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 .
This action is in response to the application filed on 08/20/2026. Claims 1-7 and 13-30 are pending in the case. This action is Final.
Applicant Response
3. In Applicant’s response dated 08/20/2026, Applicant amended Claims 1, 15 and 21 and argued against all objections and rejections previously set forth in the Office Action dated 08/24/2026.
Examiner Comments
6. In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
Claim Rejections - 35 USC § 103
7. 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.
8. Claims 1-7 and 13-30 are rejected under 35 U.S.C. 103 as being unpatentable over Rouhani (Pub: No. US 20210019605 A1, Pub. Date: 2021-01-21) in view REIMANN ( US 20200380123 A1, 2020-12-03) in view of Zheng ( NPL: Title: A DNN Fingerprint for Non-Repudiable Model Ownership Identification and Piracy Detection; Published Date: 2022, VOL. 17), in further view of STAPLETON (Pub: No. US 20200234121 A1, Pub. Date: 2023-07-23)
Rouhani teaches a method comprising:
receiving data characterizing artefacts associated with each of a plurality of layers of a first machine learning model (see Rouhani: Fig. 5A, [0083], “At 502, the detection engine 330 may identify, based at least on an activation map associated with a hidden layer of a first machine learning model, a plurality of input samples that causes the activation functions applied by a plurality of neurons forming the hidden layer to output values occupying one or more low probabilistic regions of the activation map. For instance, the activation map associated with a hidden layer of the machine learning model 100 may enumerate the different output values of the activation functions applied by the neurons forming that layer of the machine learning model.”)
generating, for each layer of the first machine learning model, fingerprints corresponding to each of the artefacts in such layer (see Rouhani: Fig. 5A, [0084], “At 504, the detection engine 330 may embed, in the hidden layer of the first machine learning model, a first digital watermark corresponding to the plurality of input samples by at least training, based on training data that includes the plurality of input samples, the first machine learning model. For example, a digital watermark may be embedded in the machine learning model 100 by at least training the machine learning model 100 based on training data that includes one or more input samples configured to alter the low probabilistic regions of the activation map.”), by computing at least one of : a mean distance of a manifold representing the artefact from a point in space (see Rouhani: Fig.2, [0062], “the mean values of the selected Gaussian distributions are as close to the digital watermark after the embedding of the digital watermark, a second loss term loss.sub.2 shown in Equation (3) below may be added to the loss function associated with the machine learning model 100 which, as noted, may be minimized during the training of the machine learning model 100. The second loss term loss.sub.2 may characterize the distance between the digital watermark present in the machine learning model 100 and the N-bit binary string defined by the owner of the machine learning model 100.”), a fluctuation of the manifold representing the artefact relative to a mean distance (see Rouhani: Fig.1, [0065], “the data movement required to embed the digital watermark in the machine learning model 100 may be minimized by integrating the first loss term loss.sub.1 and/or the second loss term loss.sub.2 into the tensor graph of the machine learning model 100 such that the gradients with respect to the mean values of the Gaussian Mixture Model (GMM) may be computed during regular back-propagation and all computations for embedding the digital watermark may be performed homogeneously on a processor (e.g., graphic processing unit (GPU) for image processing)”), or a frequency of values in the artefact relative to expected quantized values (see Rouhani: Fig.1, [0070] “watermarking keys associated with the output layer 130 may be selected to reside in low-density regions of the machine learning model 100 in order minimize the effects on the accuracy of the machine learning model 100. That is, the samples used to embed a digital watermark in the output layer 130 of the machine learning model 130 should originate from a rarely explored region of the machine learning model 100 occupied by subsets of input data infrequently encountered by the machine learning model 100, for example, during the training of the machine learning model 100.”)
Rouhani does not teach the method wherein:
the fingerprints are generated from existing artefact structure of the first machine learning model without prior modification of the first machine learning model, provide the data characterizing the determining to a consuming application or process.
determining, using a second machine learning model and based on the generated fingerprints whether the first machine learning model is an unauthorized derivative of another machine learning model, thereby detecting a security threat to a machine learning model deployment environment.
initiating, by the consuming application or process, a remediation operation relative to the first machine learning model in response to the determining identifying the first machine learning model as an unauthorized derivative, the remediation operation comprising one or more of: isolating or blocking access to the first machine learning model, partially disabling functionality of the first machine learning model, logging inputs and/or outputs of the first machine learning model, or poisoning the first machine learning model.
However, REIMANN teaches the method comprising:
the fingerprints are generated from existing artefact structure of the first machine learning model without prior modification of the first machine learning model, (see REIMANN: Fig.2, [0020], “At operation 205, a system herein receives a set of deep neural networks (DNNs) including a first plurality of the DNNs trained with an embedded trojan and a second plurality of the DNNs trained without any embedded trojan, wherein each plurality of the trained DNNs is represented by a mathematical formulation learned by the DNNs and expressing a relationship between an input of the DNNs and an output of the DNNs. The training of the first and second plurality of the DNNs results in the creation of population of networks in which a percentage has backdoor(s) and a percentage that does not have any backdoor.” … [0022], “a statistical analysis of the at least one characteristic feature extracted from the mathematical formulation of each of the plurality of trained DNNs is executed to determine whether there is a discernible difference between the first plurality of the DNNs trained with the embedded trojan and the second plurality of the DNNs trained without any embedded trojan.”)
Because both Rouhani and REIMANN are in the same/similar field of endeavor of machine learning model analysis for detecting copying, accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the invention to modify the teaching of Rouhani to include the system that generate fingerprints from existing artefact structure of the first machine learning model without prior modification of the first machine learning model, the generated fingerprints as taught by REIMANN. One would have been motivated to make such a combination in order to provide an accurate, reliable, and efficient mechanism of determining and reporting the likelihood that a deep neural network has a backdoor or trojan installed therein.
However, Zheng teaches the method comprising:
determining, using a second machine learning model and based on the generated fingerprints whether the first machine learning model is an unauthorized derivative of another machine learning model (see Zheng: Col.2, Parag.4,
PNG
media_image1.png
264
406
media_image1.png
Greyscale
thereby detecting a security threat to a machine learning model deployment environment (see Zheng: Abstract: Pg.2977, “Model transfer learning, compression and retraining are commonly used by pirates to evade detection or even redeploy the pirated models for new applications without compromising performance. This paper presents a novel non-intrusive DNN IP fingerprinting method that can detect pirated models and provide a non-repudiable and irrevocable ownership proof simultaneously. The fingerprint is derived from projecting a subset of front-layer weights onto a model owner identity defined random space to enable a distinguisher to differentiate pirated models that are used in the same application or retrained for a different task from originally designed DNN models. The proposed method generates compact and irrevocable fingerprints against model IP misappropriation and ownership fraud.”)
providing data characterizing the determining to a consuming application or process (see Zheng: Table V, Pg. 2984, Parag.2, “For cross-application scenarios, innocent models refer to those models that have the same architecture as the original model but are trained from scratch on different datasets. For example, V11_C100 and V11_ImageNet are two independently trained VGG11 models. The information of innocent models in cross-application scenarios are given in Table V.”)
Because both Rouhani, REIMANN and Zheng are in the same/similar field of endeavor of machine learning model analysis for detecting copying, accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the invention to modify the teaching of Rouhani to include the system that determine whether the first machine learning model is an unauthorized derivative of another machine learning model based on the provided data as taught by Zheng. One would have been motivated to make such a combination in order to improve efficiency and scalability of a model comparison to prevent unauthorized copy or clone of the machine learning model.
Rouhani, REIMANN and Zheng does not teach the system wherein:
initiating, by the consuming application or process, a remediation operation relative to the first machine learning model, in response to the determining identifying the first machine learning model as an unauthorized derivative, the remediation operation comprising one or more of: isolating or blocking access to the first machine learning model, partially disabling functionality of the first machine learning model, logging inputs and/or outputs of the first machine learning model, or poisoning the first machine learning model.
However, STAPLETON teaches the system wherein:
initiating, by the consuming application or process, a remediation operation relative to the first machine learning model (see STAPLETON: Fig.5, [0062], “an example of how the integrity of a trained machine learning model may be protected while also facilitating authentication of access to the model, in accordance with various embodiments.”), in response to the determining identifying the first machine learning model as an unauthorized derivative the remediation operation (see STAPLETON: Fig.3, [0052], “depicts an example of a lookup table 334 that may store information usable, e.g., by license engine 106, to determine information about ML models stored in database 104 (or ML models stored remotely, such as in database 116 or on local volatile or non-volatile memory of remote computing systems 102.sub.2, 102.sub.N, etc.).”), comprising one or more of: isolating or blocking access to the first machine learning model, partially disabling functionality of the first machine learning model, logging inputs and/or outputs of the first machine learning model, or poisoning the first machine learning model (see STAPLETON: Fig.5, [0088], “access to the ML model may be limited to eliminated until the model's integrity can be reestablished, e.g., by retraining the model and/or by providing a replacement model.” … Fig.7, [0089], “the successful integrity check may be logged. … [0096], “system may cause the digital key to be applied as input across at least a first portion of the trained machine learning model to transition one or more gate nodes (e.g., 561, 661) that are interspersed within the trained machine learning model between a locked and an unlocked state”)
Because Rouhani, REIMANN, Zheng and STAPLETON are in the same/similar field of endeavor of machine learning model analysis for detecting copying, accordingly, it would have been obvious to a person of ordinary skill in the art, before the effective filing date of the invention to modify the teaching of Rouhani to include the system that initiating, by the consuming application or process, a remediation operation relative to the first machine learning model as taught by STAPLETON. One would have been motivated to make such a combination in order to improve efficiency and scalability of a model comparison to prevent unauthorized copy or clone of the machine learning model.
Regarding Claim 2,
As shown above, Rouhani, REIMANN, Zheng and STAPLETON teach all the limitations of Claim 1. Rouhani further teaches the method wherein:
the artefacts comprise: weights, biases, running mean, and running variance (see Rouhani: Fig.1-2, [00561, “during the training of the machine learning model 100, θ may denote the parameters of the machine learning model 100 (e.g., weights, biases, and/or the like), ƒ.sub.l.sup.i (x, θ) may denote the activation map corresponding to the input sample x belonging to class i at layer l of the machine learning model.”
Regarding Claim 3,
As shown above, Rouhani, REIMANN, Zheng and STAPLETON teaches all the limitations of Claim 1. Rouhani further teaches the method wherein:
determining is conducted on a fingerprint-by-fingerprint basis (see Rouhani: Fig.3, [0092], “the detection engine 330 may determine that the second machine learning model is a duplicate of the first machine learning model based at least on the digital watermark being determined to be present in the second machine learning model. In some example embodiments, the digital watermark embedded in the machine learning model 100 may be present in the third-party machine learning model 320 if the probability of the machine learning model 320 correctly classifying the K quantity of random input samples exceeds the threshold value p.”
Regarding Claim 4,
As shown above, Rouhani, REIMANN, Zheng and STAPLETON teach all the limitations of Claim 1. Rouhani further teaches the method wherein:
the determining is conducted on a model indicator-by-model indicator basis Fig.6, “the detection engine 330 may determine that the first client 352a is the source of the third party machine learning model 352 and not the second client 352b based at least on the third digital watermark extracted from the third party machine learning model 320 matching the first digital watermark embedded in the first copy of the machine learning model 100 distributed to the first client 352a but not the second digital watermark embedded in the second copy of the machine learning model 100 distributed to the second client 352b.”
Regarding Claim 5,
As shown above, Rouhani, REIMANN, Zheng and STAPLETON teach all the limitations of Claim 4. Rouhani further teaches the method comprising:
generating, by the second machine learning model, a score characterizing the level of similarity, wherein the score forms part of the provided data (see Rouhani: Fig.3, [0064], “whether the third party machine learning model 320 is a duplicate of the machine learning model 100 may be determined based on whether a bit error rate (BER) between the digital watermark embedded in the machine learning model 100 and the digital watermark extracted from the third party machine learning model 320 is below a threshold value. For instance, if the third-party machine learning model 320 is a duplicate of the machine learning model 100, the input samples corresponding to the trigger keys X.sup.key may trigger the same or similar changes to the low probabilistic regions of the activation maps associated with the hidden layers of the third-party machine learning model 320 and may therefore result in a lower bit rate error (BER) value.”)
Regarding Claim 6,
As shown above, Rouhani, REIMANN, Zheng and STAPLETON teach all the limitations of Claim 1. Rouhani further teaches the method wherein:
at least one of the artefacts only has a single corresponding generated fingerprint (see Rouhani: Fig.5, [0089], ”The digital watermark may be embedded in the output layer 130 of the machine learning model 100 after the machine learning model 100 has completed training and one or more digital watermarks have already been embedded in the hidden layers (e.g., the input layer 110, the first intermediate layer 120a, and the second intermediate layer 120b) of the machine learning model 100)
Regarding Claim 7,
As shown above, Rouhani, REIMANN, Zheng and STAPLETON teach all the limitations of Claim 1. Rouhani further teaches the method comprising:
at least one of the artefacts has a plurality of corresponding generated fingerprints (see Rouhani: Fig.5, [0089], “The fine-tuning of the trained machine learning model 100 may include retraining the trained machine learning model such that the machine learning model 100 is able to assign, with an above threshold certainty (e.g., probability of greater than 99%), at least some of the K′ quantity of the random input samples to a corresponding class. Once the fine-tuning of the trained machine learning model 100 is complete, the watermarking keys associated with the output layer 130 may be determined by at least identifying and removing, from the K′ quantity of random input samples, random input samples that trigger false positives and/or false negatives.”)
Regarding Claim 13,
As shown above, Rouhani, Zheng and STAPLETON teach all the limitations of Claim 1. Rouhani further teaches the method wherein:
training the provenance machine learning model with historical fingerprints having a known provenance classification (see Rouhani: Fig.1, [0068], “these adjustments may introduce, into the output layer 130, a statistical bias associated with the input samples such that the machine learning model 100 is able to consistently assign these input samples as high probability members of the corresponding classes. That is, the digital watermark embedded in the output layer 130 of the machine learning model 100 may cause the machine learning model 100 to always determine, with an above threshold probability or certainty, that each of the input samples is a member of a corresponding class”)
Regarding Claim 14,
As shown above, Rouhani, REIMANN, Zheng and STAPLETON teach all the limitations of Claim 11. Rouhani further teaches the method wherein:
the determining comprises one or more of: storing the data characterizing the determining in physical persistence, loading the data characterizing the determining into memory, transmitting the data characterizing the determining over a network to a remote computing device, or causing the data characterizing the determining to be displayed in a graphical user interface (see Rouhani: Fig.7, [0099], “he input/output device 740 provides input/output operations for the computing system 700. In some implementations of the current subject matter, the input/output device 740 includes a keyboard and/or pointing device. In various implementations, the input/output device 740 includes a display unit for displaying graphical user interface”
Regarding independent Claim 15,
Claim 15 is a method claim and has similar/same claim limitation as Claim 1 and is rejected under the same rationale.
Regarding independent Claim 16,
Claim 16 is a method claim and has similar/same claim limitation as Claim 1 and is rejected under the same rationale.
Regarding Claim 17,
As shown above, Rouhani, REIMANN, Zheng and STAPLETON teaches all the limitations of Claim 16. Rouhani and further teaches the method wherein:
the fingerprints are generated by determining a means distance of a manifold representing an artefact from a point in space (see Rouhani: Fig.1, [0061], “Gaussian classes selected to carry the digital watermark, and μ.sub.l.sup.i may denote the mean value of the i.sup.th Gaussian distribution at layer l of the machine learning model 100. Equation (2) may be configured to minimize the spreading (variance) of each Gaussian Mixture Model (GMM) class used to carry the digital watermarking while maximizing the distance between the activation centers belonging to different Gaussian classes.”)
Regarding Claim 18,
As shown above, Rouhani, REIMANN, Zheng and STAPLETON teach all the limitations of Claim 16. Rouhani further teaches the method wherein:
the fingerprints are generated by measuring fluctuations of a manifold representing an artefact relative to a mean distance (see Rouhani: Fig.1, [0061], “Equation (2) may augment data features to better fit a Gaussian Mixture Model distribution. It should be appreciated that the mean values μ.sub.l.sup.i and intermediate activations ƒ.sub.l.sup.i(x, θ) in Equation (2) may be trainable variables that may be iteratively fine-tuned using back-propagation.
Regarding Claim 19,
As shown above, Rouhani, REIMANN, Zheng and STAPLETON teach all the limitations of Claim 16. Rouhani and further teaches the method wherein:
the fingerprints are generated by a combination of determining a means distance of a manifold representing an artefact from a point in space and measuring fluctuations of the manifold relative to a mean distance (see Rouhani: Fig.1, [0071], “different random input sample may be generated to replace the input sample not originating from the rarely explored space of the machine learning model 100. A corresponding ground-truth vector may also be generated and assigned to each input key. For example, if the machine learning model 100 is trained to perform a classification task, then each of the K quantity of random input sample may be associated with a randomly selected label corresponding to the ground-truth classification of that random input sample.”)
Regarding Claim 20,
As shown above, Rouhani, REIMANN, Zheng and STAPLETON teach all the limitations of Claim 16. Rouhani and further teaches the method wherein:
the fingerprints are generated by measuring frequency of values in an artefact relative to expected quantized values (see Rouhani: Fig.1, [0062], “for a selected set of projection matrices A and the N-bit binary string defined by the owner of the machine learning model 100, the selected Gaussian centers (e.g., the mean values of the selected Gaussian distributions s) may be adjusted via back-propagation such that a distance (e.g., Hamming distance) between the binarized projected center b and the actual bit b forming the digital watermark is minimized.”)
Regarding independent Claim 21,
Claim 21 is a method claim and has similar/same claim limitation as Claim 1 and is rejected under the same rationale.
Regarding Claim 22,
As shown above, Rouhani, REIMANN, Zheng and STAPLETON teach all the limitations of Claim 21. Rouhani and further teaches the method wherein:
at least a portion of the artefacts have two or more different fingerprints (see Rouhani: Fig.5, [0089], “The fine-tuning of the trained machine learning model 100 may include retraining the trained machine learning model such that the machine learning model 100 is able to assign, with an above threshold certainty (e.g., probability of greater than 99%), at least some of the K′ quantity of the random input samples to a corresponding class. Once the fine-tuning of the trained machine learning model 100 is complete, the watermarking keys associated with the output layer 130 may be determined by at least identifying and removing, from the K′ quantity of random input samples, random input samples that trigger false positives and/or false negatives.”)
Regarding Claim 23,
As shown above, Rouhani, REIMANN, Zheng and STAPLETON teach all the limitations of Claim 21. Rouhani and further teaches the method wherein:
the second machine learning model generates a similarity score characterizing a level of similarity between the first machine learning model and at least one of the reference machine learning models, and wherein the similarity score forms part of the data provided to the consuming application or process (see Rouhani: Fig.3, [0064], “whether the third party machine learning model 320 is a duplicate of the machine learning model 100 may be determined based on whether a bit error rate (BER) between the digital watermark embedded in the machine learning model 100 and the digital watermark extracted from the third party machine learning model 320 is below a threshold value. For instance, if the third-party machine learning model 320 is a duplicate of the machine learning model 100, the input samples corresponding to the trigger keys X.sup.key may trigger the same or similar changes to the low probabilistic regions of the activation maps associated with the hidden layers of the third-party machine learning model 320 and may therefore result in a lower bit rate error (BER) value
Regarding Claim 24,
As shown above, Rouhani, REIMANN, Zheng and STAPLETON teach all the limitations of Claim 21. Rouhani and further teaches the method wherein:
the determining is conducted on a model indicator-by-model indicator basis (see Rouhani: Fig.6, [0094], “the detection engine 330 may determine that the first client 352a is the source of the third party machine learning model 352 and not the second client 352b based at least on the third digital watermark extracted from the third party machine learning model 320 matching the first digital watermark embedded in the first copy of the machine learning model 100 distributed to the first client 352a but not the second digital watermark embedded in the second copy of the machine learning model 100 distributed to the second client 352b.”)
Regarding Claim 25,
As shown above, Rouhani, Zheng and STAPLETON teach all the limitations of Claim 21. Rouhani and further teaches the method wherein:
the fingerprints are generated by determining a mean distance of a manifold representing an artefact from a point in space (see Rouhani: Fig.1, [0061], “Gaussian classes selected to carry the digital watermark, and μ.sub.l.sup.i may denote the mean value of the i.sup.th Gaussian distribution at layer l of the machine learning model 100. Equation (2) may be configured to minimize the spreading (variance) of each Gaussian Mixture Model (GMM) class used to carry the digital watermarking while maximizing the distance between the activation centers belonging to different Gaussian classes.”)
Regarding Claim 26,
As shown above, Rouhani, REIMANN, Zheng and STAPLETON teach all the limitations of Claim 21. Rouhani and further teaches the method wherein:
the fingerprints are generated by measuring fluctuations of a manifold representing an artefact relative to a mean distance (see Rouhani: Fig.1, [0061], “Equation (2) may augment data features to better fit a Gaussian Mixture Model distribution. It should be appreciated that the mean values μ.sub.l.sup.i and intermediate activations ƒ.sub.l.sup.i(x, θ) in Equation (2) may be trainable variables that may be iteratively fine-tuned using back-propagation.’)
Regarding Claim 27,
As shown above, Rouhani, REIMANN, Zheng and STAPLETON teach all the limitations of Claim 21. Rouhani and further teaches the method wherein:
the fingerprints are generated by a combination of determining a mean distance of a manifold representing an artefact from a point in space and measuring fluctuations of the manifold relative to the mean distance (see Rouhani: Fig.1, [0071], “different random input sample may be generated to replace the input sample not originating from the rarely explored space of the machine learning model 100. A corresponding ground-truth vector may also be generated and assigned to each input key. For example, if the machine learning model 100 is trained to perform a classification task, then each of the K quantity of random input sample may be associated with a randomly selected label corresponding to the ground-truth classification of that random input sample.”)
Regarding Claim 28,
As shown above, Rouhani, REIMANN, Zheng and STAPLETON teach all the limitations of Claim 21. Rouhani and further teaches the method wherein:
initiating, by the consuming application or process, a remediation operation relative to the first machine learning model (see STAPLETON: Fig.5, [0062], “an example of how the integrity of a trained machine learning model may be protected while also facilitating authentication of access to the model, in accordance with various embodiments.”)
Regarding Claim 29,
As shown above, Rouhani, REIMANN, Zheng and STAPLETON teach all the limitations of Claim 28. Rouhani and further teaches the method wherein:
the remediation operation comprises one or more of: isolating or blocking access to the first machine learning model, partially disabling functionality of the first machine learning model, logging inputs and/or outputs of the first machine learning model, or poisoning the first machine learning model (see STAPLETON: Fig.5, [0088], “access to the ML model may be limited to eliminated until the model's integrity can be reestablished, e.g., by retraining the model and/or by providing a replacement model.” … Fig.7, [0089], “the successful integrity check may be logged. … [0096], “system may cause the digital key to be applied as input across at least a first portion of the trained machine learning model to transition one or more gate nodes (e.g., 561, 661) that are interspersed within the trained machine learning model between a locked and an unlocked state”)
Regarding Claim 30,
As shown above, Rouhani, REIMANN, Zheng and STAPLETON teach all the limitations of Claim 21. Rouhani and further teaches the method wherein:
the known provenance classifications comprise a set of fingerprints from models known to be related and a set of fingerprints from models known to not be related (see Rouhani: Fig.5A, [0083], “the detection engine 330 may identify, based at least on an activation map associated with a hidden layer of a first machine learning model, a plurality of input samples that causes the activation functions applied by a plurality of neurons forming the hidden layer to output values occupying one or more low probabilistic regions of the activation map.”)
Response to Arguments
Claim Rejections - 35 U.S.C. § 101,
Regarding the 35 U.S.C. 101 rejection for being directed non-statutory subject matter has been updated based on applicant amendments and withdrawn.
Claim Rejections - 35 U.S.C. § 103,
Applicants’ arguments with respect to claim amendments have been considered but are moot considering the new combination of references being used in the current rejection. The new combination of references was necessitated by Applicant’s claim amendments. Therefore, the claims are rejected under the new combination of references as indicated above.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
PGPUB
NUMBER:
INVENTOR-INFORMATION:
TITLE / DESCRIPTION
US 20210073377 A1
Bos; Joppe Willem
Title: S MACHINE LEARNING MODEL AND METHOD FOR DETERMINING IF THE MACHINE LEARNING MODEL HAS BEEN COPIED
Description: This disclosure relates generally to machine learning, and more particularly, to a machine learning model and method for determining if the machine learning model has been copied.
US 20190258783 A1
Mehta; Sameep
Title: STOLEN MACHINE LEARNING MODEL IDENTIFICATION
Description: classifying the target machine learning model as stolen based upon the stolen score reaching a predetermined threshold.
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 ZELALEM W SHALU whose telephone number is (571)272-3003. The examiner can normally be reached M- F 0800am- 0500pm.
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, Cesar Paula can be reached at (571) 272-4128. 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.
/Zelalem Shalu/Examiner, Art Unit 2145
/CESAR B PAULA/Supervisory Patent Examiner, Art Unit 2145