Computer implementation method and system for marking digital watermarks on data

Abstract

The present invention is a computer-implemented method for marking a digital watermark (1) on data, comprising the following method steps performed by a computing unit (2): a) Step of providing an initial data set (3); b) A step of reading at least one operation system from a watermark database (5) that is connected to a computing unit (2); c) A step of generating a marked data set (7) by applying an operation (6) according to the operation system to each item (4) of the initial data set (3); d) Step of providing a marked data set (7) for use; e) Step of receiving the data set (8) to be examined; f) A step of determining the probability of agreement (9) for the data set (8) to be examined; g) If the match probability (9) is greater than a specified threshold: a step of comparing an item (4) of the data set (8) to be examined with a corresponding marked data set (7) determined for the data set (8); and h) The present invention relates to a computer implementation method characterized by a step of repeating steps a) through g), wherein for items (4) having the same data type as the item (4) for which the operation (6) was confirmed to have been removed in step g), the computing unit (2) uses a different operation system in the watermark database (5).

Description

Technology Field

[0001] The present invention relates to a computer-implemented method for marking a digital watermark on data and a system suitable therefor according to the type defined in detail in the preamble of claim 9. Background Technology

[0002] It is known to uniquely mark data, such as computer-readable files, by providing digital watermarks. This can be used, for example, to verify the authenticity of each file. Thus, it is possible to track whether each file was created by an authorized source or modified by an unauthorized third party. Digital watermarks are typically embedded directly into the content to be marked using steganographic methods. However, the provision of digital watermarks must not cause the content of each file to be altered so severely that the file can no longer be used for its original intended purpose. Furthermore, ideally, digital watermarks should be implemented so that their presence cannot be easily detected on each file. This prevents a third party from removing the digital watermark from the file without authorization.

[0003] As digitalization increases, the importance of digital data is also growing. For example, vehicle manufacturers can aggregate data describing vehicle usage behavior. This data is recorded while the manufacturer's vehicles are in operation and can then be processed by a central computing device, such as the cloud. The data generated in this way can be sold, for instance, to a third party, enabling them to carry out vehicle-related technology development. For instance, vehicle components and / or vehicle-related services may be developed or optimized.

[0004] There is a risk that a vehicle manufacturer may lose revenue if its customers illegally resell data obtained from the manufacturer. Therefore, to prevent this in the future, it would be desirable to be able to verify at what point the data was transmitted. Furthermore, data generated by the vehicle manufacturer can be falsified or altered, which can cause serious harm. For example, erroneous data may lead to technical defects or malfunctions, resulting in damage to machinery or injury to personnel. Legal action against the vehicle manufacturer may also be taken as a result of such incidents. Therefore, it is desirable to provide methods and means that enable the verification of such data manipulation.

[0005] A common method of marking data with a digital watermark involves, for example, adding artificially generated content, particularly content of a specific format, to the corresponding computer-readable file. This addition then becomes the digital watermark. If such addition and the consequently digital watermark are detected, they can be simply removed from the data set. Therefore, digital watermarks should be as difficult to detect as possible without knowing the method of implementation.

[0006] Furthermore, in the field of software development, it is known that irrelevant code components are included in the program code sections related to actual functions to make them ambiguous. Code components may also be added to alter the program flow, causing unrelated task steps to be performed. This is also known as obfuscation. Even if an unauthorized third party reads the source text of the program, they cannot identify the underlying program code, or it becomes difficult for humans to read or understand. This makes "reverse engineering" of the program difficult.

[0007] Methods for performing obfuscation and methods for evaluating the correspondingly performed obfuscation are disclosed, for example, in International Publication WO 2007 / 049817 A1. Using the methods disclosed in the said document, it is possible to evaluate whether the program code is sufficiently obfuscated. If it is not sufficiently obfuscated, additional iterative obfuscation processing may be performed to improve the obfuscation of the program code. To this end, different parts of the program code may undergo obfuscation or different obfuscation methods may be used. Additionally, Chinese Patents CN 113 987 550 A and CN 110 414 194 A disclose algorithms for inserting watermarks into documents. The problem to be solved

[0008] The present invention aims to provide an improved computer implementation method for marking a digital watermark on data, which protects computer-readable data more reliably when used.

[0009] According to the present invention, this objective is achieved by a computer-implemented method for marking a digital watermark on data having the features of claim 1. Preferred configurations, improved embodiments, and a system for carrying out the method of the present invention are shown in the dependent claims of this independent claim. means of solving the problem

[0010] A general computer implementation method for marking a digital watermark on data is improved by the following method steps performed by a computing unit according to the present invention:

[0011] a) a step of providing an initial data set, wherein the initial data set is in a structured format such that the items of the initial data set have an array system defined in the initial data set;

[0012] b) a step of reading at least one operation system from a watermark database connected to a computing unit via communication;

[0013] c) a step of generating a marked data set by applying operations according to the operation system to each item of the initial data set;

[0014] d) Step of providing the marked data set for use;

[0015] e) Step of receiving the data set to be examined;

[0016] f) a step of determining the probability of agreement of the data set under examination based on an operation according to at least one operation scheme found in the data set under examination;

[0017] g) If the match probability is greater than a specified threshold: a step of comparing an item in the data set under inspection with a corresponding marked data set determined for the data set under inspection, and determining an item in the data set under inspection from which the operation applied in step c) has been removed; and

[0018] h) a step of repeating steps a) through g), wherein for items having the same data type as the item confirmed to have had the operation removed in step g), the computing unit uses a different operation scheme in the watermark database.

[0019] By using the computer-implemented method for marking data according to the present invention, a digital watermark can be reliably provided to a computer-readable file. By performing method steps a) through g) in an iterative manner, the computing unit provides corresponding markings using different operation schemes sequentially, making it more difficult to identify the digital watermark. Consequently, it becomes more difficult for an entity unaware of the underlying operation scheme to remove the corresponding operation from the marked data set. Thus, the initial data set or the correspondingly marked data set can be increasingly reliably protected from tampering by unauthorized third parties. Accordingly, an entity performing the method according to the present invention can reliably determine whether data obtained from an external source was generated within the company, that is, whether it matches the initial data set or is based on or originates from such a data set.

[0020] Since the corresponding operation is applied to each item in the initial data set, it is reliably possible to prove the operation in the data set under inspection even if the marked data set is altered by an unauthorized third party. For example, if some items are changed or removed, there are still sufficient additional items in the data set under inspection to include the operation and the digital watermarks applied accordingly. Each data set includes a header (" Header It may have additional content such as ""). This additional content may not be manipulated, or it may be appropriately manipulated to mark a digital watermark. Individual items may be manipulated multiple times to have multiple digital watermarks. The pattern used by the computing unit when applying manipulation to each item directly constitutes the digital watermark.

[0021] In this case, the computing unit may generate the initial dataset directly or receive it from a trusted source, such as a server where values ​​corresponding to the items are aggregated. The initial dataset can have or take the form of a wide variety of file formats. The only important thing is that the items in the initial dataset have a specified arrangement scheme. This allows the items of each dataset to be distinguished from one another. For example, in the case of a text file, a new line of the text file corresponds to a new item. The first line of the text file may be additional content. Items may also be separated from one another, for example, by spaces.

[0022] The operation system used is designed specifically by data type. All common data types, particularly those automatically detected according to general technical standards—such as strings, booleans, numbers, and integers—can be considered. For example, mathematical operations can be applied to numeric values ​​but not to alphabetical columns.

[0023] When a marked data set is generated, it is distributed for use. This means that a third party can process and appropriately manipulate the marked data set. If such manipulation is suspected, the data set is provided to the computing unit as a data set to be inspected. The computing unit then analyzes whether at least one manipulation consistent with the manipulation scheme stored in the watermark database is found in the data set to be inspected. Such manipulation may be coincidentally included in the data set to be inspected. To clearly distinguish between coincidental matching and actual item manipulation, the computing unit determines the probability of a match in step g) and compares it to a defined threshold for each. In the simplest case, it may be sufficient if at least one manipulation is found in the data set to be inspected. However, the probability of a match may also be determined, for example, based on the ratio of the number of actual manipulations found in the data set to be inspected to the maximum number of manipulations provided to each data set according to the computing unit's procedure.

[0024] If the probability of a match is greater than a defined threshold, it means that the dataset under examination is an initial dataset modified by an unauthorized third party, or that the dataset under examination includes items taken from a marked dataset. Thus, misuse has occurred. The threshold may be at an appropriate level, e.g., 5%, 10%, 20%, 50%, or a portion or multiple thereof. The threshold may be adjusted depending on the situation.

[0025] The objective of the present invention is to increase the reliability that the corresponding digital watermark or manipulation is not removed from a marked data set. To this end, a computing unit determines a marked data set corresponding to the data set under inspection and identifies items from which the corresponding manipulation has been removed. Thus, the manipulation scheme used for these items or these data types is known to a relevant third party, and the relevant third party can remove the corresponding manipulation from the marked data set. Accordingly, the computing unit uses a different manipulation scheme for the corresponding item or data type in step h), thereby increasing the probability that such manipulation scheme will not be known to an unauthorized third party.

[0026] In step f), the computing unit already determines which items are free from the corresponding manipulation and, consequently, the digital watermark. An item in the data set under inspection may be free from manipulation because the manipulation was intentionally removed, or because an unauthorized third party, unaware of the manipulation scheme under which the digital watermark marking operation should be provided, directly added the item. Therefore, to definitively prove that the manipulation has been removed, the corresponding item in the data set under inspection must be compared with the corresponding item in the corresponding marked data set. To determine the corresponding marked data set, the computing unit compares the data set under inspection with the marked data set stored in the database in terms of content, and may select the marked data set with the highest content match with the data set under inspection as the "correspondence." In this case as well, a threshold may be defined, and if this threshold is exceeded, the correspondence is not confirmed, even if there is a certain overlap, for example, due to one or more items matching by chance. The marked data set may be accepted into the database when generated by the computing unit.

[0027] In this regard, it should be noted that the absence of manipulation in the data set under inspection may indicate the strength of the manipulation system, while the presence of manipulation may indicate the vulnerability of the manipulation system. Therefore, it is important to distinguish between items already existing in the marked data set and items modified by unauthorized third parties. Manipulation involving items directly added by a third party implies that the third party has grasped the manipulation system. According to the present invention, the computing unit may also switch such a manipulation system. However, the absence of manipulation in items added by a third party that is present in the original items indicates that the manipulation system is robust. According to the present invention, the computing unit may not switch such a manipulation system for the future generation of the marked data set.

[0028] In this case, an advantageous improved embodiment of the method according to the present invention provides that a computing unit adds each item according to an operation pattern to manipulate items in an initial data set, particularly adding / adding numbers obtained by applying mathematical operations to items composed of numeric values, or changing at least one format of each item according to the operation pattern. Thus, there are various methods for manipulating each item. If an item contains a numeric value, it may be added as a number at any location. This number is generated specifically according to the unmanipulated numeric value. To this end, various mathematical operations may be applied to the item. For example, the number may be generated by generating a digit sum. Thus, this process is the process of generating a so-called checksum. All general methods for generating checksums may be applied.

[0029] If the item is a string, for example, alphabets or special characters may be added at any position. For example, an underscore may be added at the fourth position of the string. If the string does not contain numbers, for example, the first three letters may be capitalized.

[0030] Additionally or alternatively, manipulation patterns may provide for changing item formatting. For example, the letter size of specific alphabets in an item, e.g., the font size of the first alphabet, may be reduced or increased by 1. For example, the font color of a character may be changed from black to very dark gray at a specific location. Multiple formatting changes may also be applied in combination. For example, a white underline may be added at an arbitrary location in an item implemented in black. This makes it more difficult for an unauthorized third party to detect such manipulation. Particularly preferably, each manipulation scheme, according to a pattern in which such addition or formatting change is defined, for example, for each item or n In the nth item ( n It provides that even numbers are applied. For example, a white underline may be added every fifth item following the first four alphabets in a string, and a space may be inserted at an appropriate position every sixth item. To determine if such operations exist in the data set under inspection, the computing unit checks whether the corresponding pattern providing addition or format change is found.

[0031] According to an additionally advantageous configuration of the method according to the present invention, a computing unit adds at least one artificially generated item to an initial data set for additional marking on the initial data set, and in particular, stores the artificially generated item in an artificial item database, the artificially generated item including array information describing the array of artificially generated items in the initial data set.

[0032] In step f), the computing unit determines whether at least one artificial item stored in the artificial item database can be found in the data set under inspection in order to determine the probability of a match. Thus, in each initial data set, not only are existing items added or their formats changed, but entirely new items may also be artificially generated and inserted into the initial data set. These artificially generated items are particularly conspicuous "markers" that serve as digital watermarks. Preferably, these artificially generated items are created only when existing items in the initial data set have already been manipulated.

[0033] Artificially generated items can be created according to defined rules. For example, the creation of artificially generated items can be designed based on items already present in the initial dataset. For instance, some or all average values ​​included in the dataset can be generated and then added as artificially generated items. The creation of items artificially may also follow specific patterns. For instance, an artificially generated item may be provided to the initial dataset every 1,000th item. By adding artificially generated items, the initial dataset is also transformed into a marked dataset. In a broader sense, the artificially generated item may be added to an already marked dataset. These artificially generated items serve as a particularly strong indicator that the dataset under examination is the initial dataset or that the dataset under examination is based on it. Therefore, it can be determined with particular reliability whether the dataset under examination originated internally within the company.

[0034] Additionally, a further advantageous configuration of the method according to the present invention is that the structured format is in a tabular format. That is, the items of the initial data set or the marked data set are arranged based on rows and columns. Since this is a particularly frequently used format, it is described that digital workmarks are provided for information in this tabular format accordingly.

[0035] In addition, according to an additionally advantageous configuration of the method according to the present invention, the computing unit increases the matching probability in step f), wherein:

[0036] - Excluding checksum-type manipulations, if the ratio of the number of manipulations still existing in the data set under inspection to the total number of manipulations applicable to the item according to the manipulation scheme used is greater than 5%;

[0037] - In the case of checksum-type operations, if the ratio of the number of operations still existing in the data set under inspection to the total number of operations applicable to the item according to the operation scheme used is greater than 10%;

[0038] - It provides an increase in the match probability when the data set under examination includes at least one artificially generated item.

[0039] The data set may be modified by an unauthorized third party, causing items to be altered so that the operation applied by the computing unit in step c) is removed. An entirely new item may also be added to the marked data set by an unauthorized third party. In step f), the computing unit determines whether it can find an operation that matches the respective operation scheme for each item in the data set under inspection. The computing unit stores each initial data set and / or the marked data set in a data set database and can compare them with the content of the data set under inspection in terms of content. Thus, not only is the presence or absence of an operation in the data set under inspection determined, but also the match or difference in content can be determined. The more operations found in the data set under inspection, the higher the probability of a match. A threshold of 5% or 10% has been proven suitable for distinguishing between actual operations applied by the computing unit and accidental matches. Generally, other percentage values ​​are also possible.

[0040] Since it is nearly impossible for an unauthorized third party to directly and accidentally generate an item artificially created by the computing unit, detecting an item artificially created by the computing unit in the data set under inspection is sufficient to identify that the item was generated within the company itself. The computing unit may determine a match probability of 100%, in particular, if at least one item artificially created by the computing unit is found in the data set under inspection. However, instead, the match probability may increase to, for example, 1%, 5%, 10%, or a portion or multiple thereof.

[0041] In addition, an additionally advantageous configuration of the method according to the present invention provides that a new operation system is added to the watermark database, and the new operation system is:

[0042] - Manually defined by the developer;

[0043] - Automatically defined by artificial intelligence, particularly considering the frequency of removal of manipulations in the data set under inspection; or

[0044] - Predefined manipulation rules for each data type are automatically defined through a probabilistic combination.

[0045] This prevents the depletion of manipulation systems stored in the watermark database over time. New manipulation systems may be defined manually by the developer or automatically. That is, the developer can manually identify existing manipulation systems that have been frequently identified and neutralized by unauthorized third parties, as well as those that have not. This allows for the particularly reliable identification of manipulation systems and enables the application of the underlying manipulation procedures to newly developed manipulation systems. Consequently, the newly developed manipulation systems are also more reliably guaranteed not to be identified by unauthorized third parties. Preferably, artificial intelligence is used for the development of the new manipulation systems, thereby reducing the effort required for their development. Furthermore, this reduces the risk of human error. Artificial intelligence can identify patterns with particular reliability and, consequently, derive manipulations that are particularly difficult for unauthorized third parties to identify. Appropriately trained artificial intelligence is used.

[0046] Additionally, data-type operation rules can be predefined that can be randomly combined to create new operation schemes. Data-type operation rules can, for example, define whether an operation provides an addition to an existing item or a format change. Data-type operation rules can also pre-specify where within each item the corresponding operation should be performed. Data-type operation rules can specify what type of addition and / or what type of format change it is. All of these data-type operation rules can be randomly combined with one another. This allows new operation schemes to be generated automatically, particularly reliably and simply.

[0047] Preferably, a pre-trained generative "Transformer" is used as artificial intelligence. By using such a pre-trained generative transformer, it is possible to provide digital watermarks that are difficult for unauthorized third parties to identify with particular reliability. By analyzing numerous combinations of the marked dataset and the corresponding dataset under inspection, this transformer can reliably infer the analysis patterns of unauthorized third parties. Therefore, the transformer can detect manipulations in the dataset in the same way that unauthorized third parties can. Consequently, the pre-trained generative transformer can derive a manipulation scheme that the aforementioned unauthorized third party cannot replicate. This allows data marked with digital watermarks to be protected with particular reliability.

[0048] Furthermore, an additionally advantageous configuration of the method according to the present invention is that the computing unit generates different marked data sets from an initial data set, and for this purpose, the computing unit generates different operations by applying various operation schemes to at least one specific item, and the computing unit provides a first marked data set to be used by a first subject and at least one additional marked data set to be used by at least one additional subject. That is, the computing unit can generate different marked data sets from the same starting data for different customers. In the analysis phase of the data set under inspection, if it is confirmed that the data set was generated internally within the company, it is possible to identify which subject, i.e., which potential customer, the data set under inspection originated from based on the subject-specific operations that may be found in the data set under inspection. Thus, it is possible to identify unauthorized subjects and take appropriate measures to reduce or even completely prevent the misuse of unique data by said subjects. For example, this allows for the detection of security gaps and the cessation or restriction of cooperation with said subjects. Furthermore, various instances within a company, such as departments, may apply different operational systems to generate different marked data sets. This allows for the clear identification of the source, recipient, and / or seller of the data using digital watermarks.

[0049] According to the present invention, in a system for marking a digital watermark on data comprising a computing unit, the computing unit is configured to perform the method described above. The system of the present invention may be implemented in a local environment, e.g., on a desktop computer, or distributed in a cloud environment. The computing unit may be a server or a server cluster or a component of a server or a server cluster.

[0050] The system of the present invention has access rights to a computer-readable storage medium on which a computer program product is stored, and has computer-readable instructions, and when these instructions are executed by a processor of a computing unit, the computing unit performs the method described above.

[0051] Further preferred configurations of the computer implementation method according to the present invention for marking a digital watermark on data also arise from the embodiments described in more detail below with reference to the drawings. Brief explanation of the drawing

[0052] FIG. 1 is a schematic diagram illustrating a method known in the prior art for marking a digital watermark on a tabular data set. FIG. 2 is a schematic diagram illustrating a method according to the present invention for marking a digital watermark on a tabular data set. FIG. 3 is a schematic flowchart of a computer implementation method according to the present invention for marking a digital watermark on data. FIG. 4 is a schematic diagram illustrating an exemplary operation of an item in a data network to be provided with a digital watermark in the form of a number. Specific details for implementing the invention

[0053] Digital watermarks (1) can be used to verify the authenticity of a computer-readable file and ensure traceability. For this purpose, FIG. 1 illustrates an exemplary table (10) as a computer-readable file. The table (10) includes a header (11) in addition to items (4). Artificially generated items (4*) having specific content and / or a specific format are added as digital watermarks (1) at appropriate locations in the table (10). If each artificially generated item (4*) is present in the table (10), it means that the table (10) is authentic. The digital watermark (1) was inserted into the table by the creator of the table (10). However, as illustrated in FIG. 1, the digital watermark (1), the row indicated by the hatch in this figure, can be removed from the table (10) by an unauthorized third party. Consequently, the authenticity of the file can no longer be determined. For this to happen, the unauthorized third party must know what content or format the digital watermark (1) has. This is possible relatively simply. In addition, there is a high risk that artificially generated items (4*) will be removed from the table (10) due to table adjustments already applied by a third party.

[0054] In contrast, FIG. 2 illustrates a method according to the present invention for marking a digital watermark (1) on a computer-readable file. In this figure, an artificially generated item (4*) is added to the corresponding table (10) as a digital watermark (1), and a manipulation (6) is provided to each item (4) of the corresponding initial data set (3). Even if an unauthorized third party removes at least some of the artificially generated item (4*) and the manipulation (6), the manipulation (6) is still sufficiently present, and thus the authenticity of the data set can be determined.

[0055] In this case as well, the table (10) or the initial data set (3) may have a header (11) to which no manipulation (6) is applied to the content. However, it is conceivable that the manipulation (6) is also provided to the header.

[0056] The manipulation (6) must not be identifiable or traceable by an unauthorized third party, because otherwise it could be easily removed or imitated. Furthermore, the provision of the corresponding manipulation (6) to each initial data set (3) must not alter the initial data set to the extent that it can no longer be used for its intended purpose. Additionally, enough manipulations (6) must be provided to the items (4) of the initial data set (3) so that the authenticity of the data set under consideration can be reliably determined even if a large number of the corresponding manipulations (6) are removed. To this end, a computer-implemented method according to the present invention for marking digital watermarks (1) on data is used, and the sequence of the method is schematically illustrated in FIG. 3. Sub-figures 3a through h correspond to each method step.

[0057] In step a), the initial data set (3) is provided here as well in the form of a table (10). The initial data set (3) has a structured format. This means that the items (4) of the initial data set (3) have or follow an array system defined in the initial data set (3). The array system can be defined such that a header (11) comes before the items (4). Since the initial data set (3) in this drawing is a table (10), this array system provides that each item (4) is arranged according to rows and columns. Each item (4) corresponds to a specific data type. Any known data type verified as valid, such as boolean, byte, decimal, long, string, integer, real number, etc., may be used.

[0058] In step b), the computing unit (2) performing the method according to the present invention reads at least one operation system from a watermark database (5) that is communicationally connected to the computing unit (2).

[0059] In step c), the computing unit (2) generates a marked data set (7) by applying an operation (6) according to each operation system to each item (4) of the initial data set (3). Preferably, the computing unit (2) combines a plurality of various operation systems. Each operation system corresponds particularly to each data type of each item (4). For example, various operations (6) can be applied to text as numbers. Each operation (6) generates a digital watermark (1). The pattern of the operation (6) itself is also such a digital watermark.

[0060] As an operation (6), each item (4) may be added to / added at any designated location, or the format of any designated character may be changed in any designated way. Additionally, artificially generated items (4*) may be inserted into each initial data set (3) to generate each marked data set (7). The marked data set (7) is generated by providing the operation (6). In FIG. 3, these artificially generated items (4*) are indicated by hatching. Multiple artificially generated items (4) are added to the table (10) to form a complete artificially generated row.

[0061] In the illustrated embodiment, the data types of the items (4) are distinguished among the respective columns of the table (10). For example, the Vehicle Identification Number (VIN) may be entered in the first column, the name of each vehicle buyer or vehicle owner in the second column, and the sensor value in the third column. For data types in the form of strings, an underline may be inserted at a suitable location, for example, by operation (6). Such an underline may be added, for example, after the first three consecutive alphabets. Strings containing both alphabets and numbers may be automatically capitalized. Additionally, for numbers, the concept of generating a so-called checksum may be applied.

[0062] For example, an artificially generated item (4*) may be added to the table (10) for every 1,000 lines or rows. The artificially generated item (4*) or each initial data set (3) and / or marked data set (7) may be stored in whole or in part in a separate database via the computing unit (2). The computing unit (2) may optionally store additional metadata, such as timestamps, customer IDs, the number of rows and columns of the table (10), access dates, memory sizes, or file sizes.

[0063] Preferably, the artificially generated item (4*) is created after other operations (6) have already been implemented. That is, the artificial item (4*) can be generated according to the operations (6). For example, in the case of strings, the most frequent value or mode of each row can be randomly selected. Also, an average value can be determined for each column and summed for each number. The artificially generated item (4*) can also be created by simply copying after any table item is selected.

[0064] Next, the data set (7) marked in step d) is distributed to different users for use. In the meantime, the underlying data may be manipulated or used without authorization. For example, an unauthorized third party may directly resell the data or modify it in an unauthorized manner. In this case, it is not necessary for each marked data set (7) to be the original entity to which it was distributed. That is, there may be security gaps in these entities, and the marked data set (7) may be stolen.

[0065] At a later time in step e), the data set to be examined (8) is provided to the computing unit (2). If it is suspected that the data set to be examined (8) is based on the marked data set (7), verification is now required.

[0066] To this end, the computing unit (2) determines the probability of a match (9) of the data set (8) to be examined based on the operation (6) according to at least one operation scheme found in the data set (8) to be examined in step f). To this end, the computing unit (2) accesses the watermark database (5) and checks whether each operation (6) can be found in the data set (8) to be examined. This includes checking for the presence or absence of the corresponding artificially generated item (4*).

[0067] Since it is nearly impossible for an unauthorized third party to directly generate these artificially generated items (4*), a high probability of match (9) can be determined when these artificially generated items (4*) are present in the data set (8) under examination. The more operations (6) are detected by the computing unit (2), the higher the probability of match (9) is. As for the threshold for the ratio of operations (6) found in the data set (8) under examination to generally expected operations (6), it has been proven that 10% is appropriate for checksum-type operations (6) and 5% is appropriate for other operations (6). For example, if 15 operations (6) other than checksums are provided and seven of them are removed, the probability of match is 8 / 15. Since this is greater than 5%, it is highly probable that the data set (8) under examination is based on the marked data set (7) or the basic initial data set (3). For example, one of the total six operations (6) provided in the form of a checksum can be removed, and the probability of a match is 5 / 6. Since this is also greater than the above 10%, it is highly probable that this data set (8) subject to inspection originated from within the company.

[0068] To calculate the probability of a match (9), the computing unit (2) may apply a wide variety of rules. For example, individual percentage values ​​may be defined for how many of the operations (6) performed for different operation systems should be included. All these percentage values ​​may be aggregated together to become the probability of a match (9). For example, the computing unit (2) may generate an average value here. For example, the discovery of operations (6), particularly artificially generated items (4*), may be used as a strict boundary condition so that if at least one artificially generated item (4*) is present, the probability of a match may increase to 100%.

[0069] In step g), the computing unit (2) compares whether the corresponding operation (6) has been removed from each item (4) for the inspection target data set (8) where the match probability (9) exceeds a defined threshold. To do this, the computing unit (2) compares the contents of the inspection target data set (8) with the contents of the corresponding marked data set (7). At this stage, the computing unit (2) can identify items (4) already included in the marked data set (7) and items directly added by an unauthorized third party. Then, the operation scheme from which the corresponding operation (6) has been removed is not reused by the computing unit (2) to provide the digital watermark (1) and is replaced with another operation scheme. This may also apply to operations (6) that may be found in items directly created by the aforementioned third party. Various criteria for when each operation scheme is replaced may be defined. Thus, it can be expected that not all items (4) manipulated according to the same operation scheme have been altered by an unauthorized third party. Accordingly, regarding the frequency at which operations (6) of an operation system are removed, a threshold may be defined at which the operation system is replaced when exceeded. For example, if 75% of all operations (6) applied by a specific operation system are found and are removed from the data set (8) under inspection, the operation system may need to be replaced.

[0070] The method according to the present invention is based on the iterative generation of the corresponding marked data set (7) and the analysis of the data set (8) to be inspected that is fed back to the computing unit (2). Accordingly, the computing unit (2) learns to use increasingly reliable manipulation systems, thereby enabling reliable verification of the authenticity of the data. To this end, increasingly new manipulation systems are added to the watermark database (5). These new manipulation systems can be defined, for example, manually by a developer, automatically by AI, or automatically based on a probabilistic combination of predefined manipulation rules for each data type. This iterative procedure is illustrated in FIG. 3h).

[0071] FIG. 4 illustrates, exemplarily, the sequence for generating a checksum as an operation (6). For this purpose, each item (4) contains a number or sequence of numbers. A number suitable for the alphabet "C" must be found. In step 401, each digit of the number is considered individually, and every second digit is multiplied by 2.

[0072] In step 402, the sum of the numbers for each number obtained in this way is determined.

[0073] Now, all the numbers obtained in this way are added up in step 403.

[0074] The remainder when the result is divided by 10 must be, for example, 0. Therefore, in step 404, "C" is determined to be 9.

[0075] The verification of whether such an operation (6) exists is performed in the same manner. That is, the computing unit (2) directly determines the value for "C" and compares it with the value 9. If the two values ​​match, this means that the corresponding operation (6) provided by the computing unit (2) is present in the data set (8) to be checked. The flowchart illustrated in FIG. 4 is an example. All possible additional methods may be used to calculate the checksum and provide each operation (6).

[0076] A system for performing a computer implementation method according to the present invention is also part of the present invention.

[0077] Particularly preferably, the computing unit (2) generates different marked data sets (7) for various subjects or customers, thereby enabling the identification of when and where the data was "leaked." The authenticity of the data set (8) under inspection can be verified using the method according to the present invention. Thanks to repeated execution, the computing unit (2) uses an increasingly reliable operation system.

Claims

Claim 1 A computer-implemented method for marking a digital watermark (1) on data, the following method steps performed by a computing unit (2): a) Step b) providing an initial data set (3), wherein the initial data set (3) is in a structured format such that the items (4) of the initial data set (3) have an array system defined in the initial data set (3). c) reading at least one operation system from a watermark database (5) that is connected to the computing unit (2) via communication; Step of generating a marked data set (7) by applying an operation (6) according to the operation system to each item (4) of the initial data set (3); d) Step of providing the above-mentioned marked data set (7) for use; e) Step of receiving the data set (8) to be examined;f) Step of determining the probability of agreement (9) of the data set to be examined (8) based on the operation (6) according to at least one operation system found in the data set to be examined (8); g) If the above match probability (9) is greater than a specified threshold: comparing an item (4) of the above-mentioned data set (8) with a corresponding marked data set (7) determined for the above-mentioned data set (8), and determining an item (4) from which the operation (6) applied in step c) has been removed in the above-mentioned data set (8); and h) A computer implementation method characterized by a step of repeating steps a) through g), wherein for items (4) having the same data type as the item (4) for which the operation (6) was confirmed to have been removed in step g), the computing unit (2) uses a different operation system in the watermark database (5). Claim 2 A computer implementation method according to claim 1, wherein the computing unit (2) adds each item (4) according to an operation pattern to manipulate items (4) in the initial data set (3), in particular, adds a number obtained by applying a mathematical operation to an item (4) composed of a numeric value and / or changes at least one format of each item (4) according to an operation pattern. Claim 3 A computer-implemented method according to claim 1 or 2, wherein the computing unit (2) adds at least one artificially generated item (4*) to the initial data set (3) for additional marking, and in particular stores the artificially generated item (4*) in an artificial item database, which includes arrangement information describing the arrangement of the artificially generated item (4*) in the initial data set (3), and the computing unit (2) checks whether at least one artificial item (4*) stored in the artificial item database can be found in the inspection target data set (8) in order to determine the match probability (9) in step f). Claim 4 A computer implementation method characterized in that, in any one of claims 1 to 3, the structured format is a table (10). Claim 5 In any one of claims 1 to 4, the computing unit (2) increases the match probability (9) in step f), wherein:- Excluding checksum-type operations (6), if the ratio of the number of operations (6) still existing in the inspection target data set (8) to the total number of operations (6) applicable to the item according to the operation system used is greater than 5%; In the case of a checksum-type operation (6), if the ratio of the number of operations (6) still existing in the inspection target data set (8) to the total number of operations (6) applicable to the item according to the operation system used is greater than 10%; A computer implementation method characterized by increasing the matching probability when at least one artificially generated item (4*) is included in the above-mentioned inspection target data set (8). Claim 6 In any one of paragraphs 1 to 5, a new operation system is added to the watermark database (5), and the new operation system is:- Manually defined by the developer;- Automatically defined by artificial intelligence, particularly considering the frequency of removal of operations (6) in the above-mentioned inspection target data set (8); or- A computer implementation method characterized by automatically defining predefined data type-specific operation rules through probabilistic combination. Claim 7 A computer implementation method according to claim 6, characterized in that a pre-trained generative transformer is used as artificial intelligence. Claim 8 A computer implementation method characterized in that, in any one of claims 1 to 7, the computing unit (2) generates different marked data sets (7) from the initial data set (3), the computing unit (2) generates different operations (6) by applying various operation systems to at least one specific item (4) for this purpose, and the computing unit (2) provides a first marked data set to be used by a first subject and at least one additional marked data set to be used by at least one additional subject. Claim 9 A system for marking a digital watermark (1) on data, comprising a computing unit (2), wherein the computing unit (2) is configured to perform a method according to any one of claims 1 to 8.

Images