System and method for providing data privacy risk for data privacy related data objects to allocate access rights

By introducing data extraction, privacy risk calculation and risk mitigation modules into the system, the privacy risks in the data file combination are evaluated and mitigated, and the problem of inaccurate allocation of access rights in the prior art is solved, and the efficiency of data privacy protection is improved.

CN119096244APending Publication Date: 2024-12-06SIEMENS AG
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202380036444.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Priority Date
2022-04-26
Filing Date
2023-03-23
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

The prior art is difficult to effectively evaluate and mitigate the privacy risks in data file combinations, especially in data in different data sources and formats, resulting in the inability to accurately allocate access rights, increasing the risk of leakage of personal information and restricted entry data.

Method used

A system was designed, including a data extraction module, a privacy risk calculation module and a risk mitigation module. By extracting data, calculating privacy risks and applying mitigation functions, it generates identity space, identifying individuals and assigning access rights categories, which are suitable for different data formats and Type, reduce privacy risks.

Benefits of technology

It realizes the privacy risk assessment and mitigation of data file combinations, accurately allocate access rights, reduces the risk of leakage of personal information and restricted entry data, and improves the efficiency of data privacy protection.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119096244A_ABST
    Figure CN119096244A_ABST
Patent Text Reader

Abstract

A system for providing data privacy risks of data privacy related data objects to allocate access rights for processing and storing data objects (DO), comprising: a data extraction module (21) configured to receive more than one data object (DO) and to extract data from the data object; -an identity definition module (23) configured to generate at least one identity space, where the identity space comprises those data objects comprising data elements extracted with respect to a group of individuals or a subset of the group of individuals, and where the group of individuals is different from the group of individuals of each other identity space; a privacy risk calculation module (24) configured to calculate at least one privacy risk (PR) for each identity space (40, 41, IS) and for the contained individuals as a function of a probability of re-recognition of each individual based on the extracted data elements in the identity space (40, 41, IS) and as a function of a set of context knowledge assumed to be known; -a derivation module (26) configured to derive the extracted data and to derive a calculated privacy risk (PR).
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to a system, a computer-implemented method, and a computer program product for assigning access rights to a data object including data related to an individual (ie, related to data privacy). Background Art

[0002] Data with personal information is protected by the General Data Privacy Regulation (e.g. EU law). The General Data Privacy Regulation (GDPR) strongly encourages the application of information minimization and anonymization and pseudonymization to protect individuals and their data.

[0003] The risk of identifying a person and therefore their information should be as low as possible. Anonymized data will have the benefit that it is no longer considered personal data and can be freely used and accessed without the requirements or restrictions given by the GDPR. The benefit of data with pseudonymized information and a limited amount of personal information is that when a leak or some type of data loss occurs, the damage regarding data privacy violation will be reduced.

[0004] The same applies to data that includes information about restricted items like trade secrets, such as details of the manufacturing process, product composition, etc.

[0005] Often, it is not apparent at first glance that data is considered personal data or data relating to a restricted item, as it can also be non-directly personal or restricted item identifying data, for example a time series of temperature data measured in an apartment or around a machine, which when connected with other data (e.g. a person's shift schedule or a maintenance schedule for a machine) can produce individually relevant information.

[0006] Therefore, the storage, processing or sharing of data containing personal information or restricted items must be limited, depending on the risk of identifying individuals and restricted items through the data. This is especially the case when data from different sources and concerning different aspects of individuals or restricted items should be stored in the same data storage and / or should be accessible and processed by the same data user.

[0007] US2021 / 0150056 A1 discloses a privacy management platform that is configured to scan identity, primary and / or secondary data sources to provide users with visibility into stored personal information, risks associated with storing such information, and usage activities related to such information. The platform can associate personal information with specific data subjects to provide an index catalog across multiple data sources.

[0008] US2018 / 114037 A1 discloses a method for assessing the risk of re-identification of a single individual in a data set. An individual, subject, or patient is described by a data subject profile, such as a record in a data set. The population distribution is determined by one or more quasi-identification fields identified in the data subject profile. An anonymity value is calculated based on the aggregate information value associated with the data set and the population size. A re-identification metric for the individual is then calculated based on the anonymity value.

[0009] Identifying and mitigating privacy risks in data is difficult to achieve due to the fact that the information of an individual or restricted entry can be scattered across different data sources and files (databases, csv files, spreadsheets, images, pdfs, to mention just one view) in different formats and types. The representation of an individual or restricted entry may be defined by a combination of all files or a sub-sample of files. The risk of identifying an individual or restricted entry through such a combination of files is unknown, and therefore the risk of allowing access to such a combination of files is also unknown. Summary of the invention

[0010] Therefore, one object of the present application is to provide a solution for evaluating the privacy risk for a combination of data files so as to enable allocating access rights to the combination of data files based on the privacy risk. Another object is to provide a possibility of reducing the privacy risk of a combination of data files so as to allocate access rights.

[0011] This object is solved by the features of the independent claim. The dependent claims contain further developments of the invention.

[0012] A first aspect relates to a system for providing a data privacy risk of a data privacy-related data object including data related to an individual to assign access rights for processing and storing the data object, the system comprising:

[0013] - a data extraction module configured to receive more than one data object and extract data from said data objects,

[0014] - generating at least one identity space, wherein the identity space comprises those data objects containing data elements extracted about the same set of individuals or a subset of the set of individuals, and wherein the set of individuals is different from the set of individuals of every other identity space,

[0015] a privacy risk calculation module configured to calculate at least one privacy risk for each identity space and for the individuals contained therein, depending on a probability of identification of each individual based on the extracted data elements in said identity space and on a set of background knowledge assumed to be known,

[0016] - an export module configured to export the extracted data, for example to a data storage unit, and to export the calculated privacy risk for assigning access rights categories depending on the calculated privacy risk required to access and / or process the extracted data.

[0017] "Individuals" include personal or restricted entries, such as trade secrets, product composition or production details of specialized products, etc. The system evaluates data objects with respect to the data privacy of individuals. Data objects can be data files or data tables, but can also be a selection of data retrieved from a database. "Extracted data elements" are those extracted data that can be related to individuals. "A set of background knowledge assumed to be known" constitutes the prerequisite. The calculated privacy risk provides the probability of identifying individuals within the group of individuals by connecting a subsample of the extracted data elements with "the set of background knowledge known" by, for example, data users who have access to the stored data objects. If the calculated privacy risk is high, an access permission category is assigned that is only provided to a few users (such as individuals whose information is contained in the data object, or a group of trusted users).

[0018] The modules of the system provide a processing pipeline to evaluate more than one data object with respect to data privacy and enable access to the extracted data to be restricted by access rights categories. The extracted data can be combined / joined to get a complete view of an individual across all data sources and data types.

[0019] In one embodiment of the system, the data extraction module is configured to receive data objects structured in different file formats and / or from different types of data sources.

[0020] This allows considering various combinations of data objects with data sources, such as file systems, file shares, cloud storage, databases, and formats, such as structured or unstructured, such as images in jpeg or png format, csv files, xls files, pdf files, or data retrieved from a database.

[0021] In one embodiment, the data extraction module is configured to detect a format of the data object required to parse the data object.

[0022] This facilitates the handling of data objects formatted in a wide variety of data formats. The data format can be automatically identified and corrections can be proposed for incorrectly formatted data.

[0023] In one embodiment, the data extraction module is configured to receive a correction instruction from a user to correct the data format, and extract data from the data object based on the correction instruction.

[0024] This enables data to be imported even if the format cannot be automatically determined. The user is the person or device to whom the privacy risk and extracted data is provided.

[0025] The system also includes a risk mitigation module, which is configured to perform at least one mitigation function on the data of the identity space if the calculated privacy risk exceeds a predefined threshold, and the result is mitigated data that may reduce the probability of identification of each individual. The risk mitigation module is configured to provide the mitigated data to the privacy risk calculation module, which is configured to calculate a new privacy risk based on the mitigated data of the identity space, and output the mitigated data and the new privacy risk via the export module.

[0026] This provides the possibility to mitigate the risk by using mitigation techniques or algorithms implemented on the data in the system.The privacy risk calculated for the mitigated data provides a measure to evaluate the mitigation process and the mitigated data with respect to the privacy of the data.

[0027] In one embodiment, the system further comprises a semantic extraction module configured to detect data elements of at least one semantic type in the extracted data, wherein each semantic type represents a specific semantic content of the semantic data element.

[0028] This enables the selection of a mitigation function depending on the detected semantic type. This results in a better alignment of the selected mitigation function with the information or content of the extracted data.

[0029] In one embodiment, the semantic extraction module is configured to receive user adjustments to the detected instances of the semantic type.

[0030] This enables correcting misclassifications of semantic types. Thus, the selection of mitigation functions can be enhanced by providing the correct semantic type as a basis for selection.

[0031] In one embodiment, the semantic extraction module is configured to adapt detection methods for detecting data elements of semantic types and / or add at least one new detection method by authorized personnel.

[0032] This enables continuous improvement in semantic type detection.

[0033] In one embodiment, the identity definition module is configured to: for each individual in the identity space, select a target personal identifier from all extracted data elements related to the individual in the identity space, and wherein the target personal identifier uniquely identifies the individual in the identity space.

[0034] The target personal identifier allows more than one extracted data element to be associated with the same individual.

[0035] In one embodiment, the identity definition module is configured to create at least one further identity space on a different subset of the extracted data elements.

[0036] Another identity space includes different combinations of extracted data elements for different individuals not covered by the previously created identity space. This provides a holistic view of personal information across different data sources and formats through information combinations, forming one or several groups of individuals.

[0037] In one embodiment, the risk mitigation module is configured to output at least one proposed mitigation function and receive a selected mitigation function selected by a user to be applied to the identity space.

[0038] This allows for the inclusion of additional knowledge of the user and improves the mitigation process.

[0039] In one embodiment, the risk mitigation module is configured to propose the at least one mitigation function to the user based on the calculated privacy risk for the identity space.

[0040] This allows for a flexible adaptation of the mitigation function depending on the privacy risk and optimizes the degree of mitigation applied to the identity space.

[0041] In one embodiment, the risk mitigation module is configured to propose the at least one mitigation function based on a semantic type of data element to be converted in the identity space.

[0042] This allows for flexible adaptation of mitigation functions tailored to the meaning (ie, semantic type of the data element).

[0043] A second aspect relates to a computer-implemented method for providing a data privacy risk of a data privacy-related data object including data related to an individual to assign access rights for processing and storing the data object, the method comprising:

[0044] - receiving more than one data object and extracting data from said data objects,

[0045] - generating at least one identity space, wherein the identity space comprises those data objects containing data elements extracted about the same set of individuals or a subset of the set of individuals, and wherein the set of individuals is different from the set of individuals of every other identity space,

[0046] - calculating at least one privacy risk for each identity space and for the individuals contained therein, based on the probability of identification of each individual based on the extracted data elements in said identity space and based on a set of background knowledge assumed to be known,

[0047] - exporting the extracted data, for example to a data storage unit, and exporting the calculated privacy risk, in particular for assigning an access rights category depending on the calculated privacy risk required for accessing and / or processing the extracted data.

[0048] A third aspect relates to a computer program product directly loadable into the internal memory of a digital computer, comprising software code portions for performing the steps as previously described when said product is run on said digital computer.

[0049] A fourth aspect relates to the use of a system as claimed above, wherein the system is applied to assign access rights categories to extracted data depending on the calculated privacy risk required to store, access and / or process the extracted data.

[0050] Alternatively, it relates to a computer-readable storage medium having stored thereon computer program instructions which, when executed by a processor, perform the steps of the method according to the preceding claim. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The present invention will be explained in more detail by referring to the accompanying drawings. Similar objects will be marked with the same reference symbols.

[0052] Figure 1 An embodiment of the system of the present invention is schematically illustrated.

[0053] Figure 2 An embodiment of generating an identity space from data objects in different data sources is schematically illustrated.

[0054] Figure 3 An embodiment of the method of the invention is shown as a flow chart. DETAILED DESCRIPTION

[0055] Note that in the detailed description of the following embodiments, the drawings are only schematic, and the illustrated elements are not necessarily shown in scale. On the contrary, the drawings are intended to illustrate the functions and collaboration of the components. Here, it should be understood that any connection or coupling of a functional block, module or other physical or functional element can also be implemented by an indirect connection or coupling, such as via one or more intermediate elements. The connection or coupling of an element or module can be implemented, for example, by a wired connection, a wireless connection and / or a combination of a wired connection and a wireless connection. The module can be implemented by dedicated hardware (such as a processor, firmware) or software, and / or by a combination of dedicated hardware and firmware and software. It should be further noted that each module described for the system can perform the functional steps of the related method, and vice versa.

[0056] Data relevant to data privacy is data containing personal information that is protected by law, such as the General Data Privacy Regulation (GDPR) under EU law. In addition, data relevant to data privacy is data about restricted items that should be kept confidential and cannot be retrieved or exported by any unauthorized individual. Individuals or restricted items are further referred to as individuals. Information about individuals can be scattered across several different data sources, and the risk of inferring privacy-related information from these several data sources can be higher than the risk provided by one single data source. Data users or data processing pipelines do not have the skills to estimate whether there is any privacy risk for the data contained in the data source as a whole, and in particular do not have the skills to estimate which information in the data is responsible for the risk. The lack of skills may be due to limited data and privacy knowledge, but also due to the complexity of the information, for example, the amount of information scattered across different data sources that can be combined to form a privacy risk.

[0057] The technical means of providing data privacy is to restrict access to one or several data objects that are used to retrieve or process data that are relevant to data privacy, and to provide access rights. In order to provide adequate access, different access rights categories can be assigned based on risk so that conclusions can be drawn about individuals based on the accessed data. The proposed system evaluates the data source and outputs the privacy risk of the received data object and derives the privacy risk, which is used to assign access rights categories based on the derived privacy risk.

[0058] Figure 1 A system 20 is shown that includes a data extraction module 21, an identity definition module 23, a privacy risk calculation module 24, and an export module 26. The data extraction module 21 is configured to receive more than one data object, extract data from the data object, and forward the data object including the extracted data to the identity definition module 23. The identity definition module 23 is configured to generate at least one identity space, wherein the identity space includes those data objects that contain extracted data elements about a set of individuals or a subset of the set of individuals, and wherein the set of individuals of this identity space is different from the set of individuals of each other identity space. The privacy risk calculation module 24 is configured to calculate at least one privacy risk for each identity space and for the contained individuals based on the probability of re-identification of each individual based on the extracted data elements in the identity space and based on a set of background knowledge assumed to be known. The export module 26 is configured to export the extracted data and the privacy risks calculated for the extracted data.

[0059] In one embodiment, the system 20 additionally includes a semantic extraction module 22 configured to detect semantic data elements, which are data elements of at least one semantic type in the extracted data, wherein each semantic type represents a specific semantic content of the semantic data element. In another embodiment, the system 20 additionally includes a risk mitigation module 25 configured to select a mitigation function based on the detected semantic type.

[0060] In another embodiment, the system 20 further includes a risk mitigation module 25, which is configured to perform at least one mitigation function on the extracted data of the data object in the identity space if the calculated privacy risk exceeds a predefined threshold. The mitigation module 25 outputs the mitigated data modified in this way to provide less or no information to identify the individual. The mitigated data is forwarded and input to the privacy risk calculation module 24. The privacy risk calculation module 24 processes the mitigated data and calculates at least one new privacy risk based on the mitigated data of the identity space. The mitigated data and the new privacy risk are forwarded to the export module 26 to output the mitigated data and the new privacy risk. Optionally, the mitigated data can be provided as input to the identity definition module 23 and processed by the identity definition module 23 and the privacy risk calculation module 24 in a loop, for example, until the resulting privacy risk reaches the predefined privacy risk. The processing in the identity definition module 23 is optional in the subsequent loop because changes in the identity space may not be required.

[0061] Each module of system 20 is described in more detail below.

[0062] The data extraction module 21 is configured to receive more than one data object 10 and extract data from more than one data object 10. The data extraction module 21 is configured to receive data objects 10 structured in different formats and / or from different types of data sources. The more than one data objects 10 may include, for example, a text file 11, an audio file 12 or an image file 13. It is further configured to detect the format of the data object required to parse the data object. The data object 10 may include not only a data file or a data table, but also a table from a database.

[0063] The data extraction module 21 supports the loading / importing of data objects 10 from different data sources (e.g., file systems, databases, shares, cloud storage) and data objects of different formats and data types (e.g., structured or unstructured) (e.g., csv, excel, pdf, images, databases). The format of the data object is automatically detected. It extracts or predicts format parameters for a given data object to parse the data. In one embodiment, the system includes a user interface 27 (e.g., a graphical user interface) configured to display the extracted or predicted parameters to the user. The user can adjust the parameters and enter corresponding correction instructions via the user interface 27. The data extraction module 21 receives the correction instructions, corrects the data format, and extracts data from the data objects 11, 12, and 13 based on the correction instructions respectively.

[0064] The extracted data is optionally forwarded to the semantic extraction module 22 before being forwarded to the identity definition module 23. Semantic types are automatically detected in the extracted data based on the content and data type. Semantic types help understand the occurrence of specific information in the data at a higher and more abstract semantic level, for example, email address, address, name, etc. for all specific email addresses found. The format of semantic data elements that can be assigned to semantic types can be independently detected in structured data (such as csv or tabular data), unstructured data (such as text files), or image data (such as with facial or text recognition).

[0065] It is possible to find several semantic types for the same data element. For example, the semantic type "car name" and the semantic type "personal name" can be assigned to the same data element. The automatic detection function in the semantic extraction module 22 is based on a combination of a recognition library, a search pattern (e.g., a regular expression, and an AI algorithm) to infer and identify the most likely semantic type. The semantic data element and the semantic type assigned to it can be forwarded to the user interface 27. The user interface 27 can receive corrections to the data elements of the semantic type by deleting the data elements of the semantic type or highlighting and adding the missing data as the data elements of the semantic type. If the data element is incorrectly classified, the assigned instance of the semantic type (label) can be adjusted. In addition, by defining a recognition method (e.g., a regular expression or an artificial intelligence (AI) algorithm) and / or adding a recognition method (e.g., a regular expression or an artificial intelligence (AI) algorithm) to the system 20, a completely new semantic type can be received via the user interface 27.

[0066] In addition, a hierarchy of semantic types may be defined. As an example, a hierarchy level includes "car name" and the next lower level includes specific car names from different manufacturers. This allows for more fine-grained mitigation methods to be selected or provided in the risk mitigation module 25. Mapping semantic types to extracted data elements in a data object provides the advantage that the user better understands what kind of information is available in the data object.

[0067] The extracted data is forwarded to an identity definition module 23, which is configured to generate at least one identity space, wherein the identity space includes those data objects containing data elements extracted about a group of individuals or a subset of the group of individuals, and wherein the group of individuals is different from the group of individuals in every other identity space. In order to calculate the risk of identifying an individual, it is necessary to know which parts of the data belong to the individual. Each part of the data belonging to or related to the individual is called an extracted data element. The identity definition module 23 identifies all extracted data elements belonging to the individual on several data objects.

[0068] Figure 2 The generation of the identity space is illustrated in FIG. Two data objects DO1, DO2 are received from a data source 30, and data objects DO3, ..., DO6 are further received from the data source 31 via a data extraction module 21. Data is extracted from data objects DO1, ..., DO6. In the identity definition module 23, the extracted data elements, i.e., those data about the individual, are identified. The extracted data elements can be one or several entries in a structured data object, for example, a row in a data object 10 structured as a csv file, or one or more occurrences in an unstructured data type, for example, a face or other object in a data object DO3 as an image or a text paragraph including a name in a data object D04 as a text document. The data elements extracted from different data objects DO1, ..., DO6 must be connected to form an overall view of an individual or a group of individuals. All extracted data elements that belong to the individual and one of the group of individuals due to the connection are called "identity space".

[0069] exist Figure 2In the embodiment illustrated in FIG, the identity definition module 23 identified in the data objects DO1, DO2, DO3 extracts data elements about individuals a, b and / or a, b, c. An identity space 40 is generated including data objects DO1, DO2, DO3, which include extracted data elements about individuals a, b and / or a, b, c. No extracted data elements about individuals a, b and / or a, b, c are identified in any other received data objects DO4, DO5, DO6. The identity definition module 23 identifies extracted data elements of individuals d and / or e in data objects DO4, DO6. No extracted data elements about individuals d and / or e are identified in any other received data objects DO1, DO2, DO3, DO5. Therefore, the identity definition module 23 generates an identity space 41 including data objects DO4 and DO6. The identity definition module 23 identifies extracted data elements of individuals e and h only in the data object DO5, but not in the other data objects DO1, ..., DO4, DO6. Therefore, an identity space 42 is generated including the data object DO5.

[0070] In one embodiment, system 20 supports the user in generating an identity space by connecting data elements extracted from different data sources and data types on given common information (i.e., extracted data elements) via user interface 27. In another embodiment, system 20 supports the user in this task by proposing to the user automatically detected connections found by high-level algorithms.

[0071] For structured data objects, several extracted data elements (e.g., lines in a document) may belong to the same individual. Therefore, the extracted data elements or combinations of extracted data elements are marked in the identity space by a target personal identifier that uniquely identifies this individual (i.e., an attribute or combination of attributes that identifies the same individual). In an identity space where several data objects are connected, it is sufficient to define a target personal identifier for a central data object to which other data objects are attached.

[0072] exist Figure 2 In the identity space 40, the target identifier 32 is assigned to the data object DO1. The target personal identifier 33 is not assigned to the data objects DO2, DO3 connected to DO1. In one embodiment, the target personal identifier is automatically identified by the algorithm and forwarded to the user interface to be proposed to the user. If no target personal identifier is defined or detected, each extracted data element in the data object (e.g., structured file) is assigned to another individual. This may lead to unwanted erroneous risk calculations.

[0073] Identification and assignment of extracted data elements is more complex in data objects in unstructured format compared to data objects of structured data, where several extracted data elements belong to each other by definition, because they are in the same file and form information of the same individual (e.g. a row in a csv). In case of data objects in unstructured format, only pieces of information marked with a semantic type are found and extracted from the data object. This is a single piece of information, such as name, address, email, face. Advanced methods can be used to find within this file a piece of information belonging to a specific individual to eventually connect it to data elements extracted from other data objects. It is possible to add only one piece of information. Since the extracted data elements are collected over several data objects, several identity spaces will be created. Each identity space will have a unique group of individuals, see Figure 2 .

[0074] The privacy risk calculation module 24 is configured to calculate the privacy risk for each individual in the identity space. In addition, a risk score is calculated for the entire identity space across all individual privacy risks. The privacy risk represents the probability of re-identifying an individual in the available identity space. It describes how unique an individual is in a given population, assuming that an attacker knows specific extracted data elements of the individual's information and can therefore be identified. If the individual as defined in the identity space is completely unique, the risk of re-identifying the individual is 100%. This is the so-called prosecutor and journalist risk. However, other information can also be included in the risk, for example, security measures are in place to protect this data.

[0075] In this case the user has to indicate what information the attacker might know, and for this combination of attributes the risk is calculated for each individual in the group. This uncovers blind spots, since the user might pick the wrong combination, not knowing that specific available information (i.e. extracted data elements or semantic data elements) leads to a higher risk. In the described method, this information is not asked from the user or provided by the user. The privacy risk calculation module 24 calculates the risk for each possible combination and presents the calculated risk to the user for his further analysis, in particular assigning access rights categories to the extracted data contained in the data objects of the corresponding identity space associated with the individual or group of individuals. Depending on the number of extracted data elements and / or available semantic data elements, or on the hardware on which the system runs, this can be very time consuming.

[0076] The required computation time can be adjusted by the user by applying filters to the required computations. Filters are for example: - exclusion of an attribute because the user knows that this data element is not available to an attacker,

[0077] - the maximum or minimum number of attributes in a combination,

[0078] - includes all combinations of specific attributes, or

[0079] - Maximum tolerable privacy risk.

[0080] All specifications of this combination will lead to higher privacy risk by adding more attributes. Therefore, if the combination is already above a certain threshold, they can skip the calculation.

[0081] The calculated privacy risk is output by the export module 26 and / or provided to the user by the user interface 27. By this, the user has all the information available to decide where there is an unacceptable risk, and which attributes belong to this risk. By combining propagation, he will also be able to see which attributes are mining risks and should be potentially transformed in subsequent steps.

[0082] The risk mitigation module 25 includes a tool set for reducing privacy risks. The risk mitigation module 25 is configured to propose at least one mitigation function based on the privacy risk calculated for the identity space. The risk mitigation module is configured to output at least one proposed mitigation function and receive a selected mitigation function selected by the user to be applied to the identity space. It provides implementations of various mitigation algorithms, such as deletion, aggregation, generalization, k-anonymity, and others. The user can select which (one or more) data elements / attributes to transform, the appropriate algorithm, and the correct parameters. The mitigation algorithms ultimately form a pipeline, which is then executed by the system 20. What is new is that the previously collected information about semantic types and privacy risks helps the user select the right attributes and useful mitigation strategies.

[0083] The user can make an informed decision which data elements have to be converted. By selecting an attribute the user can be further prompted how the information can be converted based on the semantic type. For example, when selecting an attribute where the semantic type "email" is detected the user gets an appropriate suggestion how the email can be converted. For example by removing everything before the "@" symbol with a replacement algorithm and displaying appropriate parameters. The proposed mitigation algorithms are filtered based on the target data format. Certain methods are not possible or useful for e.g. images. These are not displayed to the user.

[0084] The mitigation process is applied to the identity space. Therefore, changes must be applied to copies of different primitive data types in a specific format.

[0085] The data export module 26 is configured to output the transformed, i.e. extracted data, which is modified or corrected by the data extraction module 21. The data export module 26 is configured to output the mitigated data and provide them for further use, in particular for assigning access rights categories for accessing and processing the extracted and / or mitigated data. The data export module 26 provides modes of data aggregation, export and downloading.

[0086] The described system 20 is applied to provide privacy risk to re-identify an individual for extracted data of a data object, and to assign access rights categories according to the calculated privacy risk required to access and / or process the extracted data.

[0087] Figure 3 A method for providing data privacy risks of data privacy-related data objects according to system 20 is illustrated in FIG. In a first step S1 of the method, more than one data object DO is extracted and data is extracted from the data objects. Optionally, at least one detected data element of a semantic type is detected in the extracted data, wherein each semantic type represents a dedicated semantic content of the detected data element, see step S7. At least one identity space IS, 40 is generated in step S2, wherein the identity space includes those data objects containing extracted data elements about a group of individuals or a subset of the group of individuals, and wherein the group of individuals is different from the group of individuals in each other identity space. In step S3, at least one privacy risk PR is calculated for each identity space and for the individuals contained therein based on the probability of re-identification of each individual based on the extracted data elements in the identity space and based on a set of background knowledge assumed to be known. In step S4, the extracted data and the calculated privacy risks are exported.

[0088] Optionally, the calculated privacy risk PR is compared with a predetermined threshold value T, see step S5. If the privacy risk PR is higher than the threshold value T, at least one mitigation function is performed on the data in the identity space, see step S6, so that the result is mitigated data. The mitigated data is expected to have a reduced probability of re-identification of each individual. In alternative scheme Al, the privacy risk is calculated for the mitigated data according to step S3, but based on the mitigated data rather than the original extracted data. In alternative scheme A2, step S2 is performed on the mitigated data, that is, at least one semantic type of data element is detected in the mitigated data. In both alternative schemes A1 and A2, the mitigated data is further processed by steps S3, S5 and S6 until the calculated privacy risk is lower than the threshold value T, or the process is stopped in another way. As a final step, see S4, the mitigated data and the new privacy risk are derived.

[0089] The entire system is built around a holistic, person-centric view. The possibility to combine data of various types, structures, and sources provides a higher level of authenticity and allows to consider privacy risks across domains and media types. As mentioned above, the person-centric view has implications for the entire pipeline. The definition of the identity space is new, the way of calculating risk is new, and the mitigation and modification in different sources are new approaches.

[0090] It is to be understood that the above description of the examples is intended to be illustrative, and the illustrated components are susceptible to various modifications. For example, the illustrated concepts can be applied to different technical systems, and in particular to different subtypes of the corresponding technical systems with only minor modifications.

Claims

1. A system for providing data privacy risk of a data privacy-related data object including data related to an individual to assign access rights for processing and storing said data object (DO), comprising: - a data extraction module (21) configured to receive more than one data object (DO) and to extract data from said data objects, - an identity definition module (23) configured to generate at least one identity space, wherein the identity space comprises those data objects containing extracted data elements about a set of individuals or a subset of the set of individuals, wherein the extracted data elements are those extracted data that can be associated with the individuals, and wherein the set of individuals is different from the set of individuals of every other identity space, a privacy risk calculation module (24) configured to calculate at least one privacy risk (PR) for each identity space (40, 41, IS) and for the individuals contained therein, based on a probability of re-identification of each individual based on the extracted data elements in said identity space (40, 41, IS) and based on a set of background knowledge assumed to be known, - an export module (26) configured to export said extracted data and to export said calculated privacy risk (PR), Additionally includes: - A risk mitigation module (25) configured to: If the calculated privacy risk (PR) exceeds a predefined threshold (T) value, at least one mitigation function is performed on the data of the identity space (40, 41, IS), and the result is mitigated data (MD) that reduces the probability of re-identification of each individual, The mitigated data (MD) is provided to the privacy risk calculation module (25), the privacy risk calculation module (25) being configured to: calculating new privacy risks based on the mitigated data (MD) of the identity space (40, 41, IS); and The mitigated data (MD) and the new privacy risk are output via the export module (26).

2. The system according to claim 1, wherein: The data extraction module (21) is configured to receive data objects (DO) structured in different file formats and / or from different types of data sources.

3. The system according to claim 2, wherein: The data extraction module (21) is configured to detect a format of the data object required for parsing the data object.

4. System according to one of the preceding claims, wherein: The data extraction module (21) is configured to receive a correction instruction from a user to correct a data format, and extract data from the data object (DO) based on the correction instruction.

5. System according to one of the preceding claims, comprising: A semantic extraction module (22) is configured to detect semantic data elements, which are data elements of at least one semantic type in the extracted data, wherein each semantic type represents a specific semantic content of the semantic data element.

6. The system according to claim 5, wherein: The semantic extraction module (22) is configured to receive adjustments to the detected instances of semantic types by the user.

7. The system according to claim 5 , wherein: The semantic extraction module (22) is configured to adapt detection methods for detecting data elements of semantic type and / or to add at least one new detection method by authorized personnel.

8. System according to one of the preceding claims, wherein: The identity definition module (23) is configured to: for each individual in the identity space (40, 41, IS), select a target personal identifier (32) from all extracted data elements related to the individual in the identity space (40, 41, IS), and wherein the target personal identifier (32) uniquely identifies the individual in the identity space (40, 41, IS).

9. System according to one of the preceding claims, wherein: The identity definition module (23) is configured to create at least one further identity space on a different subset of the extracted data elements.

10. A system according to any one of the preceding claims, wherein: The risk mitigation module (25) is configured to output at least one proposed mitigation function and receive a selected mitigation function selected by the user to be applied to the identity space (40, 41, IS).

11. The system according to claim 10, wherein: The risk mitigation module (25) is configured to propose the at least one mitigation function based on the privacy risk calculated for the identity space (40, 41, IS).

12. The system according to one of claims 10-11, wherein: The risk mitigation module (25) is configured to propose the at least one mitigation function based on the semantic type of the data element to be converted in the identity space (40, 41, IS).

13. A computer-implemented method for providing a data privacy risk profile of a data privacy-related data object including data related to an individual to assign access rights for processing and storing said data object, comprising: - receiving (S1) more than one data object (DO) and extracting data from said data objects, - generating (S2) at least one identity space, wherein the identity space comprises those data objects containing extracted data elements about a set of individuals or a subset of the set of individuals, wherein the extracted data elements are those extracted data that can be associated with the individuals, and wherein the set of individuals is different from the set of individuals of every other identity space, - calculating (S3) for each identity space and for the contained individuals at least one privacy risk according to a probability of re-identification of each individual based on the extracted data elements in said identity space and according to a set of background knowledge assumed to be known, - exporting (S4) the extracted data and the calculated privacy risk, Additionally includes: - if the calculated privacy risk (PR) exceeds a predefined threshold (T) value, performing at least one mitigation function on the data of the identity space (40, 41, IS) and resulting in mitigated data (MD) that reduces the probability of re-identification of each individual, - providing the mitigated data (MD) to the privacy risk calculation module (25), the privacy risk calculation module (25) being configured to: calculating new privacy risks based on the mitigated data (MD) of the identity space (40, 41, IS); and The mitigated data (MD) and the new privacy risk are output via the export module (26).

14. A computer program product directly loadable into the internal memory of a digital computer, comprising software code portions for performing the steps of claim 13 when said product is run on said digital computer.

15. Use of the system according to claims 1-12, wherein: The system is applied to assign access rights categories to the extracted data in dependence on the calculated privacy risk required to access and / or process the extracted data.