CTFR 18/332,974 CTFR 98555 DETAILED ACTION 12-151 AIA 26-51 12-51 Status of Claims Claim(s) 1-4, 8-12, and 16 are pending and are examined herein. Claim(s) 1-4, 8-12, and 16 have been Amended. Claim(s) 5-7 and 13-15 are Canceled. Claim(s) 1-4, 8-12, and 16 remain rejected under 35 U.S.C. § 112 and 35 U.S.C. § 103. Notice of Pre-AIA or AIA Status 07-03-aia AIA 15-10-aia The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA. Priority Acknowledgment is made of the applicant’s claim to foreign priority Application No. 10-2022-0168796 filed on December 6, 2022. Response to Amendment The amendment filed on June 23, 2025 has been entered. Claims 1-20 are pending in the application. Applicant’s amendments to claims have overcome the rejection under 35 U.S.C. § 101 previously set forth in the Non-Final Office Action mailed on April 21, 2025. Applicant’s amendments to the claims have been fully considered and are addressed in the rejections below. Response to Arguments Applicant's arguments , with respect to the rejection under 35 U.S.C. § 103 filed on 04/27/2026 (see remarks Pp. 6-9) have been fully considered but are not persuasive and are moot in view of the new grounds of rejection necessitated by amendments. The Examiner refers to the updated rejection under 35 U.S.C. § 103 for more details. Claim Rejections - 35 USC § 112 07-30-02 AIA The following is a quotation of 35 U.S.C. 112(b): (b) CONCLUSION. —The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention. The following is a quotation of 35 U.S.C. 112 (pre-AIA), second paragraph: The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention. 07-34-01 Claim(s) 1-4, 8-12, and 16 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA), as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor, for pre-AIA the applicant regards as the invention. Regarding Currently Amended Claim 1 , the claim recites limitations that render the scope of the claimed invention indefinite for the following reasons: The preamble of the claim recites “ A client model training method performed by a client among a plurality of participating clients in a decentralized learning environment, the client model training method comprising ” lines 1-3, which appears to define the claimed method as being subjected to an being performed by “a client participant” among participating clients. However, the limitations reciting “ wherein performing the model consensus on all the training results includes: selecting any one of the participating clients as a leader according to a round robin method; configuring test data existing in the leader as a test set; checking accuracy of all candidate models, which are generated by the participating clients by training the global model, based on the test set; and performing the model consensus with a candidate model with highest accuracy of all the candidate models ” without identifying the element that is used to perform these steps/acts. Since the claimed process for these limitations fails to define the element/elements performing the operations, it is therefore unclear whether the claimed client itself performs these steps, the leader (which may be a different participating client) performs the recited steps, or the participating clients collectively perform the recited steps. The claim also fails to clearly define the relationship between the claimed client and the recited leader. Thus, the scope of the claim is not clear, and one of ordinary skill in the art cannot determine the scope of the claim with reasonable certainty. The claim recite the limitations “ generating a candidate model by training a global model; transmitting a training result of the client for sharing the candidate model to the other clients within a critical time of the participating clients; receiving other training results from at least one of the other clients; performing model consensus on all training results, which are composed of the training result and the other training results to generate a consented model according to a predefined consensus algorithm ;” lines 4-11. The claim appear to use the terms “training results,” “all training results” and “candidate models” interchangeable without defining the relationship between them. Thus, it is unclear whether the claimed “training result” and “candidate models” represent the same thing, the claimed “training result” includes a candidate model with other undefined results, or whether the claimed “training results” and “candidate models” refer to a distinct different elements. Accordingly, one of ordinary skill in the art wouldn’t be reasonable apprised of the scope of the claim. For examination, the claimed terms interpreted as referring to the same elements. The claim recites the limitations: “ receiving other training results from at least one of the other clients ;” lines 10, and further recites “ performing model consensus on all training results, which are composed of the training result and the other training results, ” and “ checking accuracy of all candidate models, which are generated by the participating clients by training the global model, based on the test set, ” which introduces inconsistency in the claim scope. Specifically, the recited “ at least one ” language suggest that the received set to consist of a few as a single other client’s result, where the “ all training results ” defines the training results and other training results. However, the recitation of “ all candidate models ” implies the complete set of candidate models generated by all participating clients. Thus, it is unclear whether the recited consensus and accuracy checking operation performed on only the subset of training results (i.e., training results and other training results defined as all training results) or on the complete network results (i.e., all candidate models). For examination purposes, the “ all candidate models ..generated by the participating clients ” interpreted broadly as all candidate models available within the consensus process (those corresponding to the training results and other training results). The claim recites “ performing model consensus on all training results, ... to generate a consented model according to a predefined consensus algorithm ” lines 9-11, which defines the output of the consensus operations as a “consented model.” However, the claim further recites “ performing the model consensus with a candidate model with highest accuracy of all the candidate models .” First, the relationship between the last step and the earlier recitation that the consensus “ generates a consented model ” is unclear. It is unclear whether the recited step constitutes a second consensus process, a modification of the earlier consensus operation, or a selection step. Accordingly, it is unclear whether the candidate model having the highest accuracy is intended to refer to the consented model defined in the earlier limitation, or whether an additional, separate consensus operation is performed using the highest accuracy candidate model as the input. Therefore, the scope of the claim is unclear and one of ordinary skill in the art wouldn’t be appraised by the scope of the claimed invention. For examination, the claimed “ consented model ” and the limitation reciting “ performing the model consensus with a candidate model with highest accuracy of all the candidate models ” are interpreted broadly as the result of performing consensus and all candidate models. For at least the above reasons, claim 1 does not particularly point out and distinctly claim the invention and it therefore indefinite under 35 USC § 112(b). Regarding Currently Amended Claim 8 , Claim 8, which depends from parent claim 1, refers to the same limitation defined in claim 1 reciting “ wherein performing the model consensus on all the training results includes ”, with a consensus mechanism that differs from the mechanism already recited in claim 1. Claim 1 specifies performing the model consensus by selecting any one of the participating clients as a leader according to a round robin method and using test data existing in the leader as a test set, whereas claim 8 specifies configuring a group of clients, each with a respective test set including noise, and selecting the candidate having the highest aggregated accuracy of all the candidate models. It is unclear whether claim 8 is intended to introduce a different embodiment that contradicts the embodiment recited in claim 1, or whether claim 8 is intended to define an alternative consensus operation performed by a different predefined consensus algorithm. Additionally, it is not clear whether claim 8 in combination with claim 1 requires both mechanisms performing the same operation (i.e., performing model consensus). Because the claim does not recite the two mechanisms as separate alternatives (i.e., separate consensus algorithms or options), and because claim 8 inherit the limitations of claim 1 by its dependency, one of ordinary skill in the art cannot determine with reasonable certainty whether claim 8 requires the consensus mechanism of claim 1 in combination with claim 8, the mechanism of claim 8 only, or both. Regarding Currently Amended Independent Claim 9 , the claim recites substantially similar limitations as those of claim 1 and is rejected for similar reasons and rationale. Regarding Currently Amended Dependent Claims 2-4, 8, 10-12, and 16 , which depend from independent claims 1 and 9, inherit the deficiencies of their respective parent claim. In view of the above, the Examiner respectfully requests that Applicant thoroughly review the claims for compliance with the requirements set forth under 35 U.S.C. § 112. Appropriate correction is required. Claim Rejections - 35 USC § 103 07-20-aia AIA 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. 07-23-aia AIA The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows: 1. Determining the scope and contents of the prior art. 2. Ascertaining the differences between the prior art and the claims at issue. 3. Resolving the level of ordinary skill in the pertinent art. 4. Considering objective evidence present in the application indicating obviousness or nonobviousness. 07-21-aia AIA Claim (s) 1-2, 4, 9-10, and 12 are rejected under 35 U.S.C. 103 as being unpatentable over Li et al., (NPL: "A Blockchain-based Decentralized Federated Learning Framework with Committee Consensus." (2020)) in view of Qi et al., (NPL: "High-quality model aggregation for blockchain-based federated learning via reputation-motivated task participation." (March, 2022)), and further in view of Kim et al., (NPL: "FedCC: federated learning with consensus confirmation for Byzantine attack resistance (student abstract)." (June, 2022)) . Regarding Currently Amended Claim 1, Li discloses the following: A client model training method performed by a client among a plurality of participating clients in a decentralized learning environment, the client model training method comprising: ( Li , [Abstract] “we proposed a decentralized federated learning framework based on blockchain, i.e., a Blockchain-based Federated Learning framework with Committee consensus (BFLC).” [Pp. 2-3, Section: III] “Federated Learning (FL) enables the machine learning algorithms training across multiple distributed clients without exchanging their data samples. In the original FL settings, one centralized server takes control of the training process, including client management, global model maintenance, and gradient aggregation. During each training round, the server broadcasts the current model to some participating nodes. After receiving the model, nodes locally update it with their local data and submit the update gradients to the server. The server then aggregates and applies the local gradients into the model for the next round. The decentralized nature of blockchain can replace the place of the central server. As aforementioned, the functions of the centralized server can be implemented by the Smart Contract (SC) instead, and be actuated by transactions on the blockchain. To tackle this vision, we propose BFLC, which is a Blockchain-based Federated Learning framework with Committee consensus. Without any centralized server, the participating nodes perform FL via blockchain, which maintains the global models and local updates.”) generating a candidate model by training a global model; ( Li , [P. 3, Section: III] “Fig. 1. The training process of the proposed BFLC framework. (1) Training nodes acquire the newest global model and perform local training.” [Pp. 2-3, Section: III] “In the beginning, a randomly initialized model was placed into the #0 block, then the 0-th round of training starts. Nodes access the current model and execute local training, and put the verified local gradients to new update blocks.” [P. 4, Section: III-C] “Nodes other than committees perform local training each round. In FL, for the sake of security and privacy, raw data will be kept in nodes locally, and these nodes only upload the gradients to the blockchain. ... Nodes can actively obtain the current global model at any time and perform local training. The gradients will be sent to the committee and be validated.”) [ Examiner’s Note : Li teaches participant nodes (i.e., clients) locally training the model using their data (i.e., generating a candidate model).] transmitting a training result of the client for sharing the candidate model to the other clients of the participating clients; ( Li , [Pp. 3-4, Section: III-B] “Considering the computation and communication cost of consensus, we propose an efficient and secure Committee Consensus Mechanism (CCM) to validate the local gradients before appending it to the chain. Under this setting, a few honest nodes will constitute a committee in charge of verification of local gradients and blocks generation. In the meantime, the rest nodes execute local training and send the local updates to the committee.” [P. 4, Section: III-C] “Nodes can actively obtain the current global model at any time and perform local training. The gradients will be sent to the committee and be validated. When eligible updates are packaged on the blockchain, as a reward, tokens can be attached to them.” Fig. 1. The training process of the proposed BFLC framework. (1) Training nodes acquire the newest global model and perform local training. (2) Training nodes send local updates to committee (3) Committee validate the updates and record new model or updates onto blockchain.) [ Examiner’s Note : BFLC describes the transmission to committee/other participant nodes through round-based k-update, the transmitted local updates represents the training result (i.e., candidate models gradients).] receiving other training results from at least one of the other clients; ( Li , [P. 4, Section: III-B] “In the meantime, the rest nodes execute local training and send the local updates to the committee. The committee then validates the updates and assign a score on them.”) [ Examiner’s Note : committee members receive local updates (training results) from other training nodes (i.e., other clients). The committee validates these received updates. All update blocks appended, where each participating node (including committee members performing consensus). Thus, other participating nodes (i.e., other clients) receive updates (results).] performing model consensus on all training results, which are composed of the training result and the other training results to generate a consented model according to a predefined consensus algorithm; ( Li , [P. 3, Section: III-A] “When there are continuously enough update blocks, the smart contract triggers the aggregation, and a new model of the next round is generated and placed on the chain.” [Pp. 3-4, Section: III-B] “The competition-based consensus mechanisms append blocks on the chain first, whereafter, the consensus meets. Conversely, the communication-based generate mechanisms reach an agreement before appending blocks. Considering the computation and communication cost of consensus, we propose an efficient and secure Committee Consensus Mechanism (CCM) to validate the local gradients before appending it to the chain. Under this setting, a few honest nodes will constitute a committee in charge of verification of local gradients and blocks generation. In the meantime, the rest nodes execute local training and send the local updates to the committee. The committee then validates the updates and assign a score on them. Only the qualified updates will be packed onto the blockchain. At the beginning of the next round, a new committee is elected basing on the scores of nodes in the previous round, which means that the committee will not be re-elected. It is noteworthy that the update validation is a pivotal component of the CCM, therefore, we describe a feasible approach: the committee members validate the local updates by treating their data as a validation set, and the validation accuracy becomes the score. This is the minimized approach that acquires no further operation of the committee, but only the basic ability to run the learning model. After combining the scores from the various committee members, the median will become the score of this update.” [P. 4, Section: III-C] “As aforementioned, a certain number of valid updates are required for each round. Therefore, when the committee validates enough local updates, the aggregation process is activated. These validated updates are aggregated by the committee into a new global model. The aggregation can be performed on the local gradients [11] or the local models [12], and the network transmission consumptions of these two methods are equal. After the new global model is packed on the blockchain, the committee will be elected again, and the next training round begins.”) [ Examiner’s Note : The CCM is the model consensus algorithm based on model validation and the new updated model packed on the blockchain becomes the next new model (i.e., consented model).] performing an update of the global model based on the consented model, ( Li , [P. 3, Section: III-A] “When there are continuously enough update blocks, the smart contract triggers the aggregation, and a new model of the next round is generated and placed on the chain.” [P. 4, Section: III-C] “After the new global model is packed on the blockchain, the committee will be elected again, and the next training round begins.”) [ Examiner’s Note : The newly consented/aggregated model is provided to the blockchain and becomes the updated global model for the next round.] wherein performing the model consensus on all the training results includes: ... configuring test data existing in the leader as a test set; ( Li , [Pp. 3-4, Section: III-B] “the committee members validate the local updates by treating their data as a validation set, and the validation accuracy becomes the score.”) [ Examiner’s Note : the committee members (i.e., leader) validate the local updates (i.e., candidate models) by using the local validation/test set against these models to determine the accuracy.] checking accuracy of all candidate models, which are generated by the participating clients by training the global model, based on the test set; ( Li , [Pp. 3-4, Section: III-B] “the committee members validate the local updates by treating their data as a validation set, and the validation accuracy becomes the score.”) [ Examiner’s Note : the committee members (i.e., leader) validate the accuracy of the local updates (i.e., candidate models) using their validation dataset and assigning score.] While Li describes the process of local training of nodes within a decentralized environment through each round and uses a committee consensus mechanism performing validation to determine the next updated global model. The proposed BFLC framework uses validation-accuracy score to determine which candidate local models will be used as the aggregated/consented model in the next round as the global model. Additionally, the BFLC use committee members as the leader performing the model consensus. However, Li does not appear to explicitly teach: selecting any one of the participating clients as a leader according to a round robin Method; configuring test data existing in the leader as a test set; performing the model consensus with a candidate model with highest accuracy of all the candidate models. However, Li in view of Qi teaches the following: selecting any one of the participating clients as a leader according to a round robin Method; ( Qi , [P. 5, Section: III-C] “ModelAggregation(): Model aggregators can aggregate local models into global models by calling this interface. The model aggregation organization internally decides an aggregator for the model aggregation in this round with a round-robin manner... The model aggregator responsible for model aggregation can pack the global model, all local models, and data contributions into a new block by calling this interface after the model aggregation.” [P. 6, Section: IV-B] “The model aggregation organization takes turns to select one of the model aggregators to carry out this round of reputation updates, e.g., the node m is selected at the tth round to update λm,n. The updated reputation values are recorded in the blockchain through the model aggregation organization.”) [ Examiner’s Note : The paper describes the use of round-robin method for selecting a participating node (i.e., model aggregator) as the leader for each round of consensus.] configuring test data existing in the leader as a test set; ( Qi , [P. 3, Section: III-B] “In BFL, a part of data is not used for training but for testing. The client nodes holding the test data can act as model aggregators.” [Pp. 5-6, Section: IV-A] “After acquiring all the local models, the model aggregator uses the owned data set to test the local models. The fair-value game [10], a loss-based marginal approach, is applied to test the quality of local models.”) [ Examiner’s Note : The selected model aggregator (leader) uses its own data as the test data to evaluate reputation of the local models.] performing the model consensus .... with highest accuracy of all the candidate models. ( Qi , [Pp. 5-6, Section: IV-A] “After acquiring all the local models, the model aggregator uses the owned data set to test the local models. The fair-value game [10], a loss-based marginal approach, is applied to test the quality of local models. Let A m , n t denote the evaluation result of model aggregator m on data owner n , i.e., A m , n t = G ( w t ) - G ( w - n t ) (6).”) [ Examiner’s Note : The fair-value-game test uses accuracy-based evaluation of all local models, where the evaluation results is used in the reputation weighted model aggregation as described in section (V).] As shown above, Li and Qi are from the same field of endeavor and discloses a related subject matter (i.e., decentralized federated learning framework). Accordingly, at the effective filing date, it would have been prima facie obvious to one ordinarily skilled in the art of machine learning to modify the combination of Li and Qi to incorporate the blockchain-based federated learning (BFL) with a reputation mechanism as taught by Qi. One would have been motivated to make such a combination in order to improves the reliability and security of federated learning (Qi [Section: VII]). Li in view of Qi does not appear to explicitly teach: performing the model consensus with a candidate model with highest accuracy of all the candidate models. However, it would have been obvious in view of Kim. Hereinafter, Kim, in combination with Li and Qi, teaches: performing the model consensus with a candidate model with highest accuracy of all the candidate models. ( Kim , [Pp. 5-6, Section: IV-A] “The l clients compare test accuracies of models updated with the w r ~ and a previous weight w r - 1 using their own training data, then the consensus clients transmit the confirmation results to the server. Fourthly, the server confirms and decides a global weight wr. If the majority of the consensus clients submits the com paring results that the w r - 1 has higher test accuracy than the w r ~ , the server defines the w r to the w r - 1 .”[ Examiner’s Note : Kim teaches the concept of performing model consensus by comparing test accuracies of the local models and selecting the one with highest accuracy result as confirmed global model (i.e., consented model).] Li, Qi, and Kim are from the same field of endeavor because their disclosure generally relates to Federated Learning with Consensus. Accordingly, it would have been obvious to a person having ordinary skill in the art, before the effective filing date of the claimed invention, having the combination of Li, Qi, and Kim before them, to incorporate the consensus confirmation method as taught by Kim. One would have been motivated to make such a combination in order to reduces the impact of Byzantine attackers in clients on global learning performance degradation. Doing so would provide attack resistance and learning performance (Kim [P. 2, Col. 2]). Regarding Currently Amended Claim 2, the combination of Li, Qi, and Kim teaches the elements of claim 1 as outlined above, and further teaches: wherein, in generating the candidate model by training the global model, an i-th candidate model is generated through a training process for a (i−1)-th (i is a natural number) consented model that was lastly generated in the performing model consensus on all training results, where i is a natural number. ( Li , [p. 3, Section: III-A] “In the beginning, a randomly initialized model was placed into the #0 block, then the 0-th round of training starts. Nodes access the current model and execute local training, and put the verified local gradients to new update blocks. When there are continuously enough update blocks, the smart contract triggers the aggregation, and a new model of the next round is generated and placed on the chain. ... We denote the number of required updates for each round as k, and denote the number of rounds as t = 0, 1, .... Then we have: the # t×(k+ 1) block contains the model of t-th round, which is called model block, and the # [t×(k+1)+1,(t+1)× (k+1)−1] blocks contain the updates of t-th rounds, which are called update blocks.” [p. 4, Section: III-C] “These validated updates are aggregated by the committee into a new global model. ... After the new global model is packed on the blockchain, the committee will be elected again, and the next training round begins.”) Regarding Currently Amended Claim 4, the combination of Li, Qi, and Kim teaches the elements of claim 1 as outlined above, and further teaches: transmitting the other training results to clients of other clients that have not received at least one of all the training results. (Li’s BFLC framework stores local updates (training results) on the blockchain, and the blockchain propagates/broadcast blocks to all participating nodes. Additionally, Qi teaches that each model aggregator collects local models (training results), and then shares those collected models with the other model aggregators in the organization. It is noted that the model aggregators defined as participating nodes. Qi [Pp. 3-4 Section: III-B] “A model aggregator collects the local models of the data owners bound to it via peer-to-peer transfer. In order to get all local models, the model aggregators will share their collected models with each other. Blockchain serves as a bridge to share global models and reputation.” [P. 5, Section: III-C] “During a local model training cycle Tlocal, the model aggregators call this interface to collect the local models. At the end of Tlocal, the local models collected by each other will be shared within the model aggregation organization.”) Regarding Currently Amended Claim 9, The claim recites substantially similar limitation as corresponding claim 1 and is rejected for similar reasons as claim 1 using similar teachings and rationale. Claim 1 is directed to a method, and claim 9 is directed to a device. The combination of Li, Qi, and Kim also discloses a BFLC blockchain system including storage for decentralized local training on participant devices. Regarding Currently Amended Claim 10 , The claim recites substantially similar limitations as corresponding claim 2 and is rejected for similar reasons as claim 2 using similar teachings and rationale. Regarding Currently Amended Claim 12 , The claim recites substantially similar limitations as corresponding claim 4 and is rejected for similar reasons as claim 4 using similar teachings and rationale . 07-21-aia AIA Claim (s) 3 and 11 are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Li, Qi, and Kim as described above, and further in view of Zhang et al., (Pub. No.: US 20220209963 A1) . Regarding Currently Amended Claim 3, the combination of Li, Qi, and Kim teaches the elements of claim 2 as outlined above, and further teaches: While t he combination of Li, Qi, and Kim teaches the model training and update for each round including a blockchain headers of the participating nodes. The combination of Li, Qi, and Kim does not appear to explicitly suggest: computing a hash value of the (i-1)-th consented model using a hash function; and including the computed hash value in training metadata to identify the (i-1)-th consented model as a basis for generating the i-th candidate model. However, it would have been obvious in view of Zhang. Hereinafter, Zhang, in combination with Li, Qi, and Kim, teaches: computing a hash value of the (i-1)-th consented model using a hash function; and including the computed hash value in training metadata to identify the (i-1)-th consented model as a basis for generating the i-th candidate model. ( Zhang , [0025] “In an embodiment, the model transaction includes at least following parts: a transaction basic part including a transaction hash value, a timestamp, a signature and a public key; a model and proof of work part including a model hash value, a model evaluation and a Merkel root of a model training set; and a voting part including a hash value of the previous block and a Merkel root of the data set. [0026] a block includes a hash value of the block, a hash value of the previous block, a hash value of a model of the block, a timestamp, and a Merkel root of a voting result.” [0081] “The consistent block contains the optimal model hash value, which can be used as a basis for updating all nodes.” [0142] “After each round of block generation, an optimal model is selected (that is, the hash value of the model in FIG. 7). In the next round of model training, the node will refer to the training result of the previous block to adjust the model and restart training... the training time of the optimal model for each block is 391 s, 471 s, 420 s, 515 s, and 605 s, respectively. The average time for model verification is 4.52 s, and the average round of voting for consensus takes 90.40 s in total.” [0121-0121] “Operation S220, performing a hash calculation on the hash values of the leaf nodes in pairs to generate a first sub-hash value. Operation S230, performing a hash calculation on the first sub-hash values in pairs to generate the Merkel root of the voting result, the leaf node of the Merkel tree of the voting result includes a Merkel root of the valid vote.”) [ Examiner’s Note : Zhang defines and determines a hash value of the block’s model using hash algorithm, and the transaction metadata includes the prior block’s hash value, the prior block is defined to be used in the next round (i.e., selected optimal model).] Accordingly, it would have been obvious to a person having ordinary skill in the art, before the effective filing date of the claimed invention, having the combination of Li, Qi, Kim, and Zhang before them, to incorporate the method/system for using hash value for blockchain and machine learning models training as taught by Zhang. One would have been motivated to make such a combination in order to greatly reduce the time-consuming model verification and improve the consensus efficiency (Zhang [0143]). Regarding Currently Amended Claim 11 , The claim recites substantially similar limitations as corresponding claim 3 and is rejected for similar reasons as claim 3 using similar teachings and rationale . 07-21-aia AIA Claim (s) 8 and 16 are rejected under 35 U.S.C. 103 as being unpatentable over the combination of Li, Qi, and Kim as described above, and further in view of Lu et al., (IDS: “Blockchain and Federated Learning for Privacy-Preserved Data Sharing in Industrial IoT." (2019)) . Regarding Currently Amended Claim 8, the combination of Li, Qi, and Kim teaches the elements of claim 1 as outlined above, and further teaches: Li Further teaches: configuring a group including clients, among the participating clients, which generated candidate models by training the global model; ( Li , [Pp. 3-4, Section: III-B] “we propose an efficient and secure Committee Consensus Mechanism (CCM) to validate the local gradients before appending it to the chain. Under this setting, a few honest nodes will constitute a committee in charge of verification of local gradients and blocks generation. In the meantime, the rest nodes execute local training and send the local updates to the committee.” [Pp. 3-4, Section: III-B] “At the beginning of the next round, a new committee is elected basing on the scores of nodes in the previous round,... based on the validation scores, the corresponding nodes with better performance will be elected by the smart contract and constitute the new committee for the next training round.”) [ Examiner’s Note : Li discloses the process of configuring a committee (Group) from the participating client nodes, where the committee members (i.e., group) are used to validate the local gradients of the nodes (i.e., generated candidate models by training the global model).] calculating accuracy of each of the candidate models on the test set; ( Li , [Pp. 3-4, Section: III-B] “the committee members validate the local updates by treating their data as a validation set, and the validation accuracy becomes the score... After combining the scores from the various committee members, the median will become the score of this update.”) [ Examiner’s Note : Li discloses that each committee member validates the local updates (i.e., candidate models) using their own validation/test set.] Qi Also teaches: configuring a group including clients, among the participating clients, which generated candidate models by training the global model; ( Qi , [P. 3, Section: III-B] “Model Aggregator: In BFL, a part of data is not used for training but for testing. The client nodes holding the test data can act as model aggregators. Suppose that there are M model aggregators, who form the model aggregation organization denoted by M = {1,2,...,M}. Members of the model aggregation organization collabo rate with each other and are jointly responsible for model aggregation.” [P. 4, Section: III-B] “In the reputation layer, one data owner first shares its local model to all model aggregation nodes through the blockchain layer. Each model aggregation node performs model quality testing of the local model. By exchanging the test results, the model aggregation organization will collaboratively complete a reputation evaluation of this data owner and finally publish the reputation evaluation to blockchain.”) [ Examiner’s Note : Qi discloses that the group (i.e., model aggregation organization) is used to collectively evaluate candidate models (i.e., local models) generated by the participating client nodes.] calculating accuracy of each of the candidate models on the test set; ( Qi , [Pp. 5-6, Section: IV-A] “After acquiring all the local models, the model aggregator uses the owned data set to test the local models. The fair-value game [10], a loss-based marginal approach, is applied to test the quality of local models. Let A m , n t denote the evaluation result of model aggregator m on data owner n , i.e., A m , n t = G ( w t ) - G ( w - n t ) (6)... G(·) is the function for model accuracy measurement.”) [ Examiner’s Note : Qi discloses that each model aggregator in the organization determines the accuracy (i.e., quality score) for local models using their own test data.] performing the model consensus by selecting the candidate model with the highest aggregated accuracy across all test sets of the group as the consented model. ( Qi , [P. 6, Section: IV-B] “Since each model aggregator has different test data, BFL gives a comprehensive reputation evaluation of the data owners by multiple model aggregators... Model aggregator m collects the local reputation evaluations of all model aggregators for the data owner n, thus generating an indirect reputation evaluation of node m for node n by (see equation (10)).. Finally, model aggregator m can obtain the tth global reputation evaluation of node n ...” [P. 7, Section: V-A] “a reward allocation algorithm based on reputation-weighted contribution is designed. If malicious node n exaggerates its contribution, then BFL will give it a relatively poor reputation evaluation based on the reputation mechanism introduced in Section IV. The poor reputation evaluation, as the weight for reward allocation, will decrease the reward for n.” [P. 9, Section: V-C] “Suppose that data owner n is a malicious node, it has a limited and decreasing corrupting effect on the global model. Proof: Since model aggregation employs reputation to weight the reported contributions, the impact of n’s malicious behavior on model aggregation is reduced because of λn < 1 and s n < λ n s n ... That is, as the training continues, λnsn has a decreasing share in aggregating the model until it exits the task.”) [ Examiner’s Note : the reputation-weighted aggregation across multiple model aggregators (i.e., group) using their test sets would correspond to the “ aggregated accuracy across all test sets of the group. ”] While Li in view of Qi teaches committee members performing consensus mechanism to validate the local models using validation dataset and determining the qualified updates that will be packed onto the blockchain based on validation accuracy. While Li in view of Qi define the local model being evaluated including noise. However, Lu, in combination with Li, Qi, and Kim, teaches: configuring a group including clients, among the participating clients, which generated candidate models by training the global model; ( Lu , [P. 3, Section: II-B] “We regard these parties as committee nodes, which are responsible for driving the consensus in permissioned blockchain. Then, the committee nodes train a global data model M jointly by federated learning.”) configuring, by each of the clients in the group independently, a respective test set including noise; ( Lu , [P. 6, Section: III-D-1] “To protect data privacy during multiparty decentralized learning, we incorporate a differential privacy preserved mechanism into federated learning... The training process over dis tributed data owners is shown in Fig. 6. The related parties {P1,P2,...,Pn} are selected through multiparty retrieval in blockchain. The textual data they hold is transferred into normalized graph vectors Vecg = v1,v2,...,vk,e11,e12,...,ekk . When there is a data sharing request R (including a series of queries), Pi will train a local data model mi based on Vecgi toward the request first. The Vecgi, composed of sensitive terms and weights, is used to train a local model in the federated learning process. Since the local model will be shared to other participants, to protect the privacy of Vecg i, we incorporate differential privacy in the learning phase to train a mˆi from noised data. Then, Pi will send model mˆi to other participants. Once mˆi is received,Pi+1 will train a new local data model mˆi+1 based on received mˆi and its local data, then broadcast mˆi+1 to other participants. The data models are trained iteratively among participants. ... Differential private local model training: The noise calibrated by sensitivity s is added to local data Veci. The local data model mˆi is trained locally at Pi, by using machine learning algorithm on the selected noisy data Veci. 3) Collaborative multiparty learning: The Laplace mechanism is applied on local data model mi to achieve differential privacy mˆi = mi + Laplace(s/) (4) where s is the value of sensitivity, as shown in (Eq (5)). Then, the noise-added model mˆi is broadcasted as a transaction of the blockchain to other participants for federated learning.”) calculating accuracy of each of the candidate models on the test set; performing the model consensus by selecting the candidate model with the highest aggregated accuracy across all test sets of the group as the consented model. ( Lu , [P. 7, Section: III-D-2] “Since each committee node trains a local data model, the quality of the model should be verified and measured during the consensus process. We leverage prediction accuracy to quantify the performance of the trained local model. More specifically, in the classification during training, the accuracy is denoted by the fraction of correctly classified records. While in the task of regression, the accuracy is measured by mean absolute error (MAE) ... where f(xi) is the prediction value of model mi and yi is the real value of the records. The lower the MAE of model mi is, the higher the accuracy of mi will be. ... During responding to a data sharing request, a committee node Pi transmits its trained model mi to the next committee node. The transmissions are recorded as model transactions t m i , together with its MAE(mi). ... A committee node Pj collects all model transactions and stores them locally as candidate blocks. ... Each verifying node calculates the MAE(mi) for each model transaction and MAE(M). If the calculated MAE is within a certain range, an approval will be sent to the leader. If the block containing all transactions is approved by every committee node, the leader will send the block data signed with its signature to all nodes.”) Accordingly, it would have been obvious to a person having ordinary skill in the art, before the effective filing date of the claimed invention, having the combination of Li, Qi, Kim, and Lu, to incorporate the privacy-preserving data sharing mechanism as taught by Lu. One would have been motivated to improve and enable secure data sharing with high efficiency and utility in the consensus protocol (Lu [Section: V]). Regarding Currently Amended Claim 16 , The claim recites substantially similar limitations as corresponding claim 8 and is rejected for similar reasons as claim 8 using similar teachings and rationale. Conclusion 07-40 AIA 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. Any inquiry concerning this communication or earlier communications from the examiner should be directed to SADIK ALSHAHARI whose telephone number is (703)756-4749. The examiner can normally be reached Monday Friday, 9 A.M - 6 P.M. ET. 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, Li Zhen can be reached on (571) 272-3768. 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. /S.A.A./Examiner, Art Unit 2121 /Li B. Zhen/Supervisory Patent Examiner, Art Unit 2121 Application/Control Number: 18/332,974 Page 2 Art Unit: 2121 Application/Control Number: 18/332,974 Page 3 Art Unit: 2121 Application/Control Number: 18/332,974 Page 4 Art Unit: 2121 Application/Control Number: 18/332,974 Page 5 Art Unit: 2121 Application/Control Number: 18/332,974 Page 6 Art Unit: 2121 Application/Control Number: 18/332,974 Page 7 Art Unit: 2121 Application/Control Number: 18/332,974 Page 8 Art Unit: 2121 Application/Control Number: 18/332,974 Page 9 Art Unit: 2121 Application/Control Number: 18/332,974 Page 10 Art Unit: 2121 Application/Control Number: 18/332,974 Page 11 Art Unit: 2121 Application/Control Number: 18/332,974 Page 12 Art Unit: 2121 Application/Control Number: 18/332,974 Page 13 Art Unit: 2121 Application/Control Number: 18/332,974 Page 14 Art Unit: 2121 Application/Control Number: 18/332,974 Page 15 Art Unit: 2121 Application/Control Number: 18/332,974 Page 16 Art Unit: 2121 Application/Control Number: 18/332,974 Page 17 Art Unit: 2121 Application/Control Number: 18/332,974 Page 18 Art Unit: 2121 Application/Control Number: 18/332,974 Page 19 Art Unit: 2121 Application/Control Number: 18/332,974 Page 20 Art Unit: 2121 Application/Control Number: 18/332,974 Page 21 Art Unit: 2121
Prosecution Insights