Device and method for safe storage of media including personal data and erasing of stored personal data
Patent Information
- Application Number
- JP2022013847
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2021-02-09
- Filing Date
- 2022-02-01
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2042-02-01
AI Technical Summary
Existing technologies face challenges in securely storing media containing personal data while ensuring compliance with data protection regulations like GDPR, where the right to erasure (or 'right to be forgotten') complicates data management, often leading to unnecessary deletion of non-personal data.
A method and device that utilize temporary anonymous identifiers generated through one-way functions to store personal data securely, allowing selective erasure of personal data without affecting other data, using two memory systems for anonymized and non-anonymized data storage.
Enables secure and efficient storage and management of personal data, allowing selective erasure of personal data while preserving the integrity of non-personal data, and ensuring privacy by preventing unauthorized identification of individuals.
Abstract
Description
Technical Field
[0001] The present disclosure relates to technologies for securely storing and playing back media containing personal data and for reliably erasing such personal data upon request.
Background Art
[0002] The present invention addresses the storage of media containing mixtures of personal data and other data, and in particular seeks to propose technologies that meet the privacy safeguards set forth in legal documents such as Regulation (EU) 2016 / 679 (General Data Protection Regulation (GDPR)), the Chinese Cybersecurity Law, the California Consumer Privacy Act, and other US federal and state laws, regarding the protection of natural persons in the processing of personal data and the free movement of such data. Under the GDPR, each person shall have the right to require the complete erasure of their stored personal data at any time. This is referred to as the data subject's right to erasure, or the "right to be forgotten". In existing technologies, however, the right to be forgotten may not be compatible with data version control, which is an approximately essential tool for managing many datasets over the long term.
[0003] Certain difficulties arise in relation to the storage of non-discrete media such as text, images, and video data. In one scenario, a person (data subject) has previously agreed to a company (data recipient) store a video in which she appears in only a small portion of the frames, and then contacts the data recipient and requests that her personal data be erased. While the data recipient is, in principle, authorized to retain all other frames in which the person appears, storing a personal identifier for each video frame would present many problems from a data protection standpoint. This is because storing identified personal data, which carries a higher associated privacy risk, requires a technical configuration that ensures a higher level of security, which can become practically unmanageable. To satisfy this person's request for erasure, the data recipient therefore has the option of indiscriminately deleting the entire video sequence. This represents a serious and unjust loss of valuable data. [Overview of the project]
[0004] One purpose of this disclosure is to make available methods and devices for storing files containing personal data such that a data subject's request for the erasure of her personal data can be fulfilled without unnecessarily erasing other data contained in the file. Another purpose is to ensure that the file containing the personal data remains securely stored in a form that is useful for efficient management and processing until such a request for erasure is made. A specific security aspect addressed by this disclosure is to prevent unauthorized access to the stored personal data from identifying the data subject (person) to which it belongs. A specific efficiency aspect is to enable version control. Another purpose of this disclosure is to make available methods and devices for efficiently recovering files containing personal data that have been securely stored using the technology proposed herein.
[0005] These and other embodiments are realized by the present invention as defined in the independent claims. The dependent claims relate to preferred embodiments of the present invention.
[0006] A first aspect of the present invention provides a method for storing a file containing personal data. This method includes: obtaining a temporary anonymous identifier (AnonID.m) of a person, which depends on the token (FileID.m) of the file; extracting personal data items associated with the person from the file; generating, for each personal data item, a locator (Loc.mn) that enables the personal data item to be recovered into the file, and an item-specific anonymous identifier (AnonID.mn) of the person, wherein the item-specific anonymous identifier is generated by applying a predetermined one-way function to a combination of the temporary anonymous identifier (AnonID.m) and the identifier (n) of the personal data item; storing each personal data item, together with its locator and item-specific anonymous identifier, in a first memory; and storing an anonymous version of the file, excluding those personal data items, in a second memory.
[0007] Because item-specific anonymous identifiers are generated using a one-way function, unauthorized access to the first memory cannot easily make the stored personal data items attributes of that person. Furthermore, unauthorized individuals have no way of collecting all personal data items associated with the same person. This can be considered a form of anonymization of the file. Because the file is anonymized before being stored in the second memory, this memory may be subject to less stringent security requirements than the first memory. Because the person is identified by a temporary anonymous identifier, the recipient of the data can implement a method for storing the file without requiring the person to share their own unanonymous identifier. Finally, because each personal data item is stored with an item-specific anonymous identifier, comprehensive and highly accurate erasure of all items associated with a particular person is possible upon request.
[0008] In a second embodiment, a method is provided for erasing personal data associated with a person from a first memory that stores personal data items together with their corresponding locators (Loc.mn) and item-specific anonymous identifiers (AnonID.mn) of the person associated with those personal data items. This method includes obtaining the person's private identifier (PrivID), obtaining all the tokens (FileID.m) of the files from which those personal data items can be extracted, obtaining all the identifiers (n) of the personal data items associated with that person and that can be extracted from that file for each of those files, generating the person's temporary anonymous identifier (AnonID.m) by applying a predetermined one-way function to the combination of the obtained private identifier (PrivID) and the obtained file tokens (FileID.m), generating the person's item-specific anonymous identifier (AnonID.mn) for each of the generated temporary anonymous identifiers (AnonID.m) by applying a predetermined one-way function to the combination of the temporary anonymous identifier (AnonID.m) and the obtained identifiers (n) of those personal data items, and clearing all personal data items from the first memory that match any of the generated item-specific anonymous identifiers (AnonID.mn) of that person.
[0009] The fact that each personal data item is stored with an item-specific anonymous identifier makes it possible to comprehensively and accurately find and erase all items associated with a particular person. The erasure method according to the second aspect does not require the personal data items to modify or erase any anonymous versions of those files from which they are extracted. This allows the data recipient to freely decide how and to what extent the files are recovered after the erasure of that person's personal data items. The data recipient may decide to reconstruct and use a subset of the files after erasure, according to the type of data in question, the intended use case, and the appropriate level of granularity. To name a few examples, for instance, after erasing a person's face from an image, a cropped version of that image may be retained for further use. After erasing a person's face from a particular frame in a video sequence, all remaining video frames, or each of at least N remaining frames in a sequence of video subsequences, may be retained. After erasing a person's name from a database row, other rows in that database may be retained. After erasing a person's authentication information from a document, other sections of that document may be retained. This erasure method itself does not impose any significant technical limitations on recovering the file. Finally, to verify that all personal data associated with that person has been completely erased by performing this method, it is sufficient to prove that none of the generated item-specific anonymous identifiers match any of those personal data items that remain stored in the first memory.
[0010] A third aspect provides a method for recovering personal data from a file. This method includes: retrieving the personal data items extracted from the file and their corresponding locators (Loc.mn) from a first memory that stores the personal data items extracted from the file together with their corresponding locators (Loc.mn) and item-specific anonymous identifiers (AnonID.mn) of the person associated with those personal data items; retrieving an anonymous version of the file, excluding those personal data items, from a second memory; and recovering those personal data items to their anonymous version according to their corresponding locators.
[0011] These personal data items are stored together with their corresponding locators and item-specific anonymous identifiers, allowing them to be correctly recovered to an anonymous version of their files without compromising the privacy of the people associated with them. This recovery method can be performed without requiring the recipient of the data to share their own unanonymous identifier. Furthermore, because this recovery method uses an anonymous version of the file as input, it is guaranteed to work even if some or all of this personal data is erased from the first memory, for example, at the request of the associated person. More precisely, in that case, the method may be guaranteed to work and return a usable file version (consistent, readable, editable, etc.) from which the erased personal data items have no corrupting effect.
[0012] When used together, these three aspects of the present invention form a data storage ecosystem that intentionally maintains privacy.
[0013] Further provided is a device including processing circuits that are communicatively connected to the first and second memories and are arranged to perform the first, second, or third method. The present invention further relates to a computer, or more particularly, a computer program that includes instructions causing the device to perform the first, second, or third method. The computer program may be stored on or distributed on a data carrier. As used herein, “data carrier” may be a transient data carrier, such as a modulated electromagnetic wave or light wave, or a non-transient data carrier. Non-transient data carriers include volatile and non-volatile memories, such as magnetic, optical, or solid-state permanent and non-permanent storage media. Still within the scope of “data carrier,” such memories may be permanently installed or portable.
[0014] In this disclosure, “file” is used broadly and refers to any dataset that can be independently stored and / or processed in a file system, including in the context of a database, file archive, or other data storage. In order to avoid unnecessarily deleting data in response to a request for erasure, recipients of data will typically want to handle media containing personal data in the smallest possible chunks, i.e., typically on a file-by-file basis. The present invention satisfies this requirement to the extent that it enables not only files, but even subsets of files, to be recovered after a person’s personal data has been erased. Referring again to the examples listed earlier, a file may correspond to, for example, images, video sequences, databases, or documents.
[0015] This disclosure strives to use the terms “personal data,” “data subject” (i.e., a natural person), “data recipient” (i.e., the person to whom personal data is disclosed) and “anonymization” in a manner consistent with their meanings under the GDPR. As a result, a “personal data item” is an item of “personal data,” such as an area of an image containing a person’s face, a video frame showing the license plate of a vehicle owned by that person, a row in a database containing a person’s name, or a section of a document containing a person’s authentication information.
[0016] In general, all terms used in the claims shall be interpreted according to their ordinary meanings in the art unless explicitly defined herein. Any mention of “a / an / the element, apparatus, component, means, step, etc.” shall be interpreted, unless otherwise specified, as referring to at least one example of such element, apparatus, component, means, step, etc. The steps of any method disclosed herein do not need to be performed in the order disclosed unless otherwise specified.
[0017] The aspects and embodiments will be described below with reference to the attached drawings for illustrative purposes. [Brief explanation of the drawing]
[0018] [Figure 1] This is a flowchart showing how to save files containing personal data. [Figure 2] This is a flowchart showing how to erase personal data associated with a person. [Figure 3] This is a flowchart showing how to recover personal data from saved files. [Figure 4] This is a functional block diagram of a device for performing at least the method shown in Figure 1. [Figure 5] Each arrow represents a key derivation structure indicating a one-way mapping. [Modes for carrying out the invention]
[0019] Aspects of this disclosure will be described in more detail below with reference to the accompanying drawings illustrating specific embodiments of the invention. These embodiments, however, may be embodied in many different forms and should not be understood as limiting. Rather, these embodiments are provided for illustrative purposes only, so as to make this disclosure complete and comprehensive and to adequately convey the entire scope of aspects of the invention to those skilled in the art. Throughout this specification, similar reference numerals indicate similar components. Save files containing personal data
[0020] Figure 1 illustrates method 100 for storing a file containing personal data. Method 100 may be performed by or on behalf of a data recipient, a natural person or a legal entity. This person receives or generates the file and has further obtained consent from a certain person to store her personal data in that file. This consent may cover not only the personal data in the file being stored, but also any additional files. Method 100 is designed to address a scenario in which the person revokes her consent. This revocation may be equivalent to a request for the complete erasure of her personal data.
[0021] In an initial, optional step 110, a one-way function interface (see element 490 in Figure 4), configured with a predetermined one-way function, becomes available to a person, enabling her to generate a temporary anonymous identifier (AnonID.m) based on a combination of a private identifier (PrivID) and a token for the file to be stored (FileID.m). The one-way function interface is available as locally installed software or an online interface, allowing the person to generate the temporary anonymous identifier (AnonID.m) without sharing the private identifier (PrivID) with the data recipient. Sharing the private identifier (PrivID) may only be necessary if the person wishes her personal data to be erased, as described below with reference to Figure 2.
[0022] As a one-way function, implementing method 100 may use a hash function, in particular a cryptographic hash function that provides a level of security deemed appropriate given the confidentiality of the personal data being stored. Two examples of these are SHA-256 and SHA3-512. The one-way function is predetermined so that a temporary anonymous identifier (AnonID.m) can be recovered when a request for erasure is made (for example, it is reproducible).
[0023] In step 110, the one-way function interface made available to the person may further enable the person to generate a private identifier (PrivID) based on a personal identifier (ID). The personal identifier (ID) may be a civilian or official identifier such as an identification number, social security number, passport number, combination of name and birthday, etc. that the person can reconstruct without actively archiving it. If this advantage is not considered important, the person may provide any bit pattern as the private identifier (PrivID) to the one-way function interface as long as she can regenerate the bit pattern in relation to future requests for erasure and she is confident that the bit pattern is reasonably difficult for a third party to regenerate. The mapping from the personal identifier (ID) to the private identifier (PrivID) can be the same one-way function used to map the private identifier (PrivID) to a temporary anonymous identifier (AnonID.m). Alternatively, a different one-way function can be used. In either case, as long as the person does not securely store the private identifier (PrivID), she should ensure that she still has access to the same one-way function, or a plurality of one-way functions, and can still instruct the erasure of her personal data.
[0024] The token (FileID.m) of the file to be saved is arbitrary, as long as it can be reconstructed based on information about the file stored in the second memory. The token (FileID.m) may be an execution sequence number. To enable a retrospective list (for example, when erasure is requested), the token is preferably a discrete quantity. Exemplary tokens include the file creation date, media recording data, the file size (of the unnamed version of the file), the size bin to which the file size belongs (in MB, e.g., in a range such as [0, 10], [10, 20], [20, 30]), a fingerprint / digest of the file content, and a file name that identifies the file in the file system. To perform the erasure of personal data, the recipient of the data may retrospectively obtain a complete set of values for these tokens by querying the file system for all files stored in the second memory. When the file creation date is used as a token, the complete set of values represents the progression of file storage, which corresponds, for example, to the range of data that can be read from the work log in digital or non-digital format. While not essential to the present invention, using a unique token for each file (i.e., an identifier) has some merit because it ensures there are no inconsistencies between item-specific anonymous identifiers (AnonID.mn). Alternatively, the same token (FileID.m) may be used for a cluster of multiple files stored simultaneously or at close time, avoiding the need to store sensitive, temporary anonymous identifiers (AnonID.m). However, it is not prudent to use a constant token for all files because it implies that all item-specific anonymous identifiers belong to a common sequence, making it impossible to efficiently document the total amount of stored personal data items. Such documentation would be a great help when complete erasure is required.
[0025] If the token of the file to be saved (FileID.m) is unknown to that person, different from, for example, today's date, etc., that value may be shared with that person. Or, this may be directly supplied to the unidirectional function interface in the second optional step 112. A temporary anonymous identifier (AnonID.m) can then be correctly generated based on the combination of the private identifier (PrivID) and the token (FileID.m). Method 100 may or may not include step 110 or step 112, or a combination of those steps.
[0026] In the next step 114 of method 100, that person's temporary anonymous identifier (AnonID.m) is obtained. The temporary anonymous identifier (AnonID.m) changes together with the token of the file (FileID.m). A preferred way to obtain the temporary anonymous identifier (AnonID.m) is to receive it from that person. In such a way, that person does not need to share a private or personal identifier with the recipient of the data that executes method 100.
[0027] In step 116, personal data items associated with that person are extracted.
[0028] In step 118, for each of the personal data items extracted in step 116, a locator (Loc.m.n) and that person's item-specific anonymous identifier (AnonID.m.n) are generated.
[0029] A locator (Loc.mn) enables the recovery of personal data items into a file. Its structure may depend on the type of media the file is associated with. If the personal data item is an area in an image, the locator (Loc.mn) may indicate that area in the image, for example, in units of the coordinates of its bounding box. For video data, a locator that enables the recovery of an extracted area corresponding to a person's face or their car's license plate in a frame may indicate the sequence number of that frame and the image coordinates of the extracted area.
[0030] The person's item-specific anonymous identifier (AnonID.mn) is generated in step 118 by applying a predetermined one-way function to the combination of the temporary anonymous identifier (AnonID.m) and the personal data item identifier (n). The one-way function may be the same one used by the person to generate the temporary anonymous identifier (AnonID.m), or it may be a different one-way function with similar characteristics. The personal data item identifier (n) may be a file-specific sequence number. That is, a sequence counter is reset for each new file that is saved, or, in some cases, for each cluster of files saved simultaneously, so that each identifier is unique within that file (file cluster). However, personal data items belonging to two different files (file clusters) may have the same identifier. The identifier (n) may also be a global sequence number, or it may be the complete identifier of the personal data item at the operating system level, or at another relevant namespace level. Instead of numerical counters, some may use alphabetical counters, or counters that take values in any suitable discrete set or space. The counters follow a predetermined list (or increment) rule. You may use TIFF2022122266000002.tif9170.
[0031] In a further development of method 100, where successive versions of a file can be saved, step 118 is repeated when a new version is saved. This causes the locators of those personal data items to further depend on the version (or commit) v of the file, i.e., Loc.mnv. This supports the implementation of version control in a second memory.
[0032] Refer to Figure 5. This shows how the identifier ID, PrivID, AnonID.m, and AnonID.mn are related. The downward arrows each correspond to the application of a one-way function. The letters on the left side of Figure 5 further indicate which entities have access to the corresponding identifier. A: The private identifier (ID) belongs to the data subject. Sharing it with the data recipient is not necessary. B: The data subject will retain a Private Identifier (PrivID) until consent to the storage of personal data is revoked. At that time, the data subject will share the Private Identifier (PrivID) with the data recipient so that the data recipient may erase her personal data in accordance with Method 200. C: When a data subject shares one of the temporary anonymous identifiers (AnonID.m) with the data recipient, it is equivalent to consenting to the storage of that file along with her personal data. The data recipient uses this identifier and carries out the storage of that file in accordance with Method 100. D: The data recipient stores an item-specific anonymous identifier (AnonID.mn) in the first memory along with the extracted personal data items. These identifiers cannot be erased until the personal data items are erased, so that the data recipient can carry out the requested erasure of those personal data items in accordance with Method 200.
[0033] Returning to Figure 1, Method 100 further includes Step 120, where each personal data item is stored in a first memory location along with a locator (Loc.mn or Loc.mnv) and an item-specific anonymous identifier (AnonID.mn) (see Element 421 in Figure 4). A relational database may be used for this storage operation, which may be visualized as a table with the following general appearance: TIFF2022122266000003.tif36170
[0034] In any step 122, a counter entry for that person is stored, for example, in a first memory or elsewhere. The counter entry allows for verification of the comprehensiveness of the list of all personal data items associated with that person. For this purpose, the counter entry may show: - The total number of personal data items. In this case, the deletion method will apply pre-agreed list rules and start rules. - The first and last personal data items (for example, by their unique anonymous identifier, AnonID.mn, or by their locator, Loc.mn). In this case, the erasure method is to apply the list rule. - A complete list of personal data items. In this case, the erasure method can be independent. Or - A list of all locations for personal data items or their locations. In this case, the erasure method can be independent. Preferably, one counter entry is stored for each file containing the person's personal data, or for each new file token (FileID.m) used.
[0035] In some embodiments, the counter entry includes a identifiable anonymous identifier (RecAnonID.m) for that person. This is generated by applying a one-way function to a temporary anonymous identifier (AnonID.m), as also indicated by the dotted arrow in Figure 5. Because the identifiable anonymous identifier (RecAnonID.m) is generated using a one-way function, it cannot be an attribute of that person. However, the identifiable anonymous identifier (RecAnonID.m) allows verification that the counter entry for the correct person has been found in connection with performing an erasure request. When the counter entry includes a identifiable anonymous identifier (RecAnonID.m), it may have the following structure: TIFF2022122266000004.tif36170 This can be understood as follows: One or more files using token FileID.1 strictly contain N1 personal data items associated with that person. One or more files using token FileID.2 strictly contain N2 personal data items associated with that person. And so on.
[0036] In any further optional step 124 of Method 100, an action is taken to prevent the non-volatile storage of the temporary anonymous identifier (AnonID.m). Such protective measures may include erasing or overwriting the identifier in runtime memory. Alternatively, a gatekeeper function is provided that refuses to store data having the data type of temporary anonymous identifier (AnonID.m) in the first and second memories and any further available non-volatile memory. Thus, if a new file is saved at a later date, this period is assumed to be too long to retain the identifier in runtime memory, but a new temporary anonymous identifier (AnonID.(m+1)) needs to be obtained, i.e., step 114 needs to be performed again.
[0037] At this point, step 126 assesses whether the file to be saved contains personal data associated with any further person. In this case (branch Y), execution returns to its loop and performs steps 114, 116, 118, and 120 (along with any steps 110 and 112, if any) for that further person. As a result, a temporary anonymous identifier (AnonID'.m) for that further person is obtained. Personal data items associated with that further person are extracted. A locator (Loc.mn) and an item-specific anonymous identifier (AnonID'.mn) for that further person are generated for each personal data item. Note that both locators for these people may depend on an identifier (n) belonging to the same sequence.
[0038] If there are no further people to consider regarding the file to be saved (branch N from step 126), the execution of method 100 proceeds to an optional step 128, where personal data items in the file are removed and an unnamed (or censored) version of the file is obtained. This removal may be achieved by at least one of deleting (erasing) the personal data items, discarding any potential identifiable factors in the personal data items, or hiding, editing, masking, replacing, or overwriting the personal data items, or by further applying appropriate filtering or image processing. Appropriate filtering and image processing include blurring, pixelating, distorting, or other effects that tend to depict unidentifiable personal data. Such filtering or image processing is typically of the data destruction type and will reduce the informational content of the file.
[0039] In the subsequent step 130, an unnamed version of the file, which does not contain those personal data items, is stored in a second memory (see element 422 in Figure 4). The second memory may be the same memory as the first memory. However, in some embodiments, the second memory is different from the first memory.
[0040] Within step 130, all locators (Loc.mn) of the extracted personal data items may optionally be saved along with an unnamed version of the file. In other words, this option means that the locators (Loc.mn) are saved in both the first and second memory locations. Access to spare copies of the locators (Loc.mn) is useful when the file is reconstructed in situations where some of the personal data it originally contained has been erased, and this erasure was performed by wiping all rows in Table 1 associated with the requesting person, including the locators. The spare copies of the locators (Loc.mn) then allow the corresponding locations to be marked / modified / invalidated as empty despite the erasure. Alternatively, the same marking functionality can be performed if the locators are retained even after the erasure operation (i.e., the first column of Table 1 remains complete).
[0041] This completes the saving of the first version of the file.
[0042] Optionally, method 100 may be extended to store a second file which also includes personal data associated with the person. The second file may be related to the first one by belonging to the same temporary, spatial, or subject context. For example, these files may represent images acquired at consecutive points in time, or consecutive video sequences. These files may further be related to different sub-areas of the imaged space that are partitioned to acquire a manageable file size, granularity, or similar.
[0043] As a result, in step 132, it is assessed whether the second file is saved or not. In this case (branch Y), execution returns to the loop and starts in step 114. In this execution round, the person's second temporary anonymous identifier (AnonID.(m+1)) is obtained. This temporary anonymous identifier (AnonID.(m+1)) depends on the token (FileID.(m+1)) of the second file. Personal data items associated with the person are extracted from the second file. For each extracted personal data item, a locator (Loc.(m+1).n) and the person's item-specific anonymous identifier (AnonID.(m+1).n) are generated. Each personal data item, along with its locator and item-specific anonymous identifier, is stored in the first memory. Finally, an anonymous version of the second file, without those personal data items, is stored in the second memory.
[0044] If there are no more files to save (branch N from step 132), the execution of method 100 can be terminated.
[0045] In an optional additional step 134 of Method 100, an updated anonymous version of the file is stored in a second memory. The updated anonymous version may be stored under version control; that is, the previous anonymous version is not erased from the second memory. The storage of the updated anonymous version may be done without necessarily updating the information in the first memory, which remains valid. This fact preferably allows for limiting the circle of staff entrusted with processing personal data.
[0046] Optionally, step 134 affects data in both the first and second memories. Assuming it is desirable to increase the brightness in the video sequence, this image processing operation may be applied to all frames of the video sequence and all cropped images (i.e., extracted personal data items) in the first memory. The unnamed video sequence is saved as a new version in the version-controlled second memory. A new locator (Loc.mn2) is generated for the new version (v=2), saved in a new row in the data structure of the first memory. Each new row contains the modified (brightened), cropped image, but AnonID.mn is the same as for the first version. This is possible without compromising the protection of the person's privacy. Delete personal data associated with a person.
[0047] Figure 2 shows a flowchart of method 200 for erasing personal data associated with a person from a first memory. Method 200 begins when the memory stores personal data items along with their corresponding locators (Loc.mn) and item-specific anonymous identifiers (AnonID.mn) of the person associated with those personal data items. This content may be loaded into memory as a result of performing method 100 for saving a file containing the personal data described above.
[0048] Method 200 may be performed, for example, by or on behalf of a data recipient, a natural person or legal entity, who has previously received or generated one or more files and whose consent has been obtained to store her personal data contained in those files. Method 200 for erasing personal data associated with that person may be performed when that person requests to be “forgotten” (her personal data erased), or equivalently when that person revokes the consent previously given.
[0049] In the first step 210 of Method 200, the person's private identifier (PrivID) is obtained. The recipient of the data does not receive any private identifiers (PrivIDs) from the person while performing the normal execution of Method 100 for saving the file, and sharing them in this way does not improve the person's privacy. Therefore, step 210 typically involves either receiving the private identifier (PrivID) from the person or retrieving the private identifier (PrivID) from shared memory to which the person has granted the recipient of the data read access.
[0050] In the second step 212, all tokens (FileID.m) of the files from which those personal data items can be extracted are obtained. In this step 212, the data recipient may reconstruct the tokens (FileID.m) in accordance with the rules and / or documents regarding the storage of one or more files that contained the personal data. Alternatively, the tokens (FileID.m) may be received or fetched from memory.
[0051] In principle, a person may maintain their own registration record of the token (FileID.m) that they share with the data recipient in connection with requests for erasure. This would allow them to limit, for example, their requests for partial erasure of personal data contained in files recorded on a particular day. However, if the data recipient is a professional entity, a more convenient option may be to store for the data recipient sufficient information to enable it to comprehensively reproduce the token (FileID.m) as is, without any other input from that person other than the private identifier (PrivID). The storage of this information should be at least reliable and / or long-lasting, given that if that information is lost, there would be no convenient way to perform knowledge-based selective erasure of private data associated with a particular person. A more compelling option that can be put into practice if the token (FileID.m) is based on the file name, original file size, creation date, or other file attributes is to query the data recipient for a second memory that stores anonymous versions of those files from which those personal data items are extracted. Next, in substep 212.1, this query may include an ls or dir command issued to the filesystem of the second memory, which will succeed as long as the second memory stores the relevant file. The token resulting from the output of the query will be matched in the following step 224 against the generated item-specific anonymous identifier (AnonID.mn), so whether the query to the second memory returns additional file attributes unrelated to that person as a byproduct is a very minor concern.
[0052] In the next step 214, for each of those file tokens (FileID.m), all identifiers (n) of those personal data items associated with the person and that can be extracted from the corresponding one or more files are obtained. Obtaining those identifiers (n) may involve retrieving them from memory or other entities, or may be based on a retrieval process. The retrieval process may be controlled by a private identifier (PrivID) received from the person and may further depend on information stored in memory held by the data recipient.
[0053] For example, the identifier (n) of those personal data items to be deleted may be obtained by performing substep 214.1, in which a counter entry for that person is read. The concept of a counter entry was previously introduced in connection with step 122 of Method 100 for storing with an exemplary data structure that optionally includes a recognizable anonymous identifier (RecAnonID.m). To find the correct counter entry for a file from among several stored counter entries, the data recipient first applies a one-way function to the combination of a private identifier (PrivID) and the token for the file (FileID.m), which returns a temporary anonymous identifier (AnonID.m), and then reproduces it by applying the one-way function to the temporary anonymous identifier (AnonID.m). The data recipient matches the output of this operation with the relevant fields (columns) of several stored counter entries. Another field in the matching counter entry will represent the total number of personal data items, the first and last personal data items in a sequence determined by a predetermined listing rule, the entire list of personal data items, or any other option listed under step 122. As a result, one counter entry is expected for each temporary anonymous identifier (AnonID.m), so substep 214.1 returns one set of identifiers (n) for each temporary anonymous identifier (AnonID.m).
[0054] To demonstrate this, the counter entry is the identifier of the first personal data item that was saved. TIFF2022122266000005.tif8170 and the identifier of the last personal data item It is assumed that this will refer to TIFF2022122266000006.tif8170. Here, "first" and "last" refer to a sequence of identifiers and do not necessarily refer to the time when those personal data items were saved. In this case, the erasure method 200 is the same as or equivalent to the corresponding list rule used in the storage method 100. Apply TIFF2022122266000007.tif9170. This list rule applies to the identifier of the first personal data item. TIFF2022122266000008.tif9170, followed by the identifier of the last personal data item. The recipient of the data may regenerate all identifiers by recursively applying them to successive outputs until TIFF2022122266000009.tif8170 is reached. Identifier of the last personal data item stored in the counter entry. TIFF2022122266000010.tif8170 is used to verify the inclusivity of the listed identifiers. In variations of this regeneration process, the identifier of the first personal data item TIFF2022122266000011.tif8170 may be pre-agreed or pre-specified. Alternatively, a counter entry indicating the total number of identifiers may be the identifier of the last saved personal data item. It will be useful to learn about TIFF2022122266000012.tif9170.
[0055] In a further step 216, the person's temporary anonymous identifier (AnonID.m) is generated by applying a predetermined one-way function to a combination of the obtained private identifier (PrivID) and the obtained file token (FileID.m). The one-way function is predetermined in the sense that its action is equivalent to the one-way function used to generate the temporary anonymous identifier (AnonID.m) when one or more files containing personal data are saved.
[0056] Those skilled in the art will understand that steps 214 and 216 can be performed in either order or in parallel. Since the temporary anonymous identifier (AnonID.m) generated in step 216 may be used as input in substep 214.1, parallel execution can reduce the total number of times the one-way function must be evaluated.
[0057] Next comes step 218, which generates an item-specific anonymous identifier (AnonID.mn) for that person. This proceeds by applying a predetermined one-way function to each of the generated temporary anonymous identifiers (AnonID.m) for that person, in combination with the temporary anonymous identifier (AnonID.m) and the retrieved identifier (n) of those personal data items. To ensure complete erasure, step 218 should generate a complete collection of the item-specific anonymous identifiers (AnonID.mn) used to store the personal data items associated with that person. If the identifier (n) was obtained using substep 214.1, there is one set of identifiers (n) for each temporary anonymous identifier (AnonID.m). This should then be sufficient to combine the identifier (n) in one of those sets with the corresponding temporary anonymous identifier (AnonID.m). On the other hand, combining it with a different temporary anonymous identifier ('AnonID.m') is unlikely to provide a further item-specific anonymous identifier that matches any of the stored personal data items.
[0058] In step 220, it is assessed whether any further generated temporary anonymous identifiers (AnonID.m) of the person are to be processed in step 218. If not (branch N), in step 222, it is assessed whether any further file tokens (FileID.m) used to store personal data items associated with the person exist. If it is found that such further file tokens (FileID.m) exist (branch Y), steps 214 onward are repeated for each of the further file tokens (FileID.m).
[0059] Once all file tokens (FileID.m) have been processed, method 200 proceeds to step 224, where all personal data items matching any of the generated item-specific anonymous identifiers (AnonID.mn) for that person are erased from the first memory. This erasure may target only the personal data items, or the corresponding locators and / or item-specific anonymous identifiers may be erased together. The latter option corresponds to completely erasing rows of the data structure shown in Table 1 above. If it is desirable to find the part of the file from which those personal data items are extracted (for example, for marking purposes), their locators should be left intact or recovered by some other means unless copies of those locators are stored somewhere. Recover personal data from saved files.
[0060] Figure 3 is a flowchart of method 300 for recovering personal data to a file. Method 300 begins when a first memory stores personal data items along with their corresponding locators (Loc.mn) and item-specific anonymous identifiers (AnonID.mn) for the people associated with those personal data items, and a second memory stores an anonymous version of the file. These memories may receive this content as a result of executing method 100 for saving the file containing the personal data described above. From the time the file is saved, method 200 for erasing personal data associated with people may be executed. In this case, the first memory already contains an incomplete collection of those personal data items extracted from the file when saving method 100 was executed.
[0061] In the first step 310 of method 300, the personal data items extracted from the file and their corresponding locators (Loc.mn) are recovered from the first memory. If the locators (Loc.mn) are also stored in locations other than the first memory, such as a second memory, they may be recovered equally from there.
[0062] In the second step 312, an unnamed version of the file, which does not contain those personal data items, is retrieved from the second memory.
[0063] Those personal data items are subsequently recovered in step 314 to an unnamed version of the file according to the corresponding locator (Loc.mn). For example, a cropped area of a video frame corresponding to a person's face may be pasted back into the video frame as shown in the locator (Loc.mn). This restores the video frame to its original state and appearance. The restored video frame may differ from its original state if later versions of the file were saved as a result of, for example, image processing or video editing. Step 314 may inevitably involve completely restoring the file, that is, recovering all personal data items regardless of the person they are associated with.
[0064] Substep 314.1 represents a preferred method for performing step 314, where all recovered personal data items are sequentially examined in detail. Step 314 terminates after the last item. Substep 314.1 is ensured to be performed even if some or all of the personal data has been erased from the first memory, for example, at the request of the associated person. It can be configured to return usable files (uncorrupted, readable, editable, etc.) along with some remaining parts corresponding to those erased personal data items. The remaining parts may include any of the outputs applied in the execution of storage method 100, such as deletion, hiding, editing, masking, replacement, overwriting, filtering, and image processing.
[0065] In any step 316, an unnamed version of a personal data item is identified in which such a portion has been extracted but not recovered.
[0066] In a further optional step 318, the region identified in step 316 is notified to a downstream utility processing step. The identified region may be understood as invalid data or artificially modified data, as opposed to natural data or surrogate data. For example, if the downstream processing involves training a machine learning (ML) model, the identified region may be excluded from the training data sent to the ML model. Alternatively, any updates to the ML model resulting from the identified region (e.g., represented to a neural network by updated weights, or calculated slopes, or a set of errors that occurred) may be erased, neutralized, or reverted to previous values.
[0067] The file containing the recovered personal data items is held in the runtime memory of the processor performing method 300. To enhance the privacy of the data subject, method 300 may optionally include a final step 320 that takes action to prevent the non-volatile storage of the file containing those recovered personal data items. Such action may include erasing the file from runtime memory or overwriting it after it has been processed. Alternatively, gatekeeper functions are provided in the first and second memories and any further available non-volatile memory to prevent the storage of the file. Further alternatively, cleanup based on dependency tracking techniques may be performed. Device implementation
[0068] Figure 4 shows a device 410 that includes processing circuits 411, which are communicatively connected to the first and second memories 421, 422 and are arranged to perform a save method 100, an erase method 200, and / or a recovery method 300. The processing circuits 411 may include application-specific or programmable circuits, or they may be distributed and executed using networked ("cloud") resources. The processing circuits 411 may include volatile runtime memory. Alternatively, the memory may be non-volatile and may be provided with a gatekeeper or covered by dependency tracking. The first memory 421 may be used to store extracted personal data items, locators, and item-specific anonymous identifiers extracted from a file. Since the first memory 421 is used for relatively sensitive content, it should preferably have high resistance to intrusion attacks. Similarly, for example, communication connections to the first memory 421 over a network should be protected against eavesdropping. The second memory 422 may be used to store one (or more) unnamed versions of the file. Both memories 421 and 422 may be non-volatile memories. The second memory 422 may be subject to version control. The first memory 421 does not need to be subject to version control.
[0069] Figure 4 is annotated with data labels showing how storage method 100 can be performed. The file to be stored is supplied on the upper left side of device 410. A temporary anonymous identifier (AnonID.m) is supplied on the lower left side. The temporary anonymous identifier (AnonID.m) may be generated by or for the person associated with the private data using the one-way function interface 490. As shown here, the one-way function interface 490 receives the person's private identifier (PrivID) as input. This is further modified according to the file token (FileID.m) supplied by device 410. The person may generate the private identifier (PrivID) based on a personal identifier (ID) using the one-way function interface 490 (shown in Figure 4 as a further example). Based on these inputs, device 410 extracts personal data items (symbolized as rectangles and circles in Figure 4) from the supplied file and stores them, along with their corresponding locators (Loc.mn) and item-specific anonymous identifiers (AnonID.mn), in a first memory 421. The first memory may then contain triplets of personal data items, locators, and identifiers. In parallel, device 410 stores an anonymous version of the file in a second memory 422.
[0070] When device 410 performs the erase method 200, it may receive the requesting person's private identifier (PrivID). It then issues an erase command to the first memory 421. When device 410 performs the recovery method 300, it recovers the data from the first memory 421 and the second memory 422 and outputs the recovered file. Numbered Embodiments
[0071] Embodiment 1. A method (100) for saving a file containing personal data, Obtaining a temporary anonymous identifier (AnonID.m) for a person, which is dependent on the token (FileID.m) of that file (114), Extracting personal data items associated with that person from that file (116), For each personal data item, generate a locator (Loc.mn) that enables the personal data item to be recovered in its file, and an item-specific anonymous identifier (AnonID.mn) for that person (118), wherein the item-specific anonymous identifier is generated by applying a predetermined one-way function to a combination of the temporary anonymous identifier (AnonID.m) and the identifier (n) of the personal data item, thereby generating the locator identifier. Each personal data item is stored in a first memory along with its locator and an anonymous identifier unique to that item (120), To store an unnamed version of the file that does not contain those personal data items in a second memory (130), A method that includes this.
[0072] Embodiment 2. A method of Embodiment 1 for further storing a second file containing personal data associated with that person, Obtaining a second temporary anonymous identifier (AnonID.(m+1)) for that person, which depends on the token (FileID.(m+1)) of the second file (114), Extracting personal data items associated with that person from the second file (116), For each personal data item, the process involves generating a locator (Loc.(m+1).n) and a person's item-specific anonymous identifier (AnonID.(m+1).n) (118), wherein the item-specific anonymous identifier is generated by applying a one-way function to a combination of a second temporary anonymous identifier (AnonID.(m+1)) and the identifier of the personal data item (n), thereby generating a locator and identifier. Each personal data item is stored in a first memory along with its locator and an anonymous identifier unique to that item (120), (130) storing an unnamed version of the second file, which does not contain those personal data items, in the second memory, A method that includes this.
[0073] Embodiment 3. To obtain a temporary anonymous identifier (AnonID'.m) for a further person, which is a temporary anonymous identifier (AnonID'.m) that depends on the token (FileID.m) of that file (114), Extracting personal data items associated with that further person from that file (116), For each personal data item, the method involves generating a locator (Loc.mn) and a further person's item-specific anonymous identifier (AnonID'.mn) (118), wherein the item-specific anonymous identifier is generated by applying a one-way function to a combination of its temporary anonymous identifier (AnonID'.m) and the personal data item identifier (n), thereby generating a locator and identifier. Each personal data item is stored in a first memory along with its locator and an anonymous identifier unique to that item (120), It further includes, Here, an unnamed version of the file is stored in a second memory location, excluding the personal data items associated with that person and the personal data items associated with further individuals. The method of Embodiment 1 or Embodiment 2.
[0074] Embodiment 4. One of the above embodiments, wherein the second memory, and not the first memory, is subject to version control.
[0075] Embodiment 5. The first and second memories are non-volatile, according to one of the preceding embodiments.
[0076] Embodiment 6. At least one of the personal data items is an area of the image. The locator (Loc.mn) indicates that area within the image. One of the methods of the previous embodiments.
[0077] Embodiment 7. Methods (300) for recovering personal data into a file, Retrieving the personal data items extracted from the file, along with their corresponding locators (Loc.mn) and the item-specific anonymous identifiers (AnonID.mn) of the person associated with those personal data items (310), from a first memory location where those personal data items are stored together. From the second memory, retrieve an unnamed version of the file that does not contain those personal data items (312), To restore those personal data items to an unnamed version according to their corresponding locators (314), A method that includes this.
[0078] Embodiment 8. The recovery method of Embodiment 7 includes sequentially recovering the recovered personal data items (314.1) and terminating after the last item.
[0079] Embodiment 9. Identifying such parts of an unnamed version of a personal data item that has been extracted but not recovered (316), To notify the downstream utility processing step of those identified parts (318), The method of Embodiment 7 or Embodiment 8, further comprising:
[0080] Embodiment 10. Any method of Embodiments 7 to 9 further includes taking action to prevent the non-volatile storage of the file with those recovered personal data items (320).
[0081] Aspects of this disclosure have been described above, primarily with reference to several embodiments. However, as will be immediately apparent to those skilled in the art, other embodiments different from those disclosed above are equally possible within the scope of the invention, as defined in the claims.
Claims
1. A method (100) for storing files containing personal data, comprising: Obtaining (114) a temporary anonymous identifier (AnonID.m) of a person, the temporary anonymous identifier (AnonID.m) depending on the token (FileID.m) of the file; Extracting (116) personal data items associated with the person from the file; generating (118) for each of said personal data items a locator (Loc.m.n) that allows said personal data item to be restored to said file, and an item-specific anonymous identifier (AnonID.m.n) of said person, said item-specific anonymous identifier being generated by applying a predetermined one-way function to a combination of said temporary anonymous identifier (AnonID.m) and said personal data item identifier (n); storing (120) each of said personal data items in a first memory together with said locator and an anonymous identifier unique to said item; storing (130) an anonymized version of the file without the personal data item in a second memory; A method comprising:
2. 2. The method of claim 1, further comprising: storing (122) a counter entry for the person, the counter entry enabling verification of comprehensiveness of a list of all of the personal data items associated with the person.
3. 3. The method of claim 2, wherein the counter entry comprises a recognizable anonymous identifier (RecAnonID.m) of the person generated by applying the one-way function to the temporary anonymous identifier (AnonID.m).
4. carried out by the recipient of the data, making available (110) a one-way function interface configured with the predetermined one-way function to allow a person to generate the temporary anonymous identifier (AnonID.m) based on a combination of a private identifier (PrivID) and the token (FileID.m) of the file without sharing the private identifier with a recipient of the data; Optionally sharing (112) the token (FileID.m) of the file with the person or the one-way function interface; The method of claim 1 , further comprising:
5. The method of claim 4 , wherein the one-way function interface is further configured to enable the person to generate the private identifier (PrivID) based on a personal identifier (ID).
6. The method of claim 1 , further comprising taking an action to prevent non-volatile storage of the temporary anonymous identifier (AnonID.m) (124).
7. 7. The method of claim 1, further comprising removing (128) the personal data items in the file by anonymization operations including at least one of deleting, hiding, editing, masking, replacing, overwriting, filtering, and imaging to obtain the anonymized version of the file.
8. The method of claim 1 , further comprising: saving (134) an updated, anonymized version of the file in the second memory.
9. at least one of said personal data items is an area of a frame in a video sequence; said locator (Loc.m.n) indicating said frame in said video sequence and further indicating said area in said frame; 9. The method according to any one of claims 1 to 8.
10. A method (200) for erasing personal data associated with a person from a first memory in which personal data items have been previously stored in a manner as claimed in claim 1 together with a corresponding locator (Loc.m.n) and an item-specific anonymous identifier (AnonID.m.n) of the person associated with said personal data items, comprising: Obtaining (210) a private identifier (PrivID) of the person; Obtaining (212) tokens (FileID.m) of all files from which the personal data item may be extracted; obtaining (214) for each one of the file tokens (FileID.m) identifiers (n) of all the personal data items associated with the person and that may be extracted from the corresponding file; generating (216) a temporary anonymous identifier (AnonID.m) for the person by applying a predetermined one-way function to a combination of the obtained private identifier (PrivID) and the obtained file token (FileID.m); for each generated temporary anonymous identifier (AnonID.m) of the person, generating (218) an item-specific anonymous identifier (AnonID.m.n) of the person by applying the predetermined one-way function to a combination of the temporary anonymous identifier (AnonID.m) and the obtained identifier (n) of the personal data item; erasing (224) all of the personal data items from the first memory that match any of the person's generated item-specific anonymous identifiers (AnonID.m.n); A method comprising:
11. 11. The method according to claim 10, wherein the tokens (FileID.m) of all files from which the personal data items may be extracted are obtained by sending a query (212.1) to a second memory in which anonymized versions of the files from which the personal data items were extracted have been previously stored according to the method of claim 1.
12. 12. The method according to claim 10 or 11, wherein the identifiers (n) of all the personal data items are obtained by retrieving (214.1) counter entries for the person.
13. listing the identifiers (n) of all the personal data items associated with the person; verifying the comprehensiveness of the listed identifiers based on the retrieved counter entries; The method of claim 12 further comprising:
14. A device (410) comprising a processing circuit (411) communicatively connected to first and second memories (421, 422) and arranged to perform the method of any one of claims 1 to 13.
15. A computer program comprising instructions which, when said program is executed by a computer, cause said computer to carry out the method according to any one of claims 1 to 13.
