrate, the accuracy of the marking machine and its settings. Below are ways to implement this process.

Recognition from the captured image of a random authentication pattern (RAP) is performed as follows. To accurately determine the position of the pattern in the captured image, known methods of image processing and recognition are used. The values of each RAP pixel are then measured (often on a scale of 256 gray levels). For convenience and uniformity of calculations, these values can be normalized, for example, in the range from -1 to +1. Then the values are grouped into subgroups, in this example, the size of 110 pixels.

Thus, for a subgroup of pixels that initially have a given value, we get 110 values. If the value of the source pixels (in the binary system) was "0", negative values should prevail (in the range from -1 to +1), while positive values should prevail if the value was "1". Therefore, for each of 110 subgroups, 110 pixels can be assigned a value of “1” or “0”.

For each of 12100 pixels we get the measured value in the image, possibly normalized, and the estimated initial value. Thus, the number of errors can be measured, for example, by calculating the number of pixels that matches their estimated value (that is, if the values are normalized from -1 to +1, a negative value corresponds to "0" and a positive value corresponds to "1" ) Thus, it is possible to measure the correlation coefficient, etc.

The metric found (“metric” means the error rate or similarity) is then compared with a threshold value to determine if the captured image matches the original or copy. To determine the specified threshold value can be used standard statistical methods.

It should be noted that to determine the indicator, the described procedure does not use data outside the image, with the exception of the composition of the subgroups. Thus, the calculation of the number of errors can be expressed as follows.

The number of errors is equal to the sum for all subgroups

(Sum (Sign (z _ij ) = f (z _i1 , ..., z _iM ))),

where z _ij is the value of the i-th pixel of the j-th subgroup containing M elements, possibly normalized, and

f is a function that estimates the pixel value for a subgroup, for example f (z _i1 , ..., z _iM ) = Sign (z _i1 + ... + z _iM ).

Several options are possible:

- to scramble the pixel values of the same subgroup, a cryptographic key is used so that not all of them have the same value. The scrambling function may be an exclusive OR function,

- the indicator calculation function can evaluate and determine the probability that the pixel value is “1” and “0”, respectively (for binary pixel values),

- this method can be applied to other types of digital authentication codes, if their structure is suitable for this, in particular to secure information matrices having the above advantages,

- this method can be extended to non-binary pixel values and / or

- the values of the pixels of the subgroup can be determined so that they transmit the message (without the need for its decoding when reading).

Reading a digital authentication code (DAC) requires an accurate determination of its position in the captured image so that the value of each of its constituent cells is restored with the greatest possible accuracy, taking into account distortions caused by printing and possibly capture. However, captured images often contain symbols that may interfere with the positioning step. Obviously, the smaller the area occupied by the protected information matrix, the higher the likelihood that other symbols or patterns will interfere at the stage of positioning. For example, the input data in the form of an A4 page, for example, a folder containing a protected information matrix, will contain a large number of other elements. However, even captured images of a relatively small size, for example 1.5 cm by 1.1 cm, may contain characters that can be confused with a secure information matrix, for example, a black square, DataMatrix, etc., as shown in FIG. 6.

Establishing the position of a protected information matrix can be difficult due to capture conditions (poor lighting, blurring of borders, etc.), as well as arbitrary image orientation in the range of 360 degrees.

Unlike other types of two-dimensional barcode characters, which change relatively little with different types of printing, the characteristics of digital authentication codes (DACs), such as textures, are extremely variable. Thus, well-known methods, for example described in US patent 6775409, are not applicable. In fact, this method of detecting codes is based on the directivity of the luminance gradient, however, in the case of a protected information matrix, the gradient does not have a specific direction.

Some methods for positioning digital authentication codes (DACs) take advantage of the fact that the DACs are square or rectangular in shape, which results in a noticeable contrast with continuous segments that can be detected and used using standard image processing methods. However, in some cases, these methods are unsuccessful and, on the other hand, it is required to be able to use DACs, which are optionally square or rectangle (or are not necessarily inscribed in them).

Generally speaking, the area on which the digital authentication code (DAC) is printed has an increased ink density. However, although the use of ink density measurements is useful, it cannot be the only criterion: in reality, Datamatrix or other barcodes, often adjacent to DACs, have an even higher ink density. Therefore, it is not enough to use only this criterion.

To determine the image areas belonging to CDP, in the patent EP 1801692 proposed the use of increased entropy. However, although not yet printed patterns for copy recognition do have increased entropy, the latter can change dramatically during printing, capture, or measurement. For example, a simple measurement of entropy, based on a histogram of the scatter of pixel values of each section, can sometimes lead to higher values in areas that are not very filled in content, which theoretically should have low entropy, for example, this may occur as a result of artifacts during compression in JPEG format, either due to the texture of the paper appearing in the captured image, or as a result of reflection effects from the substrate. Thus, it is clear that the criterion of entropy is also insufficient.

Generally speaking, methods for measuring or characterizing textures seem more appropriate in order to characterize at the same time the intensity or spatial relationships specific to DAC textures. For example, Haralick in Statistical and Structural Approaches to Texture (Haralick, Statistical and Structural Approaches to Texture), a link to which is included here as a reference, describes many ways to measure the characteristics of textures that can be combined so that they clearly described a large number of textures.

However, digital authentication codes (DACs) can have textures that vary greatly depending on the type of print or capture, and in most cases it is impossible or at least impractical to provide texture characteristics for the DAC placement method, mainly because they need to be adjusted to depending on the specific effects of the capture means in the measurement of texture.

Thus, it is obvious that in order to reliably determine the position of the DAC, a number of criteria must be integrated in a non-rigid way, in particular the following criteria are suitable:

- Digital Authentication Code (DAC) texture: DACs usually have a higher paint level and a stronger contrast with their surroundings. It should be noted that, by itself, this criterion may not be sufficiently characteristic: for example, there may not be a strong contrast for some DACs that are full of paint,

- DACs have a strong contrast around the edges: usually the DAC is surrounded by a silence zone that does not contain a label and which itself can be surrounded by a border to obtain maximum contrast (it should be noted that some DACs do not have a border or have only a partial border),

- DACs often have a special shape (square, rectangular, round or other), which can be used to determine the position and

- DACs in their internal structure often contain fixed groups of data, of course, provided that they own the cryptographic key or the keys used or used to generate them, usually serving or serving for accurate synchronization. If the indicated data groups are not detected, this means either that the position of the protected information matrix is not set correctly, or that the synchronization of the data groups is unknown.

These four criteria, which are the general characteristics of the DAC texture, the characteristics along the edges of the DAC, the general shape and internal structure, if combined appropriately, can provide with high reliability the determination of the position of the DAC in the so-called “adverse” conditions (in the presence of other two-dimensional codes, with poor capture quality, locally variable image characteristics, etc.)

In order to establish the position of the DAC, the present invention proposes the following method. Obviously, there are many possible variants of the method that do not go beyond its essence. A method that applies to square or rectangular DAC, but can also be extended to other forms, according to which:

- divide the image into sections of the same size so that the DAC area corresponds to a sufficient number of sections,

- measure the texture index for each plot. The indicator may be multidimensional and preferably contain an amount indicating the level of filling with paint, and an amount showing local dynamics,

- it is possible for each plot to calculate the total indicator of the texture, for example in the form of a weighted sum of each indicator measured for the plot,

- determine at least one detection threshold, depending on whether one or more indicators per section are stored in memory. Generally speaking, a value greater than a threshold value suggests that the corresponding region forms part of the DAC. For images representing illumination deviations, a variable threshold value may be used. When several indicators are stored in memory, it may be necessary that all indicators are larger than their respective threshold values for areas that are considered to form part of the DAC, or only one of the indicators is larger than its corresponding threshold value,

- Identify areas that are part of the DAC, called “positive areas” (and vice versa, “negative areas”). Received binary image. Alternatively, using cleaning using sequentially stretching and blurring, for example using the techniques described in Chapter 9 of Digital Image Processing using Matlab by Gonzales, Woods and Gadzales, Woods and Eddin Eddin),

- define continuous groups of positive zones larger than the minimum area. If a continuous group is not detected, they return back to the second step of the indicated algorithm and reduce the threshold value until at least one continuous group of the minimum size is detected. Alternatively, change the selection of the criterion of plots, if each plot has several texture indicators. Sites are determined by following the contour of the group that are located along the DAC boundary, characterized in that they have at least one adjacent negative zone,

- to detect a square, two pairs of points are formed, formed by the points farthest from each other. If two corresponding segments have the same length and if they form an angle of 90 °, then conclude that they form a square. Alternatively, apply the Hough transform,

- as an option, a restriction detection filter is used to the original image or its abridged version (examples of filters are shown in chapter 10 of this book) and

- determine the threshold value, then the position of the pixels having a response to the filter greater than the threshold value. These pixels indicate the boundaries of the object, in particular the boundaries of the plot of the protected information matrix. Make sure that the areas located at the four edges of the digital authentication code contain a minimum number of pixels, which indicates the boundaries of the object.

Regarding the stage of dividing the image into sections, it should be indicated that the size of the sections can have a significant effect on the result of determining the position. If the areas are too small, the measured values will be inaccurate and / or very noisy, which will make it difficult to detect areas belonging to the DAC. On the other hand, if they are too large, the determination of the position of the DAC will be inaccurate, and it will be difficult to determine that the shape of the proposed DAC corresponds to the desired shape (eg, square). In addition, the size of the areas must be selected in accordance with the DAC area in the captured image, which may be, but is not necessarily known. For some capture tools, the images will have a fixed size, for example the often encountered 640 × 480 pixel format. Therefore, theoretically, capture resolution will not be very variable. Some capture tools are capable of supporting more than one image format, such as 640x480 and 1280x1024. Therefore, the size of the plot must be selected in accordance with the resolution. For example, for a capture tool that produces images in 640 × 480 format with a capture resolution of 1200 dpi (dots per inch), the image can be divided into sections of 10 × 10 pixels, a total of 64 × 48 sections. If the same capture tool also supports 1280 × 1024 format, resulting in a double capture resolution of up to 2400 dpi, the plot size should also be doubled to 20 × 20 pixels (pixels at the edges that do not form a whole plot can be left on one side). For images obtained from the scanner, the resolution of which is not always known, it can be assumed that the capture resolution is 1200 dpi, or determine it based on metadata.

It should be noted that it is possible to use sections of one pixel in size, provided that the risks associated with the greatest noise are eliminated or controlled at subsequent stages.

Regarding the measurement of the texture index, as indicated above, the texture of the DAC can vary greatly, and there is no ideal measurement of the texture index. However, DACs are usually distinguished by a high level of ink filling and / or its strong changes. If the paint used is black or dark, and if the pixel values are in the range from 0 to 255, for the value of the i-th pixel of the plot, you can take y _i = 255-x _i . Therefore, the indicator of the fill level of the paint plot is equal to the average value of y _i . However, you can also accept the average, smallest value or percentile of the sample of values (position / value on the histogram corresponding to a given percentage of the sample). Such values from a sample of values may be more stable or more characteristic than just the average value.

As an indicator of changes, you can measure the gradient at each point and keep its absolute value.

As a combined indicator of texture, you can add in equal or unequal proportions the fill rate of the paint level and the rate of change. Since the indicated indicators are not reduced to a single scale, first the indicators of filling the ink level and changes for all image areas are calculated, normalized so that each indicator has the same highs / lows, and then summarized to obtain combined texture indicators.

Regarding the determination of the detection threshold, it should be noted that it is very unreliable. In fact, if the threshold is too high, many DAC sites are essentially not detected. On the other hand, too low a threshold leads to the erroneous detection of a significant number of sites that do not belong to the DAC.

On Fig shows the information matrix 665, captured at an angle of about 30 ° with a resolution of about 2000 dpi. On Fig shows the measurement result 670 of the combined indicator of the texture (106 × 85) made for the image shown in Fig. In Fig.16 shows the same image as in Fig.15, after comparing with the threshold value, the image is indicated by 680. Fig.17 shows the same image as in Fig.16, after applying at least one operation stretching and one blur; the image is indicated by 685. On Fig shows the contour 690 information matrix, determined by processing the image presented on Fig. In Fig.19 shows the angles 695 of the contour shown in Fig.18, determined by processing the image shown in Fig.18.

The determination of the threshold is complicated by the fact that the properties of the images change significantly. In addition, the texture properties of images may undergo local changes. For example, due to lighting conditions, the right side of the image may appear darker than its left side, and applying the same threshold value to two sides will lead to many detection errors.

The following algorithm gives certain stability to texture changes, according to which the images are divided into four sections and four detection thresholds are applied to them:

- determine the tenth and ninety percentile (or first and last decile) of the indicator for the entire image, for example 44 and 176;

- determine the first threshold as the average of the two specified thresholds: (176 + 44) / 2 = 110;

- the matrix of sections is divided into four sections of the same size, for example 32 × 24 with a section size of 64 × 48, and the tenth percentile for each of the four sections, for example, 42, 46, 43 and 57, is calculated.

The following describes a method for local segmentation, or "adaptive threshold determination". Some captured DACs have weak contrast at the borders or exhibit changes in illumination, which may be such that some parts of the DAC may be brighter than the background (which theoretically cannot be for a DAC printed in black on a white background). In this case, there simply does not exist a common threshold value that enables correct segmentation, or at least the specified value cannot be determined by standard methods.

To solve this problem, they resort to the following algorithm, which makes it possible to determine the areas representing a given uniformity of the indicator. For example, the site to be segmented is determined starting from the starting point (or the starting site), then all adjacent sections meeting the similarity criterion are successively selected. The specified starting point is often chosen so that it contains an extreme value, for example, the lowest figure for the image. For example, if the starting point contains the (minimum) indicator X, and if the similarity criterion is that all sections in the range from X to X + A, where A is a pre-calculated positive value, for example, correspond to measurements of the dynamics of the image, select the group adjacent cells that meet the specified criteria.

If this method does not satisfy, apply an alternative method, which consists in identifying areas in which there is no abrupt transition. In addition, the method consists in detecting a starting point with an index X, then an adjacent point Ra is selected if its index Y is less than X + B (B is also a predetermined value). Then, if the indicated adjacent point Ra is selected, the selection criterion for points adjacent to Ra is replaced by Y + B.

It should be noted that these algorithms can be applied to the image several times, for example, for each operation, taking different points as starting points. Thus, you can get several candidates for the plots, some of which may overlap.

Regarding the classification of plots according to the calculated threshold value, it should be pointed out that similar methods can be used to determine the general threshold value, for example, the iterative method described on pages 40-40 of the book Digital Image Processing Using Matlab, Gonzales, Woods and Eddin ("Digital Image Processing using Matlab" by Gonzales, Woods and Eddin).

Regarding the definition of areas belonging to the DAC, it should be noted that the bordering zones can be determined by the selection of areas for which at least one of the adjacent areas does not meet the criteria for the areas (texture index exceeds the detection threshold).

However, after determining at least one candidate for the site, it should also be determined whether these sites have the shape corresponding to the desired one. For example, a large number of DACs are square, but they can also be rectangular, round, etc. In this case, it is possible to determine the "signature" (characteristic feature) of the desired shape by determining the center of gravity of the original shape and then calculating the distance from the center of gravity to the edge of the shape farthest from it for each degree angle from 0 to 360 °. Thus, the signature corresponds to a curve representing the angle-normalized distance. The specified curve is constant for a circle, and for a square contains four identical extrema, etc.

For the candidacy for the site, a signature is also calculated. Then, the specified signature is compared with the original signature, for example, by measuring the maximum of the autocorrelation function (in order to take into account a possible rotation). In addition, it may be necessary to retake the original or computed signature.

If the calculated similarity value is greater than the predetermined threshold value, the portion is stored in memory, otherwise it is discarded. If you are searching for areas containing extreme values, such as a square, subsequently it may be useful to determine the angles of the square from the points associated with the specified extrema.

The following steps may be used:

- receive the original signature, and the presentation of data characterizing the candidacy for the site,

- calculate the signature of the candidate for the site,

- measure the maximum similarity between the signature of the candidate for the site and the original signature and

- save the candidacy for the site if the indicated similarity value is higher than the threshold value, and optionally determine the points corresponding to the extrema of the signature of the candidate.

The following is a description of the method for a secure information matrix that is not very sensitive to the level of ink filling. As indicated above, an excessive level of ink filling of a protected information matrix can significantly reduce its readability and even the ability to be distinguished from one of its copies. However, despite the fact that there are tools that can, as far as possible, control the level of ink filling during printing, their use can be difficult or even impossible. It is preferable to create a secure information matrix that is resistant to a wide range of ink fill levels.

It turns out that protected information matrices are usually more sensitive to an increased level of ink filling than to a lower one. In fact, if the fill level is low, black, or containing color, the cells are usually always printed, therefore, they have little effect on the reading of the protected information matrix. In contrast, as shown in FIG. 6, image 515, if the fill level of the ink is too high, the ink tends to overfill the substrate, and the white areas are “flooded” to some extent by the paint from the surrounding black areas. A similar effect can be observed when marking with contact printing, laser engraving, etc.

The asymmetry between the negative effects of excess paint and insufficient paint leads to the conclusion that protected information matrices containing a smaller proportion of marked pixels will be more resistant to changes in the level of ink filling. However, the values of the cells are usually equally probable, which is caused by encryption and scrambling algorithms that create the maximum entropy of the matrix. For binary matrices containing black or white cells, you can always reduce the number of black pixels forming black cells. For example, if the cell size is 4 × 4 pixels, you can choose to print a square portion of 3 × 3 pixels or 2 × 2 pixels. Consequently, the level of filling with paint is reduced in relation to 9/16 and 1/4, respectively (it should be noted that white cells are not affected). Other configurations are possible, such as those shown in FIG. 12. 12 shows:

- a secure information matrix 585 with a cell size of 4 × 4 pixels, and the printed area of each cell is 4 × 4 pixels, surrounding VCDP 575 and surrounded by microtext 580,

- a secure information matrix 600 with a cell size of 4 × 4 pixels, and the printed area of each cell is 3 × 3 pixels, surrounding VCDP 590 and surrounded by microtext 595,

- a secure information matrix 615 with a cell size of 3 × 3 pixels, and the printed area of each cell is 3 × 3 pixels, surrounding VCDP 605 surrounded by microtext 610,

- a secure information matrix 630 with a cell size of 3 × 3 pixels, and the printed area of each cell forms a cross of 5 pixels surrounding VCDP 620 and surrounded by microtext 625 and

- a secure information matrix 645 with a cell size of 3 × 3 pixels, and the printed area of each cell is 2 × 2 pixels, surrounding VCDP 635 and surrounded by microtext 640.

In addition, for example, you can print sections of 2 × 2 pixels or 1 × 1 pixels in cells, for example, 4 or 2 pixels in size. Obviously, asymmetric or mutable configurations are also possible in which variability can perform functions such as storing a message or reference for authentication purposes, as shown below in FIG. 11.

In the latter case, the added message may be error-protected and protected in a manner similar to the protection of other messages inserted into the protected information matrix. Only modulation will be different. Consider an example: a secure information matrix containing 10,000 cells containing a message will contain an average of 5,000 black cells. However, their exact number will differ for each message or for each encryption and scrambling key. Therefore, firstly, it is necessary to generate a protected information matrix similarly to the case with the whole cells filled in order to determine the exact number of available pixels (it should be recalled that this number directly affects the applied permutation operation). Thus, suppose that in this case, the protected information matrix contains 4980 black cells. If the cell size is 4 × 4 pixels, 4980 * 16 = 79680 pixels are available. If you want to insert a message with a length of 8 bytes, the length of which after converting to a convolutional code of the 2nd degree with memory 8 will be 176 bits, the message can be played 452 times (and partially 453 times). The reproduced message is scrambled, i.e. perform a permutation operation and apply the exclusive-OR function. Below is a way to minimize the cost of the permutation. The scrambled message is modulated in black cells of the secure information matrix.

Figure 11 below shows an example of the result 570 of such a modulation in comparison with the secure information matrix 565 shown above with "full" black cells.

It should be noted that due to the use of this method, statistically 50% of the pixels of black cells will be coated with ink, therefore, the level of filling with ink is reduced by half. You can easily change the indicated degree of decrease in the level of filling with paint, for example, by reserving a certain number of pixels of cells that receive a given value, black or white. With the minimum number of pixels having black color, it is possible to avoid the fact that there are no black pixels in the "black" cell randomly.

The specified second message level provides great benefits. Due to the higher resolution, the message contains more errors, which, however, is compensated by a higher redundancy, which is 8 times higher for 4x4 pixel cells. Such a message is much more difficult to copy, because it has a very high resolution, and even its very presence may not be detected. The contained message or the contained messages can be encrypted and scrambled using different keys, so that a greater number of security levels can be provided.

Many other options are possible, for example, dividing a 4 × 4 pixel cell into four sections of 2 × 2 pixels: the fill level will be statistically the same, and the resolution will be lower, and the message will contain fewer errors, but will also contain a lower level of redundancy .

In addition, the protected information matrix may contain several sections whose cell density is changed so that at least one of the cell density values corresponds to the level of ink filling used in printing. In this case, the reading can be performed using areas having the most suitable level of filling with paint.

The combined method for optimizing the cell size and density of the ink fill level for cells is described below: several pairs of size / ink fill levels are tested, for example, those that are in the error range of 19-27% are selected. If multiple pairs are selected, those with the highest resolution are selected.

Regarding the coefficient, or proportion, of errors, it should be indicated that its value can be defined as = (1-corr) / 2, where corr is the measurement corresponding to the correlation between the received message and the original message in the range from -1 to +1 (in practice, negative values are unlikely). Thus, for corr = 0.75, the error coefficient will be 0.125% or 12.5%. It should be noted that the term “correlation” means “having the same meaning”. It should also be noted that the term “error” used in the present description refers to printing errors, errors resulting from a distortion of the information matrix during the life of the document, to errors in reading the values of the matrix cells and to copy errors, respectively. In order to minimize the third component (read errors), it is preferable to perform several sequential read operations and store the operation with the lowest error rate in memory.

Otherwise, in order to measure the error rate, the image can be compared with the threshold value in an adaptive way, values greater than / less than the threshold value are defined as white / black. Adaptive determination of the threshold value allows you to protect more information, and the image obtained by adaptive comparison with the threshold value usually represents greater variability than the image obtained by general comparison with the threshold value. To apply the adaptive determination of a threshold value, for example, it is possible to calculate the average threshold value for an image and apply a local systematic error corresponding to the average illumination of a data block of 10 × 10 pixels. Alternatively, for an external effect, you can apply a high-pass filter to the image, then a global filter. To determine the error rate in the case of comparing the image with a threshold value, simply count the number of cells whose value after comparing with the threshold value does not correspond to the expected one.

In the case where generation, printing and / or reading are performed taking into account gray levels, each cell has an individual error rate, and the correlation uses the specified individual error rate of the cells.

It should be recalled that in order to maximize the likelihood of recognition, copies of the protected information matrix should be printed with a print resolution that is as close as possible to the optimum distortion. However, the indicated optimum varies depending on what restriction is used in the model, a fixed physical size or a fixed number of cells. In addition, for a given cell size, or resolution, cell density can have a strong effect on the amount of distortion. Thus, preference is given to a cell density that provides the lowest error rate for a given cell size, even if it approaches an optimum. In fact, with regard to ink density, printing conditions are preferred to ensure the best print quality, so that if counterfeiters used the same printing procedure, they would not be able to print copies of better quality than the originals.

In the following example, six protected information matrices are created with the same number of cells, therefore, with different physical sizes and with six sets of values, cell size / density. The matrices are offset printed with a print resolution of 2400 ppi, then read using a flatbed scanner with a resolution of 2400 dpi to obtain a good quality image in order to minimize reading errors caused by image capture. The following table summarizes the average error rates obtained for various parameters, the minimum error rate (MIN) obtained for each cell size, the corresponding DMIN density, and the DIFF difference between the indicated MIN value and the theoretical optimum error rate of 19% for a fixed number of cells. It should be noted that the blank columns of the table correspond to impossible combinations of parameters, namely, a density exceeding the cell size. In addition, it should be noted that the density is "1", i.e. the value at which only one pixel is printed in each cell was not tested, even though this can sometimes give good results.

The test results are shown in the following table, with the numbers in the rows and columns indicating the dimensions of the cell (columns) and square sections inside the cells filled with ink (rows), so the intersection of row 3 with column 4 corresponds to the case in which it is printed exclusively a 3 × 3 pixel square in a 4 × 4 cell designed to fill with paint.

Density / Cell Size 2 (1200 ppi) 3 (800 ppi) 4 (600 ppi) 2 34% 22% 12% 3 - 26% eleven% four - - 22% MIN 34% 22% eleven% DMIN 2 2 3 Diff fifteen% 3% 8%

As can be seen from the table, the size of cell 3 (column 3) with a density of 2 (row 2) gives the value of the error coefficient, which is closest to the optimum of 19%. As indicated, the error rate for density 4 and mesh size 4 is 3% of the optimum. However, since at densities 2 and 3, significantly lower error rates are obtained, as shown in the table at the intersection of rows 2, 3 and column 4, it is not recommended to select these print options.

The following steps may be used:

- for each pair, the cell size / density of the candidacy creates one protected information matrix,

- print each created secure information matrix at least once under printing conditions, which will subsequently be used to print the document, for example three times,

- perform at least one capture of at least one print of each created secure information matrix, for example three capture,

- calculate the average error rate obtained for each captured secure information matrix,

- determine the minimum average error rate obtained by MIN for various protected information matrices created in accordance with the size of the cells, and select the appropriate density DMIN,

- for each MIN value, the DIFF difference between the absolute value and the optimum is calculated, and

- choose the size of T cells, providing the lowest value of DIFF, and the corresponding density of DMIN.

According to embodiments, if the density can change with a fixed cell size or the size can change with a fixed cell density, the same algorithm can be used, which makes it easier.

Preferably, print parameters, such as print media, the substrate used, and other print parameters, such as the size of the raster in offset printing, (if known) are included in the message that carries the secure information matrix. The specified information can be used for automatic interpretation or interpretation by a person.

For example, to indicate whether the substrate is paper, cardboard, aluminum, PVC, glass, etc., a few bits are usually sufficient. Likewise, a few bits are usually sufficient to indicate whether the printing media is offset, letterpress, screen, gravure, etc. Thus, if the printing means are intaglio printing on aluminum, this information is stored in a secure information matrix. If a high-quality copy is printed on good paper by means of offset printing, so that due to the high print quality, the copy can be mistaken for the original, the operator informed of the expected substrate, when reading the protected information matrix, can determine that the substrate does not match the expected.

There are ways to automatically determine the type of print, for example, offset or laser printing leave specific traces that can facilitate the automatic determination of the type of print based on the capture and processing of images or images. The result of applying this method can be automatically compared with the print settings stored in the protected information matrix, and the result can be integrated into the decision process on the authenticity of the document.

The following is a description of the steps for generating and reading or using the information in question, and “print options” may include measuring the ink fill level or cell density of the protected information matrix (these steps apply to all types of DAC):

- automatically print parameters for the DAC or indicator section, as shown in Fig.9 and 10, by processing the image or, for example, using the output signal using a densimeter, or, alternatively, by entering parameters by the operator,

- get the DAC print settings,

- encode print parameters, for example, in binary or alphanumeric format,

- insert the encoded parameters into the DAC message and / or microtext and

- generate DAC in accordance with a known algorithm.

To use print options:

- automatically print parameters for the DAC or indicator section, as shown in Fig.9 and 10, by processing the image or, for example, using the output signal using a densimeter, or, alternatively, by entering parameters by the operator,

- get the DAC print settings,

- read the DAC,

- retrieving print parameters from the message contained in the read DAC and

- compare the extracted and obtained parameters, and on the basis of this comparison decide on the origin of the document.

According to an embodiment, the above algorithm is applied only if the DAC is already defined as the original, either automatically or manually.

Regarding the measurement of print parameters other than the ink fill level, they are usually not changed by the print media. Therefore, the measurement can be performed using indicators that are not attached to the document, but used at the stage of testing and verification of print media.

On figa shows a secure information matrix 650 containing 21 × 21 cells representing the message. 13b shows a secure information matrix 655 containing 21 × 21 cells representing the same message. On figs shows a protected information matrix, covered four times (or tiled) matrix from fig.13b, to form a secure information matrix 660.

The following is a description of the preferred option for creating information matrices, which include an indication of the density of filling with paint. Printers typically use a densitometer to measure density or ink fill levels. A densitometer is usually applied to base rectangles containing the maximum amount of ink located on the edges of printed sheets and discarded when cutting sheets into documents. Often for a document (or product, packaging, etc.) intended for printing, the printer receives the limit values for ink density: prints for which the ink density value is outside the tolerance are not valid, and the printer should theoretically print them again. If this does not happen, i.e. if the printer printed the documents without observing the tolerance limit for the fill density for all configurations, it is highly desirable to be able to detect this on documents in circulation: in fact, the reading may be distorted, for example, the original may be recognized as a copy if the fill density is too high or too low, and in this case it is necessary to notify the copyright holder that there are problems with the density of the paint, which may be the cause indicated of erroneous reading. Therefore, this will avoid the harmful effects of false recognition and may make it possible to hold a printer who does not comply with printing parameters responsible. However, as indicated above, the base rectangles are mostly deleted during cutting.

A suitable area of approximately 4 mm ^{2 is} usually required for an appropriate measurement of the ink filling density, since the diameter of the densitometer is approximately 1.5 mm ² . It is useful to attach a portion of the indicated area inside or next to the protected information matrix printed with the color used for it, so that you can check if the ink density is sufficient when reading the protected information matrix may not give the expected result (for example, a copy ) FIG. 9 shows a secure information matrix 550 combined with a fully ink-filled portion 545 located therein. Figure 10 shows a secure information matrix 555, combined with a paint-filled area 560 located next to it.

For reading, you can use the following steps in which:

- get the lower and upper limits of the density of the paint,

- if necessary, the specified boundaries are converted into appropriate levels on a gray scale for the given capture conditions,

- enter the density of the paint image of the base area,

- the image determines the level value of the specified area on a gray scale and

- check whether the specified value is within the specified limits: if so, return a positive message, otherwise a negative message.

The following is a description of a method for generating informational matrices containing geometric patterns, in this case a circle.

An image is generated containing various geometric patterns, preferably using a key and, possibly, a message. Geometric patterns and their parameters are determined using the key.

The following steps can be used to create information matrices with geometric patterns:

- using a key generate a sequence of pseudo-random numbers,

- generate a blank image,

- depending on the generated numbers determine a lot of geometric shapes and related parameters,

- For each of the defined geometric shapes, geometric shapes are inserted into the blank image.

The following steps can be used to detect geometric parameters:

- using a key generate a sequence of pseudo-random numbers,

- depending on the generated numbers, a lot of geometric shapes and related parameters are called, called “initial parameters”,

- for each of the defined geometric shapes, the shape parameters in the image are evaluated and

- in a given metric, measure the distance between the expected parameters and the original parameters of the figure.

The following is a description of how to attach a pattern of dots with variable characteristics.

As indicated above, patterns of variable characteristic points (VCDPs) can be used to recognize copies, store information, as well as to uniquely identify the source image. In particular, they offer advantageous and additional means of protecting documents. 7 shows a secure information matrix 520 containing a central portion in which a VCDP 535 containing geometric shapes, in this case circles and microtexts 530 is placed. FIG. 8 shows a secure information matrix 535 surrounded by VCDP 540. It should be noted that in this case, elements for locating the DAC, such as its angles, can be used to determine the location and approximate positions of the VCDPs. 12 shows connected VCDPs and a secure information matrix.

In addition, by integrating VCDP into a secure information matrix, the degree of protection can be improved since the fake manufacturer must overcome the copy protection barriers of both VCDP and the secure information matrix. VCDP and a secure information matrix can be created using various cryptographic keys, therefore, declassifying one of the keys is insufficient to declassify the entire image. On the other hand, the information contained can be correlated so that the VCDP and the secure information matrix are essentially interconnected. The following is a possible algorithm:

- receive a message, a cryptographic key A for VCDP and a key B for a secure information matrix,

- from the message and key A create a secure information matrix, reserving the specified space for VCDP,

- allocate the second message from the received message, for example from its part,

- from the second message and key B create VCDP and

- insert the created VCDP into the secure information matrix.

According to an embodiment of the present invention, in which not the entire surface of the cell portion is printed, for example, based on the ink filling density, as described elsewhere, the position of the ink-filled portion in the cell is modulated according to the message, possibly randomly, as well as with VCDP. For example, a paint-filled area represented by a 3 × 3 pixel square in a 4 × 4 pixel cell can take four different positions. Thus, it is possible to increase the ability to recognize copies and / or embed additional information in the matrix.

The following is a description of the use of the information matrix for unambiguous identification using material analysis. Methods of identification and authentication of documents based on obtaining the characteristics of the material offer a high degree of protection. However, these methods can be difficult to use, since without marks indicating the portion of the document used to create the print, it may be difficult to correctly position the reader so that the corresponding part of the document is captured. However, protected information matrices create an easily identifiable landmark for setting the position of the reader. Thus, to create a print of the material, a section located in the center of the protected information matrix can be used, the position of which can be known with high accuracy due to the reference configurations of the protected information matrix. This can be done in spite of the reservation of the indicated area for VCDP insertion.

The following is a description of the integration of microtext or text into the information matrix. Microtext is usually presented in vector form. However, protected information matrices are bitmap images. As a result, for insertion into a protected information matrix, microtext must be raster. To preserve the accuracy of the text, it is preferable to provide a protected information matrix with the highest possible resolution as far as possible. For example, a secure 110 × 110 pixel information matrix for printing at 600 dpi (600 dpi) must be quadrupled (440 × 440 pixels) so that it can be printed at 2,400 dpi inch (2400 ppi), if print media permit this.

The protected information matrix often contains a frame that is black or creates a contrast with the immediate environment of the matrix, which facilitates its detection in the captured image. However, it turns out that although the corners of the frame are very useful in practice (determining the positions of each of the angles provides the ability to accurately set the position of the protected information matrix), the central parts of the frame are not very useful. Mostly replace them with microtext. For example, if the border is 3 pixels for printing with a resolution of 600 dpi (600 ppi) and, therefore, 12 pixels for printing with a resolution of 2400 dpi (2400 ppi), microtext can be up to 11 pixels high (preferably leaving one pixel for the inner edge of the matrix).

In the case of a square or rectangular protected information matrix and if the identical microtext is inscribed on four sides, for example, the name of the copyright holder, the name of the product, etc., it may be advantageous to orient the text in such a way that for any orientation in which the protected information matrix is observed or captured Reading the text was normal. A similar matrix is shown in FIGS. 7 and 12.

In addition, sections within the matrix can also be reserved for microtext insertion. In this case, in order to configure the modulation and demodulation of the message or messages accordingly, the device for creating and reading should be informed about areas containing microtext.

If the print reproduces the image and, therefore, the printed secure information matrix must change on each print, which is especially likely for digital printing, microtext can be modified on each print. In this case, the microtext can, for example, contain an identifier, a serial number, a unique number or any other text, in particular text that provides a link between the protected information matrix and the rest of the document. If the document is an identification card, microtext may contain, for example, the name of the holder. If the document is a package, microtext may contain an expiration date, a batch number, a trademark, a product name, etc.

The following are the steps for integrating a variable microtext into a secure information matrix:

- receive a message, a cryptographic key, possibly a font, sections reserved for microtext with the corresponding orientation of the text,

- create an image of a secure information matrix in accordance with the received message and key, reserved areas,

- in accordance with the received message generate a microtext image and

- insert the image containing microtext in each reserved area, possibly with rotation by an angle multiple of 90 °, in accordance with the orientation of the text.

According to an embodiment, the message used for microtext is part of the received message. According to another embodiment, the message is encrypted using a key obtained before generating microtext.

It should be noted, as an option, that the contents of the microtext during printing is a function of the content of the information matrix, or vice versa, the content of the information matrix can be a function of the contents of the microtext. The functions in question may be, for example, cryptographic functions. For example, the contents of microtext can serve as a cryptographic key when reading to determine the content of the information matrix.

Theoretically, microtext is intended to be read and interpreted by humans, but microtext can also be read automatically using image capturing tools and optical reader software. In this case, the specified software can provide the result in the form of text, which can be automatically compared with other types of information provided: data extracted from the protected information matrix, or with other symbols printed on the document, etc.

The following is a description of inserting information matrices into barcodes. Like inserting a message distributed across all cells of a secure information matrix, it itself can be inserted into cells of a two-dimensional barcode, for example, Datamatrix (registered trademark). Since protected information matrices have a high level of ink filling, theoretically they will not interfere when reading a two-dimensional barcode.

According to a preferred embodiment, each black cell of the Datamatrix barcode contains a secure information matrix. If the restrictive conditions of the application allow, each protected information matrix contains its own message, in which, for example, one part is fixed and the other part contains a pointer that can be associated with the cell position in the Datamatrix code.

Claims (31)

1. A method of protecting a document, characterized in that it contains:
- the stage of determining the printing conditions of the specified document,
- the step of determining the physical characteristics of the cells of at least one configuration in accordance with these printing conditions, for example, the percentage of cells printed with a printing error that occurs solely due to unforeseen variables is greater than the predetermined first value and less than the predetermined second value, the first the set value is at least 5%, and the second set value is not more than 35%,
- a step of presenting a unit of information by changing the appearance of cells representing the indicated physical characteristics, and
- the stage of printing the specified configuration using the specified printing conditions, and the specified configuration is made with the possibility of recognizing a copy that modifies the appearance of many of these cells. 2. The method according to claim 1, whereby the step of determining the physical characteristics of the cells includes
- the stage of determining the target error rate and
- the stage of choosing among the possible physical characteristics of the cells of at least one form of such physical characteristics of the cells of at least one form such that the proportion of cells printed in accordance with the printing conditions, with a printing error that occurs solely due to unforeseen variables, is closer to the target error rate . 3. The method according to claim 1, according to which at the stage of determining the physical characteristics of the cells determine the size of the cells intended for printing. 4. The method according to claim 1 or 2, according to which, at the stage of determining the physical characteristics of the cells, a part of the cells having a constant shape and a changing color is determined to represent different values of a unit of information, said part being certainly smaller than the specified cell. 5. The method according to claim 4, whereby the first predetermined value is more than 10%. 6. The method according to claim 4, whereby the first predetermined value is more than 15%. 7. The method according to claim 4, according to which the first set value is more than 20%. 8. The method according to claim 4, according to which the second setpoint is less than 25%. 9. The method according to claim 4, whereby the second predetermined value is less than 30%. 10. The method according to claim 1, according to which at the printing stage use their own resolution of the printing media that perform the specified printing. 11. The method according to claim 1, characterized in that it further comprises the step of creating a configuration in a digital information matrix representing a message containing redundancy. 12. The method according to claim 10, according to which, at the stage of creating the configuration, there is a proportion of redundancy sufficient to adjust the proportion of errors greater than the specified first set value. 13. The method according to claim 10 or 12, according to which at the stage of creating the configuration, the specified redundancy contains error correction codes. 14. The method according to item 13, according to which at the stage of creating the configuration of the specified redundancy contains error detection codes. 15. The method according to 14, according to which at the stage of generating the information matrix, the specified information matrix at the level of each unit cell and independently of neighboring unit cells represents a message containing redundancy. 16. The method according to clause 15, according to which at the stage of generating the information matrix, the redundancy serves to enable recognition of unrelated errors made at the stage of labeling in the label. 17. The method according to claim 1, according to which at the stage of applying labels to the label of the information matrix add an additional reliable label containing the message. 18. The method according to claim 1, whereby at the stage of generating the information matrix, the presentation of the specified message is encrypted using an encryption key, the information matrix being the specified encrypted message. 19. The method according to claim 1, whereby at the stage of generating the information matrix, the presentation of the specified message is encoded to create the specified redundancy. 20. The method according to claim 1, according to which at the stage of generating the information matrix, the representation of the specified message is reproduced in order to create several identical copies. 21. The method according to claim 1, according to which, at the stage of generating the information matrix, the positions of the presentation elements of the specified message are interchanged in accordance with the secret key. 22. The method according to claim 1, according to which, at the stage of generating the information matrix, the positions of the elements of the presentation of the specified message are partially interchanged in accordance with a secret key that differs from the key of the first permutation. 23. The method according to claim 1, according to which, at the stage of generating the information matrix, a scrambled function depending, on the one hand, on the value of the element, and on the other hand, on the value of the secret key element, is applied to at least one part of the presentation elements of the specified message . 24. The method according to claim 1, according to which at the stage of generating the information matrix, the function is “exclusive OR (XOR)”, which depends, on the one hand, on the value of the element, and on the other hand, on the value of the element of the secret key, which differs from the secret key the first replacement function is applied to at least one part of the presentation elements of the specified message. 25. The method according to item 23 or 24, according to which the specified substitution function pairwise replaces the values associated with neighboring cells in the specified configuration. 26. The method according to claim 1, according to which at the stage of generating the information matrix, at least one key is used, and this is not the associated key that is necessary for reproducing the information, but another key. 27. The method according to claim 1, according to which at the stage of generating the information matrix, a digital information matrix is generated representing at least two messages equipped with various security features. 28. The method according to item 27, according to which one of these messages is the information required when reading the information matrix to determine another message and / or detect errors in it. 29. The method according to item 27 or 28, according to which one of these messages is at least one key required to read another message. 30. The method according to claim 1, according to which at the stage of generating the information matrix, its hash value is added to the message representation. 31. A device for protecting a document, characterized in that it contains:
- means for determining the printing conditions of the specified document,
- means for determining the physical characteristics of the cells of at least one configuration in accordance with the specified printing conditions, for example, the percentage of cells printed with a printing error that occurs solely due to unforeseen variables is greater than a predetermined first value and less than a predetermined second value,
moreover, the first preset value is at least 5%, and the second preset value is not more than 35%,
- means for representing a unit of information by changing the appearance of cells representing the indicated physical characteristics, and
- means of printing the specified configuration using the specified printing conditions, and the specified configuration is made with the possibility of recognizing a copy that modifies the appearance of many of these cells.

Images