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 .
Claims 1-7 and 9-11 are pending in this application. Claims 1-2, 7, and 10-11 are amended and claim 8 is canceled by applicant’s amendment filed 15 August 2025.
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.
Claims 1 and 8-11 are rejected under 35 U.S.C. 103 as being unpatentable over Nixon, Jeremy, Jeremiah Liu, and David Berthelot (“Semi-supervised class discovery,” arXiv preprint arXiv:2002.03480 (2020); hereinafter “Nixon”) in view of Abu Al-Haija, Qasem, and Saleh Zein-Sabatto (“An efficient deep-learning-based detection and classification system for cyber-attacks in IoT communication networks,” Electronics 9.12 (2020): 2152; hereinafter “Al-Haija”), and further in view of Kadambe et al. (U.S. 2006/0241927, hereinafter “Kadambe”).
Regarding Claim 1, Nixon teaches a method for classifying an entity and belonging to an unknown class, the method being implemented by a device comprising a neural network (Abstract) and comprising:
training the neural network with a first corpus of data representative of classes of at least one known class (Introduction, first paragraph, and section 2.1—the neural network is first trained using a training dataset),
extracting data converted by the neural network, wherein the extracted data are called hidden data and are produced with a second corpus of data representative of classes of at least one unknown class (fig. 2 and associated description—data fed through the neural network are extracted from the penultimate layer of the network as hidden data),
clustering the extracted hidden data, producing at least one cluster corresponding to at least one new class (fig. 2 and associated description, and section 3.3—the extracted hidden data are clustered into new classes),
adding at least one of the at least one new class to the neural network (fig. 2 and associated description, and section 3.4—new classes are added to the neural network),
training the neural network with at least one portion of the second corpus corresponding to the at least one added new class (fig. 2 and associated description, and section 3.4—new classes are added to the training dataset, indicating a step of training with a dataset that includes a new class), and
classifying an entity belonging to an unknown class with the neural network (sections 3.5 and 4—performance on the new class is evaluated, indicating classifying entities is the dataset as belonging to an unknown class).
Nixon does not specifically teach the method is for classifying a fault affecting a complex electronic-communication system and belonging to an unknown class of fault, and that the new classes are classes of faults; and performing the clustering without a number of the at least one new class of fault being determined in advance of the clustering. However, Al-Haija teaches a method for classifying a fault affecting a complex electronic-communication system, the method being implemented by a device comprising a neural network, and that the classes are classes of faults (Abstract and p. 4, first section—a neural network classifies cyber attacks {a kind of fault} in an Internet of Things communication system.
These claimed elements were known in Nixon and Al-Haija and could have been combined by known methods with no change in their respective functions. It therefore would have been obvious to a person of ordinary skill in the art at the time of filing of the applicant’s invention to combine the classifying a fault affecting a complex electronic-communication system of Al-Haija with the method for classifying of Nixon to yield the predictable result of a method for classifying a fault affecting a complex electronic-communication system and belonging to an unknown class of fault, the method being implemented by a device comprising a neural network and comprising: training the neural network with a first corpus of data representative of faults of at least one known class, extracting data converted by the neural network, wherein the extracted data are called hidden data and are produced with a second corpus of data representative of faults of at least one unknown class, clustering the extracted hidden data, producing at least one cluster corresponding to a new class of fault, adding at least one of the at least one new class of fault to the neural network, training the neural network with at least one portion of the second corpus corresponding to the at least one added new class, and classifying the fault belonging to an unknown class with the neural network. One would be motivated to make this combination for the purpose of protecting a communication system from vulnerabilities by classifying new types of cyber attacks (Al-Haija, Abstract).
Nixon/Al-Haija provides examples using K-means clustering, but also teaches that other clustering algorithms can be utilized (Nixon, fig. 2 and section 3.3, first paragraph). However, Nixon/Al-Haija does not explicitly teach performing the clustering without a number of the at least one new class of fault being determined in advance of the clustering. However, Kadambe teaches clustering data without a number of the at least one new class of fault being determined in advance of the clustering (fig. 8; ¶ [0040] – [0041]—a dynamic clustering algorithm dynamically adjusts the number of new classes based on criteria such as a minimum number of samples in each class. The number of new classes is therefore not determined in advance, but is changed as the clustering algorithm proceeds).
All of the claimed elements were known in Nixon/Al-Haija and Kadambe. As explained above, Nixon/Al-Haija teaches the use of any clustering algorithm, and Kadambe teaches a dynamic clustering algorithm. It therefore would have been obvious to a person of ordinary skill in the art at the time of filing of the applicant’s invention to substitute the dynamic clustering of Kadambe for the K-means clustering of Nixon/Al-Haija to yield the predictable result of clustering the extracted hidden data, producing at least one cluster corresponding to at least one new class of fault, without a number of the at least one new class of fault being determined in advance of the clustering. One would be motivated to make this combination for the purpose of enabling the prediction of future faults in order to be proactive with equipment maintenance (Kadambe, ¶ [0006] – [0008]).
Regarding Claim 7, Nixon/Al-Haija/Kadambe teaches wherein adding is preceded by, for each of the at least one new class of fault, selecting the new class of fault if the corresponding cluster has a minimum degree of distinction or of independence with respect to other clusters corresponding to known classes of faults (Kadambe, fig. 8; ¶ [0035] – [0036] and [0041]—classes with a small distance between them {i.e. a small degree of distinction} may be combined into a single class, indicating that a new class is selected if its cluster has a minimum degree of distinction or of independence with respect to other clusters corresponding to known classes of faults).
Regarding Claim 9, Nixon/Al-Haija/Kadambe teaches wherein a single new class of fault is selected after the clustering (Nixon, section 3.5—the general case of clustering to produce multiple new classes is described, but it can clearly be reduced to the case of a single new class. The process of Kadambe, shown in fig. 8 and described in ¶ [0041], may also result in adding a single new class of fault).
Regarding Claim 10, Nixon teaches a device (Abstract and section 4.3—a neural network system trained on a GPU or CPU implies a device) comprising:
a neural network, for classifying an entity belonging to an unknown class (Abstract);
an input interface for receiving data representative of classes; an output interface for outputting information relative to a class; at least one processor; and at least one memory coupled to the at least one processor, storing instructions that when executed by the at least one processor configure the at least one processor to implement operations (Abstract and section 4.3—it is understood that the device using Tensorflow for neural network models includes input and output interfaces, at least one processor, and at least one memory). Nixon, Al-Haija, and Kadambe teach the operations of the present claim in the same manner as for claim 1, above.
Regarding Claim 11, Nixon teaches a non-transitory computer-readable data medium comprising instructions of a computer program stored thereon which, when executed by at least one processor, configure the at least one processor to classify an entity, which belongs to an unknown class, by implementing operations (Abstract and section 4.3—it is understood that a system using Tensorflow for neural network models and running on a GPU or CPU a computer readable medium comprising instruction that are executed by a processor to perform operations). Nixon, Al-Haija, and Kadambe teach the operations of the present claim in the same manner as for claim 1, above.
Claims 2-5 are rejected under 35 U.S.C. 103 as being unpatentable over Nixon in view of Al-Haija in view of Kadambe, as applied to claim 1, above, and further in view of Wang, Zifeng, et al. (“Open-world class discovery with kernel networks,” 2020 IEEE International Conference on Data Mining (ICDM). IEEE, 2020; hereinafter “Wang”).
Regarding Claim 2, Nixon/Al-Haija teaches wherein the neural network comprises an output layer with at least as many neurons as known classes (Al-Haija, p. 15, fig. 16—the example neural network includes 5 neurons in the output layer for 5 classes), but does not specifically teach wherein the neural network comprises an output layer with at least as many neurons as known classes of faults, and wherein adding at least one new class means adding a neuron to the output layer. However, Wang teaches a neural network comprises an output layer with at least as many neurons as known classes, and wherein adding at least one new class means adding a neuron to the output layer (p. 633, fig. 3 shows the neural network with output nodes for old classes and new classes. Section V. A. on pp. 633-634 describes adding output nodes {neurons} for each new class, thus implying that there is one output neuron for each class).
All of the claimed elements were known in Nixon/Al-Haija/Kadambe and Wang and could have been combined by known methods with no change in their respective functions. It therefore would have been obvious to a person of ordinary skill in the art at the time of filing of the applicant’s invention to combine the output neurons and adding an output neuron of Wang with the neural network and classes of faults of Nixon/Al-Haija/Kadambe to yield the predictable result of wherein the neural network comprises an output layer with at least as many neurons as known classes of faults, and wherein adding at least one new class means adding a neuron to the output layer. One would be motivated to make this combination for the purpose of improving classification accuracy by transferring knowledge of old classes to new classes without overfitting (Wang, section I).
Regarding Claim 3, Nixon/Al-Haija/Kadambe/Wang teaches wherein the neural network is a multilayer perceptron and further comprises an input layer and at least one intermediate layer, between the input and output layers (Wang, pp. 633-634, section V. A. and fig. 1).
Regarding Claim 4, Nixon/Al-Haija/Kadambe/Wang teaches wherein the hidden data are extracted from a last intermediate layer before the output layer (Nixon, fig. 2; also Wang, pp. 633-634, section V. A. and fig. 1—the penultimate layer is a last intermediate layer before the output layer).
Regarding Claim 5, Nixon/Al-Haija/Kadambe/Wang teaches wherein, from the input layer to the output layer, a size of a layer with respect to a size of a preceding layer is decreased by a factor higher than or equal to 2 (Al-Haija, p. 15, fig. 16 and Wang, section V.A. discuss layers and the number of neurons per layer. Sizing the layers so that a size of a layer with respect to a size of a preceding layer is decreased by a factor higher than or equal to 2 is a matter of design choice that is obvious in view of Al-Haija and Wang).
Claim 6 is rejected under 35 U.S.C. 103 as being unpatentable over Nixon in view of Al-Haija in view of Kadambe, as applied to claim 1, above, and further in view of Moustafa, Nour, et al. (“Outlier dirichlet mixture mechanism: Adversarial statistical learning for anomaly detection in the fog,” IEEE Transactions on Information Forensics and Security 14.8 (2019): 1975-1987; hereinafter “Moustafa”).
Regarding Claim 6, Nixon/Al-Haija/Kadambe does not specifically teach wherein the clustering uses a Dirichlet process Gaussian-mixture model. However, Moustafa teaches clustering using a Dirichlet process Gaussian-mixture model (Abstract and p. 1976, first paragraph after bullet points. Clusters are further described on p. 1978, last paragraph of the first column, with the algorithm described in section III.A).
All of the claimed elements were known in Nixon/Al-Haija/Kadambe and Moustafa and could have been combined by known methods with no change in their respective functions. It therefore would have been obvious to a person of ordinary skill in the art at the time of filing of the applicant’s invention to combine the Dirichlet process of Moustafa with the clustering of Nixon/Al-Haija/Kadambe to yield the predictable result of wherein the clustering uses a Dirichlet process Gaussian-mixture model. One would be motivated to make this combination for the purpose of enabling a system to self-adapt against new kinds of attacks (Moustafa, Abstract).
Response to Arguments
The amendments to the claims are accepted as overcoming the previous rejections under 35 U.S.C. 112(b).
Applicant’s arguments with respect to claims 1-7 and 9-11 have been considered but are moot because the new ground of rejection does not rely on any reference applied in the prior rejection of record for any teaching or matter specifically challenged in the argument. Although Nixon does not explicitly teach the new limitations added to independent claims 1, 10, and 11, new prior art reference Kadambe teaches these limitations, as detailed above. In particular, Kadambe teaches an embodiment that performs clustering of data without a number of the at least one new class of fault being determined in advance of the clustering (fig. 8; ¶ [0040] – [0041]—a dynamic clustering algorithm dynamically adjusts the number of new classes based on criteria such as a minimum number of samples in each class. The number of new classes is therefore not determined in advance, but is changed as the clustering algorithm proceeds).
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant’s disclosure. This art includes:
Xiao et al. (U.S. 2018/0218369) teaches determining if a cluster of transaction types experiences anomalous growth, which indicates a new class of anomaly
Shyr et al. (U.S. 2015/0286704) teaches clustering outliers using a distance threshold to determine if an entity is an outlier
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 HAL W SCHNEE whose telephone number is (571) 270-1918. The examiner can normally be reached M-F 7:30 a.m. - 6:00 p.m.
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, Michael Huntley can be reached at 303-297-4307. 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.
/HAL SCHNEE/ Primary Examiner, Art Unit 2129