Prosecution Insights
Last updated: November 20, 2026
Application No. 18/658,041

AUTHENTICATION BASED SECURE FEDERATED LEARNING

Non-Final OA §103§112
Filed
May 08, 2024
Priority
May 09, 2023 — provisional 63/465,089
Examiner
HAILEMARIAM, TIBEBU AYELE
Art Unit
2125
Tech Center
2100 — Computer Architecture & Software
Assignee
Micron Technology Inc.
OA Round
1 (Non-Final)
Grant Probability
Favorable
1-2
OA Rounds

Examiner Intelligence

Grants only 0% of cases
0%
Career Allowance Rate
0 granted / 0 resolved
-55.0% vs TC avg
Minimal +0% lift
Without
With
+0.0%
Interview Lift
resolved cases with interview
Typical timeline
Avg Prosecution
3 currently pending
Career history
3
Total Applications
across all art units
This examiner has no resolved cases yet (career too new); statute-level performance unavailable. The Grant Probability card shows Tech Center averages instead.

Office Action

§103 §112
DETAILED ACTION 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 . Specification The disclosure is objected to because of the following informalities: Paragraph [0029], recites reference characters "114-1, 11-2." In context, "11-2" appears to be a typographical error for the corresponding model-update reference character. The paragraph should be corrected to 114-2. Paragraph [0030], uses reference characters 114 as the "Initial Model." Figure 1 identifies the “Initial Model” as reference character 112. The written description should be amended so those reference characters are used consistently with the original disclosure and drawings. Paragraph [0046] refers to "machine-readable storage medium 499". There is no reference number 499 in the drawings. However, machine-readable storage medium is represented by reference 498 in Figure 4. Reference character 499 should be corrected to 498 if that correction is supported by the application as originally filed. Correct these informalities consistently throughout the specification and drawings. No amendment may introduce new matters. Appropriate correction is required. Drawings Figure 1 is objected to under 37 CFR 1.84(p)(5) because reference characters 108 and 110 appear in the drawing but are not identified in the written description. The specifications and drawings should be amended so that each reference character appearing in the drawings is properly identified in the description. Figure 3 is objected to because the block represented by reference character 332 contains the phrase "Plurality of Edge Device." The phrase should be corrected to "Plurality of Edge Devices." Corrected drawing sheets in compliance with 37 CFR 1.121(d) are required in reply to the Office action to avoid abandonment of the application. Any amended replacement drawing sheet should include all the figures appearing on the immediate prior version of the sheet, even if only one figure is being amended. The figure or figure number of an amended drawing should not be labeled as “amended.” If a drawing figure is to be canceled, the appropriate figure must be removed from the replacement sheet, and where necessary, the remaining figures must be renumbered and appropriate changes made to the brief description of the several views of the drawings for consistency. Additional replacement sheets may be necessary to show the renumbering of the remaining figures. Each drawing sheet submitted after the filing date of an application must be labeled in the top margin as either “Replacement Sheet” or “New Sheet” pursuant to 37 CFR 1.121(d). If the changes are not accepted by the examiner, the applicant will be notified and informed of any required corrective action in the next Office action. The objection to the drawings will not be held in abeyance. Claim Objections Claim1 and Claim 8 are objected to because of the following informalities: Claim 1 recites "each of the plurality of edge device." This seems like a typographical error and should be read "each of the plurality of edge devices." Claim 8, which depends on Claim 7, recites "a federated ANN model". However, claim 7 has already introduced a federated ANN model. The claim language is ambiguous as to whether or not the ‘federated ANN model’ of Claim 8 refers to the same ‘federated ANN model’ of Claim 7. If Claim 8 is intended to refer to the same federated ANN model, the phrase should be revised to "the federated ANN model." Claims 2-6 and 9 are also objected since they are either directly or indirectly dependent on the objected Claims 1 and 8 respectively as discussed above. Appropriate correction is required. Claim Rejections - 35 USC § 112 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. Claim 19 is rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA ), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention. The term “more efficiently” in Claim 19 is a relative term which renders the claim indefinite. The term “more efficiently” is not defined by the claim, the specification does not provide a standard for ascertaining the requisite degree, and one of ordinary skill in the art would not be reasonably apprised of the scope of the invention. Claim Rejections - 35 USC § 103 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. 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. This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention. Claims 1, 4 - 17 are rejected under 35 U.S.C. § 103 as being unpatentable over Beser (US 2019/0012592 A1) in view of Xu (US 2021/0143987 A1). Regarding Claim 1, Beser teaches a method, comprising: receiving respective encrypted updates (see e.g. para [0036], “When an update from the first local model is available, the first artificial neural network (ANN1) 12 uploads the first update to the central sever 18, using authentication and encryption”) to an initial artificial neural network (ANN) model from each of a plurality of edge devices that have trained the initial ANN model ( see e.g. para [0024], “the shared central (e.g., global/seed) model is securely trained under the control of the central server from the federation of participating ANNs.”)wherein the initial ANN model is configured to cause each of the plurality of edge device(s) to perform a function; (see e.g., para. [0022] “this central algorithm is therefore commonly applied to multiple ANNs in a federation …the local ANNs draw local inferences from local datasets, such as by using a local inference neural engine (LINE) as part of an ANN”.) wherein the respective encrypted updates are each encrypted ... (see e.g., para. [0025], “messages are sent between the central server and the ANNs, they are, in various embodiments, authenticated and encrypted using cryptographic protocols”, see e.g., para. [0036], Fig. 1, “the first artificial neural network (ANN1) 12 uploads the first update to the central sever(server) 18, using authentication and encryption…the second artificial neural network (ANN2) 14 uploads the second update to the central sever(server) 18, using authentication and encryption” and see e.g., para. [0026], “a private key to each individual ANN of the federation”). determining whether each respective encrypted update is authentic; (see e.g., para. [0025], “bi-directional authentication (e.g., via authentication tokens, digital signatures, identity certificates, etc.) ensures communication is only between the central server and particular ANNs within a federation”; and see e.g., para. [0027], “metadata is also used as an additional security measure between the ANNs and the central server” also see e.g., para. [0036], Fig. 1, “the first artificial neural network (ANN1) 12 uploads the first update to the central sever 18, using authentication and encryption”) Beser uses see e.g., para. [0025] “authentication tokens, digital signatures, identity certificates” to determine the authentication of the updates. aggregating each authenticated update into a federated ANN model configured to cause each of the plurality of edge devices to perform the function; and (see e.g., para. [0025], “bi-directional authentication”; and see e.g., para. [0022], “the central algorithm aggregates the collected updates into the central algorithm …the central server then forms a new, modified central algorithm, obtained through federated learning”, and see para. [0036], Fig. 1, “the central server 18 aggregates the updates into the central model for distribution back to the first artificial neural network (ANN1) 12 and to the second artificial neural network (ANN2) 14 of the first federation (FED1) 20”). Further the reference discloses that the resulting model is applied to participating ANNs to inference operations (to perform a function) see e.g., paras. [0022], “this central algorithm is therefore commonly applied to multiple ANNs in a federation …ANNs draw local inferences from local datasets”). deploying the federated ANN model to at least one of the plurality of edge devices. (see e.g., Beser, para. [0022], “the central server then forms a new, modified central algorithm, obtained through federated learning, which it then again distributes to the ANNs of the federation”; and further see e.g., para. [0036], Fig. 1, “the central server 18 aggregates the updates into the central model for distribution back to the first artificial neural network (ANN1) 12 and to the second artificial neural network (ANN2) 14 of the first federation (FED1) 20”). However, Beser does not expressly disclose using a respective edge device-specific identifier. In the same field of endeavor, Xu teaches individual key assigned by the Key Server (see e.g., para. [0079], Fig. 6, “The Unique Key 655 corresponds to an individual key assigned by the Key Server 105 …encrypt them using the Unique Key 655”). Xu further discloses assigning individual encryption keys to participating devices (see e.g., para. [0083], Fig. 7, block 720, “the Key Server 105 generates and transmits a unique individual key to the selected participant”). Xu teaches encrypting respective model updates using cryptographic keys assigned to individual participating devices, corresponding to the claimed respective edge device-specific identifier under the broadest reasonable interpretation consistent with the application's specification. It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to modify Xu's participant-specific encryption mechanism to edge specific-identifier and use it in Beser's federated ANN system to encrypt respective model updates using individual cryptographic keys assigned to participating edge devices. Such a modification would have improved the confidentiality of transmitted model updates by protecting each device's contribution with its assigned key, as suggested by Xu's disclosure of unique key 655 and encryption component 645 (see e.g., Xu, FIG. 6; FIG. 7, block 720). Regarding Claim 4, Beser discloses aggregating each authenticated update into the federated ANN model without aggregating unauthenticated updates (see e.g., para. [0036], “When an update from the first local model is available, the first artificial neural network (ANN1) 12 uploads the first update to the central sever 18, using authentication and encryption”). Beser discloses that the first, second or third ANNs upload their respective updates using authentication and encryption, see paras [0036-0037], and the central model corresponds to the claimed federated ANN model because it incorporates updates generated by participating ANNs. Server. Beser does not expressly state that updates failing authentication are excluded from aggregation. However, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the claimed invention, to configure Beser’s central server to aggregate each update that passes authentication and exclude any update that fails authentication the reason being preventing unverified sources from being included in the central model. Regarding Claim 5, Beser discloses deploying the federated ANN model to at least one of the plurality of edge devices comprises deploying the federated ANN model to those of the plurality of edge devices from which authenticated updates are received (see e.g. para [0022], “The central server then forms a new, modified central algorithm, obtained through federated learning, which it then again distributes to the ANNs of the federation” and see e.g. para [0025], “When messages are sent between the central server and the ANNs, they are, in various embodiments, authenticated and encrypted using cryptographic protocols”) Beser teaches authenticated and encrypted communication between the central server and particular ANNs and how the ANNs return their local updates to the central server. Then the central server aggregates those updates and distributes back to the participating ANNs. Regarding Claim 6, Xu discloses wherein receiving respective encrypted updates to the initial ANN model (see e.g. para [0059] and Fig. 3, “At blocks 330 and 335, after the local training is completed for a given Participant 120A-N, the Participant 120A-N transmits its encrypted (and potentially noisy) model parameters back to the Aggregator 110. At block 340, the Aggregator 110 receives and stores the encrypted parameters from each Participant 120A-N.”) comprises receiving signals indicative of at least one of an updated weight, bias, and activation function of the initial ANN model in cyphertext (see e.g. para [0058], “the Aggregator 110 can transmit a request for the internal weights of each local model, after a round of training”) and (see e.g. para [0058], “After the requested round of training is complete, at block 320 and 325, each respective Participant 120A-N extracts the relevant and/or requested model parameters from their local model, and encrypts them using their unique public key (e.g., assigned by the Key Server 105”) Regarding Claim 7, Beser discloses an apparatus comprising: a memory device; a processor coupled to the memory device and configured to (see e.g., para. [0039], Fig. 2, “comprises one or more controllers 32 having one or more processors 34 operating in conjunction with one or more tangible, non-transitory memories 36 configured to implement digital or programmatic logic”) …a first encrypted update to an initial artificial neural network (ANN) model received from a first edge device… (see e.g., para. [0036], Fig. 1, “When an update from the first local model is available, the first artificial neural network (ANN1) 12 uploads the first update to the central server 18, using authentication and encryption” and also see e.g., [0025], “messages are sent between the central server and the ANNs, they are, in various embodiments, authenticated and encrypted using cryptographic protocols”) …a second encrypted update to the initial ANN model received from a second edge device… (see e.g., para. [0031], Fig. 1, “the second artificial neural network (ANN2) 14 maintains a second upload connection U2 with the central server 18,” and see e.g., para. [0036], Fig. 1, “and the second artificial neural network (ANN2) 14 then trains on second local data (LD2) 26. When an update from the first local model is available, the first artificial neural network (ANN1) 12 uploads the first update to the central server 18, using authentication and encryption”) …decrypt the first and the second encrypted updates (see e.g., para. [0026], “When an ANN receives an encrypted message from the central server, or vice versa, it uses the ANN's private key to decrypt the message. Thus, while the public key is widely disseminated to the ANNs of the federation, the ANNs keep private their private keys, in various embodiments. As such, these software keys must compare favorably for messages to be decrypted between the central server and ANNs”) aggregate the first and the second decrypted updates into a federated ANN model…; and deploy the federated ANN model to the first and the second edge devices (see e.g., para. [0036], “When the central sever 18 receives the first update from the first artificial neural network (ANN1) 12 and the second update from the second artificial neural network (ANN2) 14, the central server 18 aggregates the updates into the central model for distribution back to the first artificial neural network (ANN1) 12 and to the second artificial neural network (ANN2) 14 of the first federation (FED1) 20.”) However, Beser does not explicitly teach determining whether a first edge device-specific identifier associated with a first encrypted update … is authentic; determining whether a second edge device-specific identifier associated with a second encrypted … is authentic; and in response to determining that the first and the second edge device-specific identifiers are authentic. Also, Beser does not disclose a secure multiparty computation protocol In the same field of endeavor, Kwon teaches: determining whether a first edge device-specific identifier associated with a first encrypted update is authentic (see e.g., [0167], Fig. 8, “identification information, such as identification information of the electronic device” and see e.g., [0168], Fig. 8, “The electronic device may transmit, to the server, federated learning identification information stored in the secure zone together with a federated learning parameter”) compare the received identification information with corresponding identification information registered at the server to determine whether the transmitting electronic device is authenticated (see e.g., para. [0169], Fig. 8, “The server may compare, with the first federated learning identification information received from the electronic device…such as the identification information of the electronic device” also, see e.g., para. [0172], Fig. 8 “thereby identifying that the federated learning data has been received from the electronic device authenticated by the server” Accordingly, Know describes the identification and authentication of data as originating from an authenticated electronic device. The disclosure supports determining authenticity through comparison of device-specific identification information transmitted with a federated learning parameter. The device authentication procedure discussed also applies to determining whether a second edge device-specific identifier associated with a second encrypted update is authentic (also, see e.g., para. Para. [0169], “the second federated learning identification information, such as the identification information of the electronic device, the identification information of the application training the AI model built in the electronic device” see e.g., para. [0172], Fig. 8 “thereby identifying that the federated learning data has been received from the electronic device authenticated by the server”) in response to determining that the first and the second edge device-specific identifiers are authentic (see e.g., para. [0171], “based on (in response to) a result of comparing the first federated learning identification information to the second federated learning identification information …the server may identify(determine) whether the electronic device that transmitted federated learning data is trustable (authentic)” It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to combine Kwon's device-authentication mechanism into Beser's federated ANN system to verify whether the first and second edge device-specific identifiers associated with the respective encrypted updates are authentic. Such a modification would have allowed the central server to decrypt the first and second encrypted updates upon successful authentication of both identifiers, thereby improving the security and reliability of the system by preventing unauthorized devices from contributing updates to the federated ANN model, as suggested by Kwon (para. [0172], FIG. 8). However, Kwon does not expressly disclose performing this aggregation according to the claimed secure multiparty computation protocol. In the same field of endeavor, Xu explicitly teaches a multi-input functional encryption (see e.g., para. [0025], “a secure multiparty computation (SMC) protocol based on functional encryption” and see e.g. para. [0027], Fig. 1, “Multi-Input Functional Encryption (MIFE)” …in which… “an aggregator can compute the average value for the parameters” without learning the individual parameter values contributed by the participants. Additionally, see e.g., para. [0028], Fig. 1, “the aggregator transmits an aggregation vector indicating the desired aggregation function of the decryption”.) It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to combine Xu's secure multiparty computation protocol into Kwon's federated-learning system to protect model parameters received from authenticated devices during aggregation. Such a modification would have provided an additional layer of privacy by allowing the server to aggregate model parameters without disclosing the individual contributions of participating devices, as suggested by Xu (para. [0025]) Regarding Claim 8, Beser discloses wherein the processor is further configured to: (see e.g., para. [0039], Fig. 2, “one or more processors 34 operating in conjunction with one or more tangible, non-transitory memories 36 configured to implement digital or programmatic logic”) ... a third encrypted update to the initial ANN model received from a third edge device (see e.g., para. [0031], Fig. 1, “the third artificial neural network (ANN3) 16 maintains a third upload connection U3 with the central server 18” and see e.g., para. [0037], Fig. 1, “the third artificial neural network (ANN3) 16 uploads the third update to the central sever(server) 18, using authentication and encryption”) aggregate the first and the second decrypted updates into a federated ANN model ... (see e.g., para. [0036], Fig 1, “the central server 18 aggregates the updates into the central model for distribution back to the first artificial neural network (ANN1) 12 and to the second artificial neural network (ANN2) 14” and further discloses decryption of encrypted communications using cryptographic keys see e.g., para. [0026], “it uses the ANN's private key to decrypt the message”). However, Beser does not expressly disclose determine whether a third edge device-specific identifier associated with a third encrypted update …is authentic; and aggregate… without the third encrypted update in response to determining that the third edge device-specific identifier is not authentic. In the same field of endeavor, Kwon teaches verifying device identification information associated with received federated-learning parameters (see e.g., para. [0167], Fig. 8, “identification information of the electronic device”; para. [0168], Fig. 8, “federated learning identification information stored in the secure zone together with a federated learning parameter”). Kwon further teaches comparing received identification information with corresponding information registered at the server to determine whether the transmitting device is authenticated (see e.g., paras. [0169]–[0173], Fig. 8, operations S830 and S850). Regarding ... without the third encrypted update in response to determining that the third edge device-specific identifier is not authentic., Kwon teaches determining whether a received federated-learning result is trustworthy (see e.g., Fig. 2, “the server 20 may identify whether a result of federated learning performed by the electronic device 10 is trustable”. Kwon further discloses excluding untrustworthy parameters from federated learning see e.g., Fig. 2, discussion of operation 270, “the server 20 may remove the federated learning parameter received from the electronic device 10 without reflecting the federated learning parameter to federated learning”). It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate Kwon's device-identification verification and model-protection mechanisms into Beser's federated ANN system to reject updates associated with devices whose identification information fails authentication while continuing to aggregate updates received from authenticated devices. Such a modification would have improved the security and reliability of Beser's federated-learning process by preventing contributions from unauthenticated devices from affecting the resulting ANN model, as suggested by Kwon's teachings of verifying registered device identification information and excluding untrustworthy federated-learning parameters (see e.g., FIG. 8, operations S830 and S850; FIG. 2, operations 250 and 270). Regarding Claim 9 Beser discloses wherein the processor is further configured to deploy the initial ANN model to the first, the second, and the third edge devices. (see e.g., para. [0039], FIG. 2, “one or more processors 34 operating in conjunction with one or more tangible, non-transitory memories 36 configured to implement digital or programmatic logic” and based on that see e.g., para. [0035], Fig. 1, “they download the central server's 18 central model at the first artificial neural network (ANN1) 12 and the second artificial neural network (ANN2) 14, respectively”; and further see e.g., para. [0037], Fig. 1, “they download the central server's 18 central model at the second artificial neural network (ANN2) 14 and the third artificial neural network (ANN3) 16, respectively”). Regarding Claim 10, Beser discloses wherein the processor is configured to encrypt the federated ANN model prior to deploying the federated ANN model to the first and the second edge devices. (see e.g., para. [0039], FIG. 2, “one or more processors 34 operating in conjunction with one or more tangible, non-transitory memories 36 configured to implement digital or programmatic logic” and see e.g., para. [0025], “When messages are sent between the central server and the ANNs, they are, in various embodiments, authenticated and encrypted using cryptographic protocols”. Also see e.g., para. [0026], “the central server uses the central server's public key to encrypt the message”). Regarding deploying the federated ANN model to the first and the second edge devices, Beser discloses (see e.g., para. [0036], Fig. 1, “the central server 18 aggregates the updates into the central model for distribution back to the first artificial neural network (ANN1) 12 and to the second artificial neural network (ANN2) 14 of the first federation (FED1) 20”). Beser further discloses (see e.g., para. [0022], “the central server then forms a new, modified central algorithm, obtained through federated learning, which it then again distributes to the ANNs of the federation”). Thus, Beser teaches generating an aggregated ANN model, encrypting server communications, and distributing the resulting model to participating ANNs. Although Beser does not expressly state that the federated ANN model itself is encrypted before deployment, its disclosure of encrypted server-to-ANN communications provides a basis for transmitting the resulting model in encrypted form. It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to apply Beser's disclosed encryption mechanism to the resulting federated ANN model before transmitting it to the first and second edge devices. Such an implementation would have protected the confidentiality of the aggregated model during transmission and prevented unauthorized recipients from deciphering the model, consistent with Beser's teaching that encrypted communications are decipherable only by the central server and participating ANNs (see e.g., Beser, para. [0025]). Regarding Claim 11 Beser teaches …wherein the processor is configured to encrypt the federated ANN model using the apparatus-specific identifier (see e.g., para. [0039], Fig. 2, “one or more processors 34 operating in conjunction with one or more tangible, non-transitory memories 36 configured to implement digital or programmatic logic” and see e.g., para. [0025], “When messages are sent between the central server and the ANNs, they are, in various embodiments, authenticated and encrypted using cryptographic protocols”. Also see e.g., para. [0026], “the central server uses the central server's public key to encrypt the message” and see e.g., para. [0025], “authenticated and encrypted using cryptographic protocols”) However, Beser does not teach the one or more blocks of the memory device are secured by hardware that isolates access to the one or more blocks and the one or more blocks store an apparatus-specific identifier In the same field of endeavor, Kwon teaches wherein the one or more blocks of the memory device are secured by hardware that isolates access to the one or more blocks (see e.g., para. [0080], Fig. 9, “a memory secure zone encoded based on hardware (central processing unit (CPU)/graphics processing unit (GPU)) such that data is not forged/falsified via external access”). Kwon further discloses that the hardware secure architecture may be implemented as a trusted execution environment (TEE) (see e.g., para. [0080]) wherein the one or more blocks store an apparatus-specific identifier (see e.g., para. [0113], Fig. 3, “the electronic device 10 may store, in the secure zone 18, the federated learning identification information, such as identification information of the electronic device 10”). Kwon teaches storing device-associated identification information in a hardware-protected memory region. However, Kwon does not expressly identify the protected region as the claimed one or more memory blocks having hardware-isolated access or identify the stored information as an apparatus-specific encryption identifier of the federated-learning server. It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to protect cryptographic information used in Beser's federated ANN system by employing Kwon's hardware-secured memory architecture. Such a modification would have reduced the risk of unauthorized access to sensitive cryptographic information, as suggested by Kwon (para. [0080]) Regarding Claim 12, Beser discloses a system, comprising: a plurality of edge devices, each configured to: (see e.g., para. [0029], Fig. 1, “…comprises…a first artificial neural network (ANN1) 12, a second artificial neural network (ANN2) 14, a third artificial neural network (ANN3) 16, and a central server 18”) receive respective data from operation in a respective location; (see e.g., para. [0030], Fig. 1, “located within a single installation, multiple installations, a single location, multiple locations, etc.” and see e.g., para. [0036], Fig. 1, “The first artificial neural network (ANN1) 12 then trains on first local data (LD1) 24, and the second artificial neural network (ANN2) 14 then trains on second local data (LD2) 26”). train an initial artificial neural network (ANN) model with the respective data; (see e.g., para. [0036], Fig. 1, “they download the central server's 18 central model at the first artificial neural network (ANN1) 12 and the second artificial neural network (ANN2) 14, respectively …The first artificial neural network (ANN1) 12 then trains on first local data (LD1) 24, and the second artificial neural network (ANN2) 14 then trains on second local data (LD2) 26”) generate a respective update to the initial ANN model; (see e.g., para. [0036], Fig. 1, “When an update from the first local model is available, the first artificial neural network (ANN1) 12 uploads the first update to the central sever(server) 18 … When an update from the second local model is available, the second artificial neural network (ANN2) 14 uploads the second update to the central sever 18”) encrypt the respective update... (see e.g., para. [0025], “they are, in various embodiments, authenticated and encrypted using cryptographic protocols” and see e.g., para. [0026], “asymmetric cryptography is a cryptographic protocol that uses multiple (e.g., two) keys between receivers and senders…the central server uses the central server's public key to encrypt the message”) transmit the respective encrypted update to a server; (see e.g., paragraph [0036], Fig. 1, “the first artificial neural network (ANN1) 12 uploads the first update to the central sever 18, using authentication and encryption” and “the second artificial neural network (ANN2) 14 uploads the second update to the central sever(server) 18, using authentication and encryption”) and the server, configured to: determine whether the respective encrypted update is authentic; (see e.g., para. [0039], Fig. 2, “one or more processors 34 operating in conjunction with one or more tangible, non-transitory memories 36 configure to implement digital or programmatic logic” and see e.g., para. [0025], “bi-directional authentication (e.g., via authentication tokens, digital signatures, identity certificates, etc.) ensures communication is only between the central server and particular ANNs within a federation” The authentication process determines whether the updates are authentic or not.) decrypt the respective encrypted update in response to determining that the respective encrypted update is authentic; (see e.g., para. [0026], “When an ANN receives an encrypted message from the central server, or vice versa, it uses the ANN's private key to decrypt the message …these software keys must compare favorably for messages to be decrypted between the central server and ANNs” and further see e.g., para. [0025], “bi-directional authentication (e.g., via authentication tokens, digital signatures, identity certificates, etc.) ensures communication is only between the central server and particular ANNs within a federation”). Beser discloses authenticated and encrypted communication between the central server and participating ANNs, including authentication through tokens, digital signatures, or identity certificates (para. [0025]), and decryption of encrypted messages (para. [0026]). “Authenticate then decrypt” means “decrypt in response to authentication.” The [bi-directional] authentication step is to verify the sending participant’s authenticity before decrypting the encrypted update, thereby restricting decryption to updates from unauthenticated participants. aggregate a plurality of decrypted respective updates into a federated ANN model (see e.g., para. [0026], “When an ANN receives an encrypted message from the central server, or vice versa, it uses the ANN's private key to decrypt the message” and see e.g., para. [0036], Fig. 1, “When the central sever 18 receives the first update from the first artificial neural network (ANN1) 12 and the second update from the second artificial neural network (ANN2) 14, the central server 18 aggregates the updates into the central model for distribution back to the first artificial neural network (ANN1) 12 and to the second artificial neural network (ANN2) 14 of the first federation (FED1) 20”) However, Beser does not expressly disclose using edge device-specific identifier to encrypt the respective update. In the same field of endeavor, Xu teaches individual key assigned by the Key Server (see e.g., para. [0079], Fig. 6, “The Unique Key 655 corresponds to an individual key assigned by the Key Server 705…encrypt them using the Unique Key 655” and further see e.g., Figs. 7 & 5, blocks 705, disclosing that key server 105 identifies participating devices and generates unique keys for the respective participants). Xu's individually assigned cryptographic key corresponds to the claimed respective edge device-specific identifier under the broadest reasonable interpretation consistent with the application's specification, which describes the edge device-specific identifier as a key. It would further have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention, to incorporate Xu's participant-specific encryption mechanism into Beser's federated ANN system to encrypt each respective model update using a unique cryptographic key associated with the transmitting edge device. Such a modification would have improved the confidentiality and security of transmitted model updates by providing individualized encryption for each participating device, as suggested by Xu's teaching of encrypting locally generated model parameters using an individually assigned key (see e.g., Xu, FIG. 6, unique key 655 and encryption component 645). Claim 13 is similar to Claim 3 except that it is a system Claim, and it recites aggregation of decrypted updates. It is rejected for substantially the same reasons given to Claim 3. Regarding …aggregate the plurality of decrypted respective updates… (see e.g., para. [0036], Fig. 1, “the central server 18 aggregates the updates into the central model for distribution back to the first artificial neural network (ANN1) 12 and to the second artificial neural network (ANN2) 14 of the first federation (FED1) 20”). Regarding Claim 14, Beser discloses the server is further configured to deploy the federated ANN model to the plurality of edge devices. (see e.g. para [0022], “The central server then forms a new, modified central algorithm, obtained through federated learning, which it then again distributes to the ANNs of the federation”) Beser teaches that after the server aggregates participant updates into a modified central model, the updated model is distributed back to the participating ANNs Claim 15 is substantially similar to Claim 10 except that it is a system Claim, and it is rejected for the same reasons given to Claim 10 Claim 16 is substantially similar to Claim 5 except that it is a system Claim, and it is rejected for the same reasons given to Claim 5 Claim 17 is substantially similar to Claim 6 except that it is a system claim, and it is rejected for the same reasons given to Claim 6 Claims 2, 3 ,18, and 20 are rejected under 35 U.S.C. § 103 as being unpatentable over Beser in view of Xu and further in view of Kwon (US 2022/0237523 A1). Regarding Claim 2, Xu teaches wherein the respective edge device-specific identifier is stored in one or more blocks of a memory device of the respective edge device; and (see e.g., para. [0078], Fig. 6, “the Storage 620 includes a set of local Training Data 650, a Unique Key 655, and a Local Model 125” and see e.g., para. [0079], Fig. 6, “The Unique Key 655 corresponds to an individual key assigned by the Key Server 105 … encrypt them using the Unique Key 655”). Xu teaches storing a participant-specific cryptographic key in a storage device and using the key to encrypt the participant's model parameters. However, Xu does not expressly teach about the one or more blocks of the memory device secured by hardware that isolates access to the one or more blocks In the same field of endeavor, Kwon teaches storing device identification information in a hardware-secured memory region (see e.g., para. [0113], Fig. 3, “the electronic device 10 may store, in the secure zone 18, the federated learning identification information, such as identification information of the electronic device 10” and see e.g., para. [0080], “a memory secure zone encoded based on hardware (central processing unit (CPU)/graphics processing unit (GPU)) such that data is not forged/falsified via external access”). Kwon teaches storing device-associated identification information in a hardware-protected secure memory zone to prevent unauthorized modification through external access. It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention to store Xu's participant-specific cryptographic key in the hardware-secured memory zone taught by Kwon. Such a modification would have improved the security of Beser's federated ANN system by protecting the cryptographic key used to encrypt model updates against unauthorized access and modification, as suggested by Kwon's use of hardware secure architecture to protect federated-learning information as suggested by Kwon see e.g. Kwon para [0019]. Regarding Claim 3, Beser discloses wherein aggregating each respective encrypted update comprises aggregating according to… (see e.g., para. [0036], Fig. 1, “the central server 18 aggregates the updates into the central model for distribution back to the first artificial neural network (ANN1) 12 and to the second artificial neural network (ANN2) 14 of the first federation (FED1) 20” and see e.g., para. [0027] “the central server and ANNs use authentication and encryption to ensure effective, secure communications”). However, Beser does not disclose aggregating according to a secure multiparty computation protocol. In the same field of endeavor, Xu teaches about a secure multiparty computation protocol (see e.g., para. [0025], “for privacy-preserving federated learning employing a secure multiparty computation (SMC) protocol based on functional encryption.” and see e.g., para. [0027], “In MIFE, a functional secret key skƒ can correspond to an n-ary function ƒ (⋅) that takes multiple ciphertexts as input. For example, given a set of encrypted parameters from individual participants and the functional secret key skƒ corresponding to an aggregation vector, an aggregator can compute the average value for the parameters (e.g., corresponding to the inner product of the aggregation vector and the encrypted parameters from individual participants), without learning the actual value provided by any one participant.”) Xu expressly describes a privacy-preserving federated-learning approach using secure multiparty computation based on functional encryption in which a function operates on multiple ciphertexts. Xu's Algorithm 1 further shows participants training local models, encrypting their model information, returning the encrypted information to the aggregator, and the aggregator applying the MIFE function with an aggregation vector to update the global model. It would have been obvious to a person of ordinary skill in the art before the effective filing date to modify Beser’s central-server aggregation of model updates to employ Xu’s secure multiparty computation protocol based on functional encryption. The modification would enable the server to obtain aggregate model parameters while preventing disclosure of individual participants parameter values thereby addressing the privacy risks associated with sharing model updates. Xu expressly identifies this privacy protections reduced communication overhead as suggested by Xu para. [0027]-[0029] Claim 18 is substantially similar to Claim 2 and Claim 11 except that it is a system claim, and it is rejected for the same reasons given to those claims. Regarding Claim 20, Beser discloses the plurality of edge devices comprise(s) drones (see e.g. [0002], “This disclosure relates to neural networks, such as are suitable for artificial intelligence. Such networks are used, for example, in airplanes, automobiles, boats, cameras, computers, data centers, data gathering devices, drones, medical applications, point-of-sale registers, registration set-ups, robots, surveillance applications, trains, trucks, under desks, in workspaces, etc.”); and wherein the memory comprises flash memory (see e.g., para. [0040], ” the computer-readable media 38 contains, maintains, and/or stores computer instructions, programs, and/or software that is used by the one or more controllers 32, including physical media, such as, for example, magnetic, optical, and/or semiconductor media, including, for example, flash, magnetic, and/or solid-state devices”) Beser expressly identifies drones among devices in which the disclosed artificial neural networks may be used. Beser also describes processor/memory components and identifies flash memory among suitable physical storage media. The use of flash memory in a drone-based ANN participant would have been the predictable use of a known nonvolatile storage medium for its ordinary storage function. Claims 19 is rejected under 35 U.S.C. § 103 as being unpatentable over Beser in view of Lal and further in view of Zhang (Zhang et al., IEEE Transaction, vol. 69, pp. 14104-14109, Nov. 2020). Regarding Claim 19, Beser discloses the system, wherein the plurality of edge devices comprises drones (See e.g. para [0002], “This disclosure relates to neural networks, such as are suitable for artificial intelligence. Such networks are used, for example, in airplanes, automobiles, boats, cameras, computers, data centers, data gathering devices, drones, medical applications, point-of-sale registers, registration set-ups, robots, surveillance applications, trains, trucks, under desks, in workspaces, etc.”) the initial ANN model is configured to be executed by the respective processor (See e.g. para [0039] and Fig. 2, “non-transitory memories 36 store instructions that are implemented by the one or more processors 34 for performing various functions,”) to causes the drone to perform a function; and wherein the federated ANN model is configured to be executed by the respective processor to cause the drone to perform the function more efficiently than the initial ANN model. However, Beser doesn’t teach how the drone perform a function or perform more efficiently than the initial ANN model. Zhang, in the same field of endeavor, provides a federated-learning implementation in which the participating (Unmanned Aerial Vehicles) UAVs themselves are the local learners. Each UAV sends the locally acquired model to a ground fusion center and receives the resulting global model for the next round of local training. Zhang reports high classification accuracy at relatively low communication cost. (See Zhang, Fig.3) PNG media_image1.png 578 770 media_image1.png Greyscale Beser, Zhang, and the instant application are all in the same field of endeavor. It would have been obvious to one of ordinary skill in the art, before the effective filing date of the claimed invention to modify the federated learning system of Beser to incorporate Zhang’s teaching of performing local updates from UAVs over multiple communication rounds. Zhang teaches that this configuration achieves high classification accuracy (better performance) at relatively low communication (Zhang Abstract and Fig.3). Conclusion Any inquiry concerning this communication or earlier communications from the examiner should be directed to TIBEBU HAILEMARIAM whose telephone number is (571)272-0221. The examiner can normally be reached M-F 8-4. 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, Kamran Afshar can be reached at (571) 272-7796. 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. /TIBEBU HAILEMARIAM/ Examiner, Art Unit 2125 /KAMRAN AFSHAR/ Supervisory Patent Examiner, Art Unit 2125
Read full office action

Prosecution Timeline

May 08, 2024
Application Filed
Oct 20, 2026
Non-Final Rejection mailed — §103, §112 (current)

Strategy Recommendation AI-generated — please review before filing

Get a prosecution strategy drawn from examiner precedents, rejection analysis, and claim mapping.
Typically takes 5-10 seconds — AI-generated, attorney review required before filing

Prosecution Projections

1-2
Expected OA Rounds
Grant Probability
Low
PTA Risk
Based on 0 resolved cases by this examiner. Grant probability derived from career allowance rate.

Sign in with your work email

Enter your email to receive a magic link. No password needed.

Personal email addresses (Gmail, Yahoo, etc.) are not accepted.

Free tier: 3 strategy analyses per month