Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 02/26/2026 and 03/21/2025 has been considered by the examiner.
Response to Arguments
Based on the applicant’s amendment to claims 1, 12 and 17 along Applicant’s arguments, see pages 8-10, filed on 07/16/2026, with respect to claims 1-20 have been fully considered and are persuasive. The rejection of claims 1-20 under 35 U.S.C. 101 has been withdrawn.
Applicant’s arguments, see pages 11-14, filed on 07/16/2026, with respect to the rejection(s) of claim(s) 1-20 under 35 U.S.C. 103 as being unpatentable over Peng et al. (US 20250156300 A1—hereinafter—"Peng”) in view of Cook et al. (US 20240005022 A1—hereinafter—" Cook”) have been fully considered and are persuasive. Therefore, the rejection has been withdrawn. However, upon further consideration, a new ground(s) of rejection is made in view of in further view Cesar et al. (US 20260003996 A1 –hereinafter—"Cesar”).
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-20 are rejected under 35 U.S.C. 103 as being unpatentable over Peng et al. (US 20250156300 A1—hereinafter—"Peng”) in view of Cook et al. (US 20240005022 A1—hereinafter—" Cook”) in further view Cesar et al. (US 20260003996 A1 –hereinafter—"Cesar”).
As per claim 1. Peng discloses a method comprising:
receiving, by a processing device, a dataset record including confidential data ([0212] Model alignment toolkit 17 can integrate one or more dataset(s) 17-1 for aligning development model 16. Curated dataset(s) 17-1 can include labeled or unlabeled training data. Dataset(s) 17-1 can be obtained from public domain datasets. Dataset(s) 17-1 can be obtained from private datasets associated with one or more developer system(s) for the alignment of bespoke machine-learned model(s) customized for private use-cases. [0213] Pre-training pipelines 17-2 can include a machine-learned model training workflow configured to update development model 16 over large-scale, potentially noisy datasets. For example, pre-training can leverage unsupervised learning techniques (e.g., de-noising, etc.) to process large numbers of training instances to update model parameters from an initialized state and achieve a desired baseline performance. Pre-training pipelines 17-2 can leverage unlabeled datasets in dataset(s) 17-1 to perform pre-training. Workbench 15 can implement a pre-training pipeline 17-2 to pre-train development model 16. [0214] Fine-tuning pipelines 17-3 can include a machine-learned model training workflow configured to refine the model parameters of development model 16 with higher-quality data. Fine-tuning pipelines 17-3 can update development model 16 by conducting supervised training with labeled dataset(s) in dataset(s) 17-1. Fine-tuning pipelines 17-3 can update development model 16 by conducting reinforcement learning using reward signals from user feedback signals. Workbench 15 can implement a fine-tuning pipeline 17-3 to fine-tune development model 16);
generating, by the processing device, a sketch as a probabilistic data structure using differential privacy ([0052] A prediction system (e.g., a content provider system, a tag value server, etc.) can generate a sketch that compiles predicted characteristics for the served content events. The sketch can be indexed by, for instance, a tag value. The sketch can be noised or otherwise obscured. The noised sketch can satisfy a differential privacy criterion. The noised sketch can be transmitted to the reference data system. [0068] Prediction system 102 can implement a sketch pipeline 106 to generate a sketch of one or more event data objects and the one or more predicted values corresponding thereto);
the generating including applying noise to the dataset record ([0069] Prediction system 102 can implement a noising pipeline 108 to add noise to sketches generated by sketch pipeline 106. The noised sketches 109 can be transmitted to a reference system 110. Reference system 110 can store reference data 112. Reference data 112 can include event data objects with corresponding reference values associated therewith. Reference data 112 can include a subset of event data objects that are in log data 104. Reference system 110 can implement a confusion matrix estimation pipeline 114 that estimates a confusion matrix for a noised sketch generated by noising pipeline 108 based on reference data 112. Confusion matrix estimation pipeline 114 can generate confusion matrix data 116); and.
communicating, by the processing device, the sketch to be stored in a database ([0169] In some implementations of example method 300, the reference system sums values of the noised sketch stored in the target positions. For instance, confusion matrix estimation pipeline 114 can select these positions of v.sup.d and sum the values stored in those positions. For instance, confusion matrix estimation pipeline 114 can obtain the following sum: [00010].Math.i∈Igvd[i]. [0170] In some implementations of example method 300, the sum of the values of the noised sketch stored in the target positions corresponds to a quantity of predictions having a true attribute determined by the reference attribute data and a predicted attribute determined by a prediction class with which the noised sketch is associated. For instance, when v.sup.d is unnoised, this sum can equal the (g, d)-th cell of the confusion matrix, such as the number of events associated with the tag values in reference data 112 that have true class g and predicted class d. When v.sup.d is noised, this sum can be an unbiased estimate of the (g, d)-th cell of the confusion matrix, such as the number of events associated with the tag values in reference data 112 that have true class g and predicted class d).
Peng does not explicitly disclose the database that supports a probabilistic result to a query operation and the sketch limiting exposure of the confidential data. Cook, in analogous art however, discloses the database that supports a probabilistic result to a query operation ([0050] FIG. 3 is a logical block diagram illustrating an example system where two entities (310, 330) produce privacy-preserving dataset sketches (340, 350) of their respective datasets (312, 332), where the dataset sketches (340, 350) are stored in a data catalog of published sketches (390), and where a third entity (360) joins the privacy-preserving dataset sketches (340, 350) in a sketch join execution unit (370) to produce a joint probability distribution of the datasets (372), according to some embodiments. FIG. 3 is an example where two entities, entity 1 (310) and entity 2 (330), each have a respective private dataset, private dataset 1 (312) and private dataset 2 (332), with private information about individuals which must remain within the entity. For example, neither entity may send even one record from their dataset to the other entity. However, there is a column or combination of columns in both datasets (312, 332) which uniquely identifies each individual, such as name or name and address. Therefore, the system of FIG. 3 allows Entity 3 (360) to learn things about the combined datasets which could not be learned from either dataset alone, using a joint probability distribution (372) of the datasets. To do this, both entities each generate a sketch which contains information about their dataset. Privacy-preserving sketch producer (316) of Entity 1 (310) takes private dataset 1 (312), a sketch noise profile (314), and shared sketch parameters (320) as input and produces a privacy-preserving sketch of dataset 1 (340). Privacy-preserving sketch producer (336) of Entity 2 (330) takes private dataset 2 (332), a sketch noise profile (334), and the shared sketch parameters (320) as input, and produces a sketch of dataset 2 (350). Privacy-preserving sketch of dataset 1 (340) and privacy-preserving sketch of dataset 2 (350) is generated in a differentially private way, so it is safe for Entity 1 (310) and Entity 2 (330) to share it with Entity 3 (360), or to publish it in the data catalog of published sketches (390). Information about individuals is not revealed, including whether or not an individual is in the dataset, in these embodiments. [0051] In FIG. 3, Entity 1 (310) and Entity 2 (330) publishes a one-time sketch that both preserves privacy and can be used by other parties, such as Entity 3 (360) to jointly estimate a PDF (372). The privacy-preserving sketch of dataset 1 (340) can be joined with the privacy preserving sketch of dataset 2 (250) a shared identifier, such as email address or name, that can be specified in the shared parameters (320) to do tasks involving columns from either or both private dataset 1 (312) or private dataset 2 (332). In FIG. 3, the task is viewing an approximate joint distribution of column values with the joint probability distribution (372). Entity 1 (310) publishes a one-time sketch (340) of its data. Entity 2 (330) publishes a one-time sketch (350) of its data. These sketches enable Entity 3 (360) to join the sketches in the sketch join execution unit (370). When computing the joint PDF (372), the goal can be measure the number of individuals who have each combination of possible values across the two datasets. [0052] FIG. 4 is a logical block diagram illustrating an example system where a dataset A server (430) receives a query (482) from a client, produces a sketch (450) of its own dataset (432), receives a privacy-preserving dataset sketch (440) of another dataset (412) that was produced (416) by another server (410), joins the two datasets together in the sketch join execution unit (470), and responds to the query (484), according to some embodiments. As shown, the system 400 may include one or more clients 460, which communicates with a service provider network 405 over a network 480. [0057] In some embodiments, the dataset service or server 430 may receive query requests 482, and responsively return query results 484. In some embodiments, the dataset server 430 may implement a relational database. In some embodiments, the query 482 may be specified in a Structured Query Language (SQL) request. In some embodiments, the query may specify the joining of two or more datasets, such as for examples dataset A (432) and dataset B (412), as shown. In some embodiments, the dataset service or server 430 may provide a variety of database management functions, such as the insertion, updating, and deletion of data in the database; the creation, deletion, and modification of database objects such as tables or indexes; and the loading of data via various data loading tools such as extract, transform, and load (ETL) tools. [0058] In some embodiments, the dataset service or server 430 may be configured to maintain an encrypted database. For example, in some embodiments, the datasets in the database, such as dataset A 432 may be encrypted. Depending on the embodiment, the tables or other data set objects may be encrypted using different techniques. In some embodiments, the individual data items or attributes in the tables may be encrypted via an encryption scheme. In some embodiments, the database system or service may not decrypt the encrypted data in the encrypted tables, and the query results 484 may be returned in encrypted form. Depending on the embodiment, a variety of different security techniques may be employed to hide the data in the encrypted tables from unauthorized users. Depending on the embodiment, a variety of different encryption schemes may be used. [0059] As shown, in some embodiments, the dataset service or server 430 may implement a query parser 490, a sketch join execution unit 470, and a result assembler 495. In some embodiments, some of these components may be implements as part of a larger querying handling subsystem of the database service. In some embodiments, the sketch join execution unit 470 may be part of a query engine implemented by the dataset service or server 430. In some embodiments, the query engine may implement a variety of querying functionalities of the dataset service or server 430. [0060] In some embodiments, the query 482 may be provided to the query parser 490, which may parse the query in received form into a representation that may be used for processing by the query engine. In some embodiments, the query 482 may be received as a query token. In some embodiments, the query token may be encrypted or compressed. In some embodiments, the query token may specify the select columns of the query, one or more filter conditions, and one or more join conditions, among other things. In some embodiments, the query parser 490 may extract such information from the query token and translate this information into parameters that can be used by the query engine. For example, in some embodiments, the query parser 490 may translate columns (e.g., the join columns for datasets A and B) into column identifiers that can be used by the query engine to access the correct columns. In some embodiments, as shown, the query parser 490 may provide the column identifiers, for example, the join columns for datasets A and B, to the sketch join execution unit 470); and the sketch limiting exposure of the confidential data ([0034] . The sketches are generated in a differentially private way, in these embodiments, so it is safe for the organizations to share them with each other or publish them. Information about individuals is not revealed, including whether or not an individual is in either dataset, in these embodiments. [0043] For example, suppose a vaccine clinic knows who has been vaccinated for a specific health condition, and a hospital knows the specific health condition's status of some patients. A researcher would like to combine their data to understand vaccine effectiveness. This can be done as follows: The hospital publishes a sketch of the status of the specific health condition, or one can be reused from previously. The vaccine clinic publishes a sketch of vaccination status. The researcher downloads both sketches, and produces an approximate contingency table from them. This contingency table can be produced without revealing that any specific person is “+” for the health condition, or even that any specific person was tested at all. The contingency table is produced in a differentially private manner, i.e., the researcher cannot tell whether any specific person (e.g., John Doe) is or is not in the data. [0046] Privacy-preserving sketch producer (116) of Entity 1 (110) takes private dataset 1 (112), a sketch noise profile (114), and shared sketch parameters (120) as input and produces a privacy-preserving sketch of dataset 1 (140). Sketch producer (136) of Entity 2 (130) takes private dataset 2 (132), and the shared sketch parameters (120) as input, and produces a sketch of dataset 2 (150). Privacy-preserving sketch of dataset 1 (140) is generated in a differentially private way, so it is safe for Entity 1 (110) to share it with Entity 2 (130), or to publish it). Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the invention to modify the claimed limitations of the database disclosed by Peng that supports a probabilistic result to a query operation and the sketch limiting exposure of the confidential data. This modification would have been obvious because a person having ordinary skill in the art would have been motivated by the desire to implement non-interactive join of privacy-preserving dataset sketch with an entity that can publish a one-time sketch of their dataset that would enable another entity to join their data without exposing private information as suggested by Cook ([0024-0025]).
Peng and Cook do not explicitly disclose selecting a privacy parameter based, at least in part, on an amount of querying data available and the applied noise is based on the privacy parameter. Cesar, in analogous art however, discloses selecting a privacy parameter based, at least in part, on an amount of querying data available and the applied noise is based on the privacy parameter ([0018] Additionally, according to some embodiments, the method described offers an innovative approach to maintaining privacy while allowing for repeated queries without automatically reducing the privacy budget when the underlying data remains unchanged. This is achieved through the use of deterministic pseudorandom noise that is based not only on the query parameters but also on the state of the database at the time of query execution)); and the applied noise is based on the privacy parameter ([0043] To satisfy Definition 1, the techniques add noise (e.g., from a Laplace distribution) to the true result of a statistical query function. This function could represent, for example, the number of members who clicked on an article or the histogram of titles of those members. The amount of noise added is determined by the L.sub.1 sensitivity of the query, which is the maximum change in the query output when a single member is added or removed from the dataset. Additionally, the level of noise is influenced by the desired privacy guarantee, represented by the parameter epsilon (e). The noisy result is then released, ensuring that individual privacy is preserved while providing useful aggregate data. [0044] Definition 2: The L.sub.1 sensitivity of a query function, ƒ: D à R.sup.d is defined as: [0045] Δ(ƒ)=max.sub.D,D′∥ƒ(D)−F(D′)∥.sub.1, for all pairs of datasets (D, D′) different in at most one row. [0046] Theorem 1: Given a query function ƒ: D à R.sup.d, the randomized mechanism K that adds noise drawn independently from the noise distribution (e.g., Laplace distribution) with parameter [00001]Δ(f)ε to each of the d dimensions of ƒ(D) satisfies ε—differential privacy. [0087] The pseudorandom Laplace noise generation algorithm 200 does not take into account changes in the underlying database between repeated queries because it is designed to be deterministic and data-independent. This algorithm generates noise based on the input parameters: the fixed secret (s), statistics type (stat_type), entity ID (eid), demographic attribute-value pair (d_attr, d_val), atomic time range (ta), and privacy parameter (epsilon). [0088] An aspect of this design is the use of a deterministic pseudorandom function (GeneratePseudorandFrac) that always produces the same output for a given set of inputs. This function doesn't query the database or incorporate any information about the current state of the data. Instead, it uses cryptographic techniques (likely a hash function, as seen in the pseudocode 300 implementation) to generate a consistent pseudorandom value based on the input parameters. [0089] As a result, if the same query parameters are used (e.g., the same combination of s, stat_type, eid, d_attr, d_val, ta, and epsilon), the algorithm 200 will generate identical noise regardless of whether the underlying data has changed. This property ensures consistency in the noise added to repeated queries, which is useful for certain use cases like maintaining consistent results in analytics dashboards over time. Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the invention to modify the claimed limitations of the database disclosed by Peng and Cook to selecting a privacy parameter based, at least in part, on an amount of querying data available and the applied noise is based on the privacy parameter. This modification would have been obvious because a person having ordinary skill in the art would have been motivated by the desire to generate deterministic pseudorandom noise from both the query and a representation of the database’s current state. If the same query is run again on unchanged data, the same noise is reused and the result stays consistent. If the database changes, the state input changes, causing a different noise value to be produced. This lets the system preserve privacy, resist averaging attacks, and avoid unnecessary privacy-budget consumption for repeated queries on unchanged data as suggested by Cesar ([0015]-[0024], [0095]-[0098], [0120]-[0123]).
As per claim 2. Peng in view of Cook in further view of Cesar discloses the method as described in claim 1, wherein the applying of the noise to the dataset record is performed using a differential privacy technique based, at least in part, on an Epsilon parameter (Cesar [0046] Theorem 1: Given a query function ƒ: D à R.sup.d, the randomized mechanism K that adds noise drawn independently from the noise distribution (e.g., Laplace distribution) with parameter Δ(f)ε to each of the d dimensions of ƒ(D) satisfies ε—differential privacy. [0047] Noise can be added from other types of distributions (e.g., other than a Laplace distribution) according to the requirements of the particular implementation at hand. For example, some possible alternatives to using a Laplace distribution for adding noise in differential privacy include: [0048] Gaussian Distribution: Noise is added based on the Gaussian (normal) distribution. This method may be used in the context of (ε, δ)-differential privacy, which provides a relaxation of pure differential privacy and can offer better accuracy under certain conditions. [0049] Exponential Distribution: This may be useful in specific types of differential privacy mechanisms like the exponential mechanism, which selects an output from a set of possible outputs based on a probability distribution weighted by a utility function. [0050] These alternatives can be chosen based on the specific requirements of the differential privacy application, such as the desired level of privacy, the nature of the data, and the computational efficiency).
As per claim 3. Peng in view of Cook in further view of Cesar discloses the method as described in claim 1, wherein the applying of the noise is performed using a bit-flip technique (Peng [0100] In an example that uses the binary vector expansion, another noising mechanism can include randomly performing bitflips in the vectors with a probability p. Bit-flipping with probability p can provide e-DP with ϵ=ln(1-pp). [0114] In an example with bitflip noise, if the bitflip probability p is known, the number of positive predictions can be scaled by a probability of no bitflip (q=1−p) to give a number of expected actual pre-noise predictions that survived the bitflip. The number of negative predictions can also be scaled by a probability of bitflip p to give a number of expected actual pre-noise predictions that were canceled by the bitflip. These scaled totals can be combined and rescaled to obtain a total number of expected actual pre-noise predictions {circumflex over (n)} for each ground truth class g, prediction class d, and frequency level f).
As per claim 4. Peng in view of Cook in further view of Cesar discloses the method as described in claim 1, wherein the dataset record includes an identity key, a respective attribute, and confidential information (Cook [0068] FIG. 8 is an example of an organizational-level model for differential privacy, with vertically partitioned data, according to some embodiments. Some embodiments of the privacy-preserving dataset sketches that can be joined non-interactively operate in an organizational-level model, i.e., vertically partitioned data. It operates in neither the global nor the local model. Each organization has its own data and the goal is to learn properties of a join. For example, party A has salary information for a group of identities (i.e., persons), and party B has the specific health condition positivity data for different group of identities (i.e., persons), where some of the identities overlap between party A and party B. In this approach, when the data is joined, the identities of the individuals who do and do not contribute to the join is not revealed. [0109] FIG. 16 is a high-level flow chart that illustrates a process for obtaining a privacy-preserving first dataset sketch, obtaining a second dataset, joining the first sketch with the second dataset to produce a joined dataset, and performing a data processing task on the joined dataset to generate a data artifact, according to some embodiments. The flowchart begins at block 1610 which obtains a privacy-preserving first sketch of a first private dataset that was produced according to a plurality of shared parameters. This first private dataset includes multiple identities, where each identity is associated with a value of a first set of values. The privacy-preserving first sketch at least contains, for a first value, all the identities that correspond to the first value mapped to entries of a vector using a hash function. The privacy-preserving first sketch might also contain, for a second value of the first set of values, all the identities that correspond to the second value mapped to entries of a second vector using the hash function. The privacy-preserving first sketch might also contain, for other values of the first set of values, all the identities that correspond to the respective other values mapped to entries of respective other vectors using the hash function. The privacy-preserving first sketch also contains added noise. The flowchart then transitions to 1620 which obtains a second dataset. Next, step 1630 joins the privacy-preserving first sketch of the first private dataset with the second dataset using at least some of the plurality of shared parameters, and using a join identifier common to both the privacy-preserving first sketch and the second dataset, to produce a joined dataset. Finally, the flowchart transitions to block 1640 which performs a data processing task on the joined dataset to generate a data artifact that maintains the privacy features of the privacy-preserving first sketch).
As per claim 5. Peng in view of Cook in further view of Cesar discloses the method as described in claim 4, wherein the applying of the noise to the dataset record includes applying the noise to the identity key of the dataset record ([0050] Privacy-preserving sketch producer (316) of Entity 1 (310) takes private dataset 1 (312), a sketch noise profile (314), and shared sketch parameters (320) as input and produces a privacy-preserving sketch of dataset 1 (340). Privacy-preserving sketch producer (336) of Entity 2 (330) takes private dataset 2 (332), a sketch noise profile (334), and the shared sketch parameters (320) as input, and produces a sketch of dataset 2 (350). Privacy-preserving sketch of dataset 1 (340) and privacy-preserving sketch of dataset 2 (350) is generated in a differentially private way, so it is safe for Entity 1 (310) and Entity 2 (330) to share it with Entity 3 (360), or to publish it in the data catalog of published sketches (390)).
As per claim 6. Peng in view of Cook in further view of Cesar discloses the method as described in claim 4, wherein the sketch, as stored in the database, does not support direct identification of the confidential information via the database (Cook [0062] Privacy-preserving sketch producer (416) of dataset B server (410) takes private dataset B (412), a sketch noise profile (414), and shared sketch parameters (420) as input and produces a privacy-preserving sketch of dataset 1 (440). Sketch producer (436) of dataset A server (430) takes private dataset A (432), and the shared sketch parameters (420) as input, and produces a sketch of dataset A (450). Privacy-preserving sketch of dataset B (440) is generated in a differentially private way, so it is safe for dataset B server (410) to share it with dataset A server (430), or to publish it. Information about individuals is not revealed, including whether or not an individual is in the dataset, in these embodiments).
As per claim 7. Peng in view of Cook in further view of Cesar discloses the method as described in claim 4, wherein the confidential information is a membership identifier (ID) of a respective entity associated with the attribute for the identity key (Cook [0068] FIG. 8 is an example of an organizational-level model for differential privacy, with vertically partitioned data, according to some embodiments. Some embodiments of the privacy-preserving dataset sketches that can be joined non-interactively operate in an organizational-level model, i.e., vertically partitioned data. It operates in neither the global nor the local model. Each organization has its own data and the goal is to learn properties of a join. For example, party A has salary information for a group of identities (i.e., persons), and party B has the specific health condition positivity data for different group of identities (i.e., persons), where some of the identities overlap between party A and party B. In this approach, when the data is joined, the identities of the individuals who do and do not contribute to the join is not revealed).
As per claim 8. Peng in view of Cook in further view of Cesar discloses the method as described in claim 4, further comprising: forming a query for processing by the database; receiving a probabilistic result to the query from the database; and resolving the sketch, as included in the probabilistic result, to the confidential information based on a mapping (Cook [0062] FIG. 4 is an example where two entities, dataset B server (410) and dataset A server (430), each have a respective private dataset, private dataset B (412) and private dataset A (432), with private information about individuals which must remain within the entity. For example, neither entity may send even one record from their dataset to the other entity. However, there is a column or combination of columns in both datasets (412, 432) which uniquely identifies each individual, such as name or name and address. Therefore, the system of FIG. 4 allows Entity A (430) to learn things about the combined datasets which could not be learned from either dataset alone, using a joint probability distribution of the datasets. To do this, both entities each generate a sketch which contains information about their dataset. Privacy-preserving sketch producer (416) of dataset B server (410) takes private dataset B (412), a sketch noise profile (414), and shared sketch parameters (420) as input and produces a privacy-preserving sketch of dataset 1 (440). Sketch producer (436) of dataset A server (430) takes private dataset A (432), and the shared sketch parameters (420) as input, and produces a sketch of dataset A (450). Privacy-preserving sketch of dataset B (440) is generated in a differentially private way, so it is safe for dataset B server (410) to share it with dataset A server (430), or to publish it. Information about individuals is not revealed, including whether or not an individual is in the dataset, in these embodiments).
As per claim 9. Peng in view of Cook in further view of Cesar discloses the method as described in claim 1, further comprising forming a mapping of confidential information included in the dataset record to the sketch and wherein the sketch is configured to be stored independent of the confidential information (Cook [0063] The privacy-preserving sketch of dataset B (440) can be joined with other datasets on a shared identifier, such as email address or name, to do tasks involving columns from private dataset A (432). In FIG. 4, the task can be assembling an approximate joint distribution of column values with the joint probability distribution. Dataset B server (410) publishes a one-time sketch (440) of its data that enables dataset A server (430) to join (470) a sketch (450) of its data privately. When computing the joint PDF, a goal can be measure the number of individuals who have all combination of possible values across the two datasets, in some embodiments).
As per claim 10. Peng in view of Cook in further view of Cesar discloses the method as described in claim 9, wherein the mapping is configured to resolve the sketch as included in the probabilistic result to the confidential information (Cook [0051] In FIG. 3, Entity 1 (310) and Entity 2 (330) publishes a one-time sketch that both preserves privacy and can be used by other parties, such as Entity 3 (360) to jointly estimate a PDF (372). The privacy-preserving sketch of dataset 1 (340) can be joined with the privacy preserving sketch of dataset 2 (250) a shared identifier, such as email address or name, that can be specified in the shared parameters (320) to do tasks involving columns from either or both private dataset 1 (312) or private dataset 2 (332). In FIG. 3, the task is viewing an approximate joint distribution of column values with the joint probability distribution (372). Entity 1 (310) publishes a one-time sketch (340) of its data. Entity 2 (330) publishes a one-time sketch (350) of its data. These sketches enable Entity 3 (360) to join the sketches in the sketch join execution unit (370). When computing the joint PDF (372), the goal can be measure the number of individuals who have each combination of possible values across the two datasets).
As per claim 11. Peng in view of Cook in further view of Cesar discloses the method as described in claim 1, wherein the generating and the applying are performed in a protected environment and the data that supports the probabilistic result is maintained in a shared environment (Peng [0049] In general, a system can be associated with sets of client devices, for example by maintaining a client tag value that is associated with the respective client device. Each client tag value can correspond to attribute information that describes the association between a content server and client device. Attribute information can include information about the relationship between the client device and the content server (e.g., web-browsing history, interaction data, session time, network analysis data), and can include protected or otherwise private information received from the respective client device. [0050] This traditional approach can be suboptimal. For example, this raw shared information can include protected or private information. Sharing this raw information can lead to decreased privacy. Further, such communications between the content servers and the centralized server can involve network communications that provide increased numbers of attack vectors for security breaches).
As per claim 12. Peng discloses a computing device comprising:
a processing device (Figure 9: computing device 50, Server computing system 60); and
a computer-readable storage medium storing instruction that, responsive to execution by the processing device (Computing Device 98), causes the processing device to perform operations including:
receiving a query for processing by a database ([0212] Model alignment toolkit 17 can integrate one or more dataset(s) 17-1 for aligning development model 16. Curated dataset(s) 17-1 can include labeled or unlabeled training data. Dataset(s) 17-1 can be obtained from public domain datasets. Dataset(s) 17-1 can be obtained from private datasets associated with one or more developer system(s) for the alignment of bespoke machine-learned model(s) customized for private use-cases. [0213] Pre-training pipelines 17-2 can include a machine-learned model training workflow configured to update development model 16 over large-scale, potentially noisy datasets. For example, pre-training can leverage unsupervised learning techniques (e.g., de-noising, etc.) to process large numbers of training instances to update model parameters from an initialized state and achieve a desired baseline performance. Pre-training pipelines 17-2 can leverage unlabeled datasets in dataset(s) 17-1 to perform pre-training. Workbench 15 can implement a pre-training pipeline 17-2 to pre-train development model 16. [0214] Fine-tuning pipelines 17-3 can include a machine-learned model training workflow configured to refine the model parameters of development model 16 with higher-quality data. Fine-tuning pipelines 17-3 can update development model 16 by conducting supervised training with labeled dataset(s) in dataset(s) 17-1. Fine-tuning pipelines 17-3 can update development model 16 by conducting reinforcement learning using reward signals from user feedback signals. Workbench 15 can implement a fine-tuning pipeline 17-3 to fine-tune development model 16);
the database including a plurality of sketches ([0069] Prediction system 102 can implement a noising pipeline 108 to add noise to sketches generated by sketch pipeline 106. The noised sketches 109 can be transmitted to a reference system 110. Reference system 110 can store reference data 112. Reference data 112 can include event data objects with corresponding reference values associated therewith. Reference data 112 can include a subset of event data objects that are in log data 104. Reference system 110 can implement a confusion matrix estimation pipeline 114 that estimates a confusion matrix for a noised sketch generated by noising pipeline 108 based on reference data 112. Confusion matrix estimation pipeline 114 can generate confusion matrix data 116),
each said sketch configured as a probabilistic data structure using a differential privacy technique ([0052] A prediction system (e.g., a content provider system, a tag value server, etc.) can generate a sketch that compiles predicted characteristics for the served content events. The sketch can be indexed by, for instance, a tag value. The sketch can be noised or otherwise obscured. The noised sketch can satisfy a differential privacy criterion. The noised sketch can be transmitted to the reference data system. [0068] Prediction system 102 can implement a sketch pipeline 106 to generate a sketch of one or more event data objects and the one or more predicted values corresponding thereto. [0098] Noise can be added to the sketches until a differential privacy criterion is satisfied. The term “differential privacy” (DP) can refer to sharing information about a dataset by describing the patterns of groups within the dataset while limiting information about individuals leaking through the sharing of the dataset. DP can be a constraint on the algorithms used to publish aggregate information about a statistical database which limits the disclosure of private information of records whose information is in the database. For instance, a DP constraint for an aggregate dataset can be satisfied when the presence of any given user's information does not shift the dataset beyond a privacy value ∈ (referred to as “∈-DP”). In some implementations, DP can be achieved by adding noise to the dataset); and
generating a sketch configured as a probabilistic data structure of the dataset having the determined amount of noise ([0127] generating a data sketch descriptive of the plurality of predicted attributes. For instance, a sketch pipeline 106 of a prediction system 102 can generate a data sketch of predictions in log data. For example, prediction system 102 can store data describing events processed by prediction system 102 and corresponding predictions generated for the events as log data (e.g., log data 104). For instance, log data 104 can include event data objects describing one or more characteristics of a content request or content delivered to a client device. Log data 104 can include any predicted values generated based on the event data object or the characteristics described by the event data object. The predicted values can include, for instance, outputs of a prediction model as described herein. Sketch pipeline 106 of prediction system 102 can include one or more data mapping tools to generate a privatized representation of log data 104, or data “sketch,” that obscures the original data describing content service events. The term “sketch” can refer to one or more data structures containing one or more data elements, data records, variables, counter registers, floating point values, strings, index values, memory pointer values, or any combination thereof as described herein. The term “sketch” and “data structure” may sometimes be used interchangeably. [0128] At 208, example method 200 can include noising the data sketch, wherein the noised data sketch satisfies a differential privacy criterion. Noising pipeline 108 can add noise to or otherwise transform sketches generated by sketch pipeline 106 to satisfy one or more privacy or security criteria. For example, various types of noise can be added to the sketches. Laplace noise can be added to the sketches. Gaussian noise can be added to the sketches. Noise can be added to the sketches until a differential privacy criterion is satisfied);and
presenting the probabilistic result for output in a user interface ([0237] Model host 31 can leverage various other resources and tools to augment the inference task. For instance, model host 31 can communicate with tool interfaces 35 to facilitate tool use by model instance(s) 31-1. Tool interfaces 35 can include local or remote APIs. Tool interfaces 35 can include integrated scripts or other software functionality. Model host 31 can engage online learning interface(s) 36 to facilitate ongoing improvements to machine-learned model(s) 1. For instance, online learning interface(s) 36 can be used within reinforcement learning loops to retrieve user feedback on inferences served by model host 31. Model host 31 can access runtime data source(s) 37 for augmenting input(s) 2 with additional contextual information. For instance, runtime data source(s) 37 can include a knowledge graph 37-1 that facilitates structured information retrieval for information associated with input request(s) 33 (e.g., a search engine service). Runtime data source(s) 37 can include public or private, external or local database(s) 37-2 that can store information associated with input request(s) 33 for augmenting input(s) 2. Runtime data source(s) 37 can include account data 37-3 which can be retrieved in association with a user account corresponding to a client 32 for customizing the behavior of model host 31 accordingly.
Peng does not explicitly disclose generating a probabilistic result by processing the query using the database based on a corresponding operation. Cook, in analogous art however, discloses generating a probabilistic result by processing the query using the database based on a corresponding operation ([0050] FIG. 3 is a logical block diagram illustrating an example system where two entities (310, 330) produce privacy-preserving dataset sketches (340, 350) of their respective datasets (312, 332), where the dataset sketches (340, 350) are stored in a data catalog of published sketches (390), and where a third entity (360) joins the privacy-preserving dataset sketches (340, 350) in a sketch join execution unit (370) to produce a joint probability distribution of the datasets (372), according to some embodiments. FIG. 3 is an example where two entities, entity 1 (310) and entity 2 (330), each have a respective private dataset, private dataset 1 (312) and private dataset 2 (332), with private information about individuals which must remain within the entity. For example, neither entity may send even one record from their dataset to the other entity. However, there is a column or combination of columns in both datasets (312, 332) which uniquely identifies each individual, such as name or name and address. Therefore, the system of FIG. 3 allows Entity 3 (360) to learn things about the combined datasets which could not be learned from either dataset alone, using a joint probability distribution (372) of the datasets. To do this, both entities each generate a sketch which contains information about their dataset. Privacy-preserving sketch producer (316) of Entity 1 (310) takes private dataset 1 (312), a sketch noise profile (314), and shared sketch parameters (320) as input and produces a privacy-preserving sketch of dataset 1 (340). Privacy-preserving sketch producer (336) of Entity 2 (330) takes private dataset 2 (332), a sketch noise profile (334), and the shared sketch parameters (320) as input, and produces a sketch of dataset 2 (350). Privacy-preserving sketch of dataset 1 (340) and privacy-preserving sketch of dataset 2 (350) is generated in a differentially private way, so it is safe for Entity 1 (310) and Entity 2 (330) to share it with Entity 3 (360), or to publish it in the data catalog of published sketches (390). Information about individuals is not revealed, including whether or not an individual is in the dataset, in these embodiments. [0051] In FIG. 3, Entity 1 (310) and Entity 2 (330) publishes a one-time sketch that both preserves privacy and can be used by other parties, such as Entity 3 (360) to jointly estimate a PDF (372). The privacy-preserving sketch of dataset 1 (340) can be joined with the privacy preserving sketch of dataset 2 (250) a shared identifier, such as email address or name, that can be specified in the shared parameters (320) to do tasks involving columns from either or both private dataset 1 (312) or private dataset 2 (332). In FIG. 3, the task is viewing an approximate joint distribution of column values with the joint probability distribution (372). Entity 1 (310) publishes a one-time sketch (340) of its data. Entity 2 (330) publishes a one-time sketch (350) of its data. These sketches enable Entity 3 (360) to join the sketches in the sketch join execution unit (370). When computing the joint PDF (372), the goal can be measure the number of individuals who have each combination of possible values across the two datasets. [0052] FIG. 4 is a logical block diagram illustrating an example system where a dataset A server (430) receives a query (482) from a client, produces a sketch (450) of its own dataset (432), receives a privacy-preserving dataset sketch (440) of another dataset (412) that was produced (416) by another server (410), joins the two datasets together in the sketch join execution unit (470), and responds to the query (484), according to some embodiments. As shown, the system 400 may include one or more clients 460, which communicates with a service provider network 405 over a network 480. [0057] In some embodiments, the dataset service or server 430 may receive query requests 482, and responsively return query results 484. In some embodiments, the dataset server 430 may implement a relational database. In some embodiments, the query 482 may be specified in a Structured Query Language (SQL) request. In some embodiments, the query may specify the joining of two or more datasets, such as for examples dataset A (432) and dataset B (412), as shown. In some embodiments, the dataset service or server 430 may provide a variety of database management functions, such as the insertion, updating, and deletion of data in the database; the creation, deletion, and modification of database objects such as tables or indexes; and the loading of data via various data loading tools such as extract, transform, and load (ETL) tools. [0058] In some embodiments, the dataset service or server 430 may be configured to maintain an encrypted database. For example, in some embodiments, the datasets in the database, such as dataset A 432 may be encrypted. Depending on the embodiment, the tables or other data set objects may be encrypted using different techniques. In some embodiments, the individual data items or attributes in the tables may be encrypted via an encryption scheme. In some embodiments, the database system or service may not decrypt the encrypted data in the encrypted tables, and the query results 484 may be returned in encrypted form. Depending on the embodiment, a variety of different security techniques may be employed to hide the data in the encrypted tables from unauthorized users. Depending on the embodiment, a variety of different encryption schemes may be used. [0059] As shown, in some embodiments, the dataset service or server 430 may implement a query parser 490, a sketch join execution unit 470, and a result assembler 495. In some embodiments, some of these components may be implements as part of a larger querying handling subsystem of the database service. In some embodiments, the sketch join execution unit 470 may be part of a query engine implemented by the dataset service or server 430. In some embodiments, the query engine may implement a variety of querying functionalities of the dataset service or server 430. [0060] In some embodiments, the query 482 may be provided to the query parser 490, which may parse the query in received form into a representation that may be used for processing by the query engine. In some embodiments, the query 482 may be received as a query token. In some embodiments, the query token may be encrypted or compressed. In some embodiments, the query token may specify the select columns of the query, one or more filter conditions, and one or more join conditions, among other things. In some embodiments, the query parser 490 may extract such information from the query token and translate this information into parameters that can be used by the query engine. For example, in some embodiments, the query parser 490 may translate columns (e.g., the join columns for datasets A and B) into column identifiers that can be used by the query engine to access the correct columns. In some embodiments, as shown, the query parser 490 may provide the column identifiers, for example, the join columns for datasets A and B, to the sketch join execution unit 470) and the sketch limiting exposure of the confidential data ([0034] . The sketches are generated in a differentially private way, in these embodiments, so it is safe for the organizations to share them with each other or publish them. Information about individuals is not revealed, including whether or not an individual is in either dataset, in these embodiments. [0043] For example, suppose a vaccine clinic knows who has been vaccinated for a specific health condition, and a hospital knows the specific health condition's status of some patients. A researcher would like to combine their data to understand vaccine effectiveness. This can be done as follows: The hospital publishes a sketch of the status of the specific health condition, or one can be reused from previously. The vaccine clinic publishes a sketch of vaccination status. The researcher downloads both sketches, and produces an approximate contingency table from them. This contingency table can be produced without revealing that any specific person is “+” for the health condition, or even that any specific person was tested at all. The contingency table is produced in a differentially private manner, i.e., the researcher cannot tell whether any specific person (e.g., John Doe) is or is not in the data. [0046] Privacy-preserving sketch producer (116) of Entity 1 (110) takes private dataset 1 (112), a sketch noise profile (114), and shared sketch parameters (120) as input and produces a privacy-preserving sketch of dataset 1 (140). Sketch producer (136) of Entity 2 (130) takes private dataset 2 (132), and the shared sketch parameters (120) as input, and produces a sketch of dataset 2 (150). Privacy-preserving sketch of dataset 1 (140) is generated in a differentially private way, so it is safe for Entity 1 (110) to share it with Entity 2 (130), or to publish it). Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the invention to modify the claimed limitations of the database disclosed by Peng generating a probabilistic result by processing the query using the database based on a corresponding operation. This modification would have been obvious because a person having ordinary skill in the art would have been motivated by the desire to implement non-interactive join of privacy-preserving dataset sketch with an entity that can publish a one-time sketch of their dataset that would enable another entity to join their data without exposing private information as suggested by Cook ([0024-0025]).
Peng and Cook do not explicitly disclose selecting a privacy parameter based, at least in part, on an amount of querying data available and determining an amount of noise applied to a dataset based on the privacy parameter, the database including confidential data. Cesar, in analogous art however, discloses selecting a privacy parameter based, at least in part, on an amount of querying data available and the applied noise is based on the privacy parameter and determining an amount of noise applied to a dataset based on the privacy parameter, the database including confidential data ([0018] Additionally, according to some embodiments, the method described offers an innovative approach to maintaining privacy while allowing for repeated queries without automatically reducing the privacy budget when the underlying data remains unchanged. This is achieved through the use of deterministic pseudorandom noise that is based not only on the query parameters but also on the state of the database at the time of query execution)); and the applied noise is based on the privacy parameter ([0043] To satisfy Definition 1, the techniques add noise (e.g., from a Laplace distribution) to the true result of a statistical query function. This function could represent, for example, the number of members who clicked on an article or the histogram of titles of those members. The amount of noise added is determined by the L.sub.1 sensitivity of the query, which is the maximum change in the query output when a single member is added or removed from the dataset. Additionally, the level of noise is influenced by the desired privacy guarantee, represented by the parameter epsilon (e). The noisy result is then released, ensuring that individual privacy is preserved while providing useful aggregate data. [0044] Definition 2: The L.sub.1 sensitivity of a query function, ƒ: D à R.sup.d is defined as: [0045] Δ(ƒ)=max.sub.D,D′∥ƒ(D)−F(D′)∥.sub.1, for all pairs of datasets (D, D′) different in at most one row. [0046] Theorem 1: Given a query function ƒ: D à R.sup.d, the randomized mechanism K that adds noise drawn independently from the noise distribution (e.g., Laplace distribution) with parameter [00001]Δ(f)ε to each of the d dimensions of ƒ(D) satisfies ε—differential privacy. [0087] The pseudorandom Laplace noise generation algorithm 200 does not take into account changes in the underlying database between repeated queries because it is designed to be deterministic and data-independent. This algorithm generates noise based on the input parameters: the fixed secret (s), statistics type (stat_type), entity ID (eid), demographic attribute-value pair (d_attr, d_val), atomic time range (ta), and privacy parameter (epsilon). [0088] An aspect of this design is the use of a deterministic pseudorandom function (GeneratePseudorandFrac) that always produces the same output for a given set of inputs. This function doesn't query the database or incorporate any information about the current state of the data. Instead, it uses cryptographic techniques (likely a hash function, as seen in the pseudocode 300 implementation) to generate a consistent pseudorandom value based on the input parameters. [0089] As a result, if the same query parameters are used (e.g., the same combination of s, stat_type, eid, d_attr, d_val, ta, and epsilon), the algorithm 200 will generate identical noise regardless of whether the underlying data has changed. This property ensures consistency in the noise added to repeated queries, which is useful for certain use cases like maintaining consistent results in analytics dashboards over time. Therefore, it would have been obvious to a person having ordinary skill in the art before the effective filing date of the invention to modify the claimed limitations of the database disclosed by Peng and Cook to selecting a privacy parameter based, at least in part, on an amount of querying data available and the applied noise is based on the privacy parameter and determining an amount of noise applied to a dataset based on the privacy parameter, the database including confidential data. This modification would have been obvious because a person having ordinary skill in the art would have been motivated by the desire to generate deterministic pseudorandom noise from both the query and a representation of the database’s current state. If the same query is run again on unchanged data, the same noise is reused and the result stays consistent. If the database changes, the state input changes, causing a different noise value to be produced. This lets the system preserve privacy, resist averaging attacks, and avoid unnecessary privacy-budget consumption for repeated queries on unchanged data as suggested by Cesar ([0015]-[0024], [0095]-[0098], [0120]-[0123]).
As per claim 13. Peng in view of Cook discloses the computing device as described in claim 12, wherein the differential privacy technique is configured to apply noise to a dataset record used to form a sketch included in the probabilistic result, the noise determined based, at least in part, on an Epsilon parameter (Cesar [0046] Theorem 1: Given a query function ƒ: D à R.sup.d, the randomized mechanism K that adds noise drawn independently from the noise distribution (e.g., Laplace distribution) with parameter Δ(f)ε to each of the d dimensions of ƒ(D) satisfies ε—differential privacy. [0047] Noise can be added from other types of distributions (e.g., other than a Laplace distribution) according to the requirements of the particular implementation at hand. For example, some possible alternatives to using a Laplace distribution for adding noise in differential privacy include: [0048] Gaussian Distribution: Noise is added based on the Gaussian (normal) distribution. This method may be used in the context of (ε, δ)-differential privacy, which provides a relaxation of pure differential privacy and can offer better accuracy under certain conditions. [0049] Exponential Distribution: This may be useful in specific types of differential privacy mechanisms like the exponential mechanism, which selects an output from a set of possible outputs based on a probability distribution weighted by a utility function. [0050] These alternatives can be chosen based on the specific requirements of the differential privacy application, such as the desired level of privacy, the nature of the data, and the computational efficiency).
As per claim 14. Peng in view of Cook discloses the computing device as described in claim 12, wherein the differential privacy technique includes a bit-flip technique (Peng [0100] In an example that uses the binary vector expansion, another noising mechanism can include randomly performing bitflips in the vectors with a probability p. Bit-flipping with probability p can provide e-DP with ϵ=ln(1-pp). [0114] In an example with bitflip noise, if the bitflip probability p is known, the number of positive predictions can be scaled by a probability of no bitflip (q=1−p) to give a number of expected actual pre-noise predictions that survived the bitflip. The number of negative predictions can also be scaled by a probability of bitflip p to give a number of expected actual pre-noise predictions that were canceled by the bitflip. These scaled totals can be combined and rescaled to obtain a total number of expected actual pre-noise predictions {circumflex over (n)} for each ground truth class g, prediction class d, and frequency level f).
As per claim 15. Peng in view of Cook discloses the computing device as described in claim 12, wherein the probabilistic result includes a sketch formed based on an operation performed using one or more of the plurality of sketches included in the database (Cook [0058] In some embodiments, the dataset service or server 430 may be configured to maintain an encrypted database. For example, in some embodiments, the datasets in the database, such as dataset A 432 may be encrypted. Depending on the embodiment, the tables or other data set objects may be encrypted using different techniques. In some embodiments, the individual data items or attributes in the tables may be encrypted via an encryption scheme. In some embodiments, the database system or service may not decrypt the encrypted data in the encrypted tables, and the query results 484 may be returned in encrypted form. Depending on the embodiment, a variety of different security techniques may be employed to hide the data in the encrypted tables from unauthorized users. Depending on the embodiment, a variety of different encryption schemes may be used. [0059] As shown, in some embodiments, the dataset service or server 430 may implement a query parser 490, a sketch join execution unit 470, and a result assembler 495. In some embodiments, some of these components may be implements as part of a larger querying handling subsystem of the database service. In some embodiments, the sketch join execution unit 470 may be part of a query engine implemented by the dataset service or server 430. In some embodiments, the query engine may implement a variety of querying functionalities of the dataset service or server 430. [0060] In some embodiments, the query 482 may be provided to the query parser 490, which may parse the query in received form into a representation that may be used for processing by the query engine. In some embodiments, the query 482 may be received as a query token. In some embodiments, the query token may be encrypted or compressed. In some embodiments, the query token may specify the select columns of the query, one or more filter conditions, and one or more join conditions, among other things. In some embodiments, the query parser 490 may extract such information from the query token and translate this information into parameters that can be used by the query engine. For example, in some embodiments, the query parser 490 may translate columns (e.g., the join columns for datasets A and B) into column identifiers that can be used by the query engine to access the correct columns. In some embodiments, as shown, the query parser 490 may provide the column identifiers, for example, the join columns for datasets A and B, to the sketch join execution unit 470)).
As per claim 16. Peng in view of Cook discloses the computing device as described in claim 15, wherein the one or more sketches are generated using the differential privacy technique (Peng [0098] Noise can be added to the sketches until a differential privacy criterion is satisfied. The term “differential privacy” (DP) can refer to sharing information about a dataset by describing the patterns of groups within the dataset while limiting information about individuals leaking through the sharing of the dataset. DP can be a constraint on the algorithms used to publish aggregate information about a statistical database which limits the disclosure of private information of records whose information is in the database. For instance, a DP constraint for an aggregate dataset can be satisfied when the presence of any given user's information does not shift the dataset beyond a privacy value ∈ (referred to as “∈-DP”). In some implementations, DP can be achieved by adding noise to the dataset).
As per claims 17-20:
Claims 17-20 are directed to one or more computer-readable storage media storing instructions that, responsive to execution by a processing device, causes the processing device to perform operations to perform substantially similar corresponding limitations of claims 12-16 respectively and therefore claims 17-20 are rejected with the same rationale given above to corresponding limitations of claims 17-20 respectively.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Applicant's amendment necessitated the new ground(s) of rejection presented in this Office action. Accordingly, THIS ACTION IS MADE FINAL. See MPEP § 706.07(a). Applicant is reminded of the extension of time policy as set forth in 37 CFR 1.136(a).
A shortened statutory period for reply to this final action is set to expire THREE MONTHS from the mailing date of this action. In the event a first reply is filed within TWO MONTHS of the mailing date of this final action and the advisory action is not mailed until after the end of the THREE-MONTH shortened statutory period, then the shortened statutory period will expire on the date the advisory action is mailed, and any nonprovisional extension fee (37 CFR 1.17(a)) pursuant to 37 CFR 1.136(a) will be calculated from the mailing date of the advisory action. In no event, however, will the statutory period for reply expire later than SIX MONTHS from the mailing date of this final action.
Bhowmick et al. (US 20180349636 A1) discuses a privacy mechanism to protect user data when transmitting the data to a server that estimates a frequency of such data amongst a set of client devices. A differential privacy mechanism is implemented using a count-mean-sketch technique that can reduce resource requirements required to enable privacy while providing provable guarantees regarding privacy and utility. For instance, the mechanism can provide the ability to tailor utility (e.g. accuracy of estimations) against the resource requirements (e.g. transmission bandwidth and computation complexity).
Cesar et al. (US 20260003996 A1) discuses techniques for executing privacy-preserving aggregation queries on a database table include receiving a query, obtaining the true output, and generating deterministic pseudorandom noise based on the query and current database state. This noise is added to the true output to create a privacy-protected result. For subsequent queries, this process is repeated, potentially with updated state data, ensuring that privacy protection adapts to changes in the underlying data. The approach maintains consistency for repeated queries on unchanged data while providing fresh noise when data changes. It balances privacy protection with data utility, allowing for various query types while guarding against privacy attacks. The techniques include displaying the noisy output to a user interface. The techniques offer adaptive privacy protection, query flexibility, and efficient resource utilization, making them suitable for dynamic data environments requiring both privacy and analytical capabilities.
NOURI et al. (US 20240249019 A1) discuses generation of a synthetic dataset that simultaneously represents the original data and preserves privacy are provided. The objective of answering statistical queries in a differentially-private manner is addressed by providing data-independent and data-dependent algorithms based on space partitioning for differentially private dataset release. These algorithms are intuitive and transparent, resulting in theoretical results on the utility-privacy tradeoffs where utility is measured with respect to kernel density preservation. The data-dependent approach overcomes the curse of dimensionality and leads to a scalable algorithm.
Contact Information
Any inquiry concerning this communication or earlier communications from the examiner should be directed to TECHANE GERGISO whose telephone number is (571)272-3784. The examiner can normally be reached 9:30am to 6:30pm.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, LINGLAN EDWARDS can be reached at (571) 270-5440. The fax phone number for the organization where this application or proceeding is assigned is 571-273-8300.
Information regarding the status of published or unpublished applications may be obtained from Patent Center. Unpublished application information in Patent Center is available to registered users. To file and manage patent submissions in Patent Center, visit: https://patentcenter.uspto.gov. Visit https://www.uspto.gov/patents/apply/patent-center for more information about Patent Center and https://www.uspto.gov/patents/docx for information about filing in DOCX format. For additional questions, contact the Electronic Business Center (EBC) at 866-217-9197 (toll-free). If you would like assistance from a USPTO Customer Service Representative, call 800-786-9199 (IN USA OR CANADA) or 571-272-1000.
/TECHANE GERGISO/Primary Examiner, Art Unit 2408
Prosecution Insights