DETAILED ACTION
This action is in response to the application filed on 04/13/2026. Claims 1-20 are pending examination
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 .
Response to Arguments
Rejections Under 35 U.S.C. § 101
Applicant’s arguments with respect to claims 1, 7 and 13 being rejected under 35 U.S.C 101 have been fully considered and are persuasive. Hence, the rejection for claims 1, 7 and 13 are being withdrawn.
Rejections under 35 U.S.C. § 103
Applicant’s arguments with respect to claim(s) 1-20 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. A new reference Cao et al. (US 202102156351 A1) has been introduced to disclose the concept of using a generative model to create adversarial attack signals. Random input and sample data are provided to a generator, which produces adversarial noise that is injected into the sample. The resulting adversarial examples are evaluated by a classifier to determine attack effectiveness, and the classifiers to determine attack effectiveness, and the classifiers feedback is used to update and train the generator so that future generated attacks are more likely to cause misclassification. Cao therefore discloses generation, injection, evaluation and iterative optimization of attack parameters.
Claim Rejections - 35 USC § 103
In the event the determination of the status of the application as subject to AIA 35 U.S.C. 102 and 103 (or as subject to pre-AIA 35 U.S.C. 102 and 103) is incorrect, any correction of the statutory basis (i.e., changing from AIA to pre-AIA ) for the rejection will not be considered a new ground of rejection if the prior art relied upon, and the rationale supporting the rejection, would be the same under either status.
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claim(s) 1, 5-8 and 13 is/are rejected under 35 U.S.C 103 as being unpatentable over Verma et al. (US 20220180190 A1) hereinafter referred to as Verma, in view of Cao et al. (US 20210256351 A1), hereinafter referred to as Cao.
As per claim 1, Verma discloses a computer-implemented method of training a machine learning model for vulnerability assessment of a cyber-physical system, the method comprising:
generating a random parameter vector; (Sampling a mini-batch of m noise samples {z (1), Z (2) z(m)} from noisy data generating distribution p data (z), Verma, para [0038]. The noise samples {z(...)} from noisy data generating distribution p data (z)).
generating a random attack dataset; (The generator module generates synthesized data objects to fool the unsupervised discriminator module, Verma, para [0008]. Synthesized data objects correspond to artificially generated datasets. Data generated to fool a discriminator is functionally adversarial. Such data is analogous to an attack dataset as it is used to cause misclassification or error in another model component)
However, Verma does not explicitly disclose the limitations:
by implementing, in a system experiment or a time- series simulation of the cyber-physical system, an attack policy using the random parameter vector to inject an attack signal and obtaining runtime effectiveness metric data and runtime stealthiness metric data for the attack signal;
passing a batch of samples through a generator to obtain generated attack policy parameters;
generating, by the generator, a generated attack dataset by implementing, in the system experiment or the time-series simulation, the attack policy using the generated attack policy parameters to inject a generated attack signal and obtaining runtime effectiveness metric data and runtime stealthiness metric data for the generated attack signal;
training first and second discriminators using the random attack dataset and the generated attack dataset, wherein the first and second discriminators are trained to approximate an effectiveness function and a stealthiness function, respectively, for attack policy parameters; and
training the generator using the trained discriminators and a loss function-- to increase a probability that generated attack policy parameters are within a feasible set defined by effectiveness and stealthiness thresholds.
Cao discloses:
by implementing, in a system experiment or a time- series simulation of the cyber-physical system, an attack policy using the random parameter vector to inject an attack signal and obtaining runtime effectiveness metric data and runtime stealthiness metric data for the attack signal; (This generator provides the source of randomness for adversarial noise, which is added to an image and fed into the substitute classifier. Using gradient ascent, or some other equivalent method, the classical generator can then be optimized so that misclassification occurs at the output of the substitute classifier, Cao, para [0022]. Here, the quantum generator output corresponds to the random parameter vector or random latent input. The classical generator uses that random input to produce adversarial noise. Adding noise to an image corresponds to injecting an attack signal into a system input. Processing the resulting image through the classifier can be characterized as a system experiment)
passing a batch of samples through a generator to obtain generated attack policy parameters; (The training may include first sampling from the data set to produce a sample with a corresponding label, many such samples/label combinations may be used sequentially to improve the generator further, Cao, para [0012]. Multiple samples and labels are used to improve the generator. Each sample and quantum generator output is supplied to the classical noise generator. The resulting generator output is interpreted as a generated attack policy parameter because it determines the adversarial perturbation applied to the sample. The samples may be processed sequentially rather than one minibatch)
generating, by the generator, a generated attack dataset by implementing, in the system experiment or the time-series simulation, the attack policy using the generated attack policy parameters to inject a generated attack signal and obtaining runtime effectiveness metric data and runtime stealthiness metric data for the generated attack signal; (Generating a set of data that can fool a classical classifier, Cao, para [0019]. The generated collection of adversarial examples corresponds to a generated attack dataset. Each adversarial example is a sample modified by generator produced noise)
training first and second discriminators using the random attack dataset and the generated attack dataset, wherein the first and second discriminators are trained to approximate an effectiveness function and a stealthiness function, respectively, for attack policy parameters; and (A typical black box attack begins with training a substitute classifier—a classifier that is trained by the attacker to replicate the behavior of the target classifier, Cao, para [0020]. Here, the substitute classifier is trained to replicate a target classifier and is subsequently used to evaluate generated adversarial examples, which is a discriminator. The substitute classifier approximates the target classifier. Its response to a generated perturbation indicates whether the attack causes an incorrect label, which functions as an effectiveness evaluator)
training the generator using the trained discriminators and a loss function-- to increase a probability that generated attack policy parameters are within a feasible set defined by effectiveness and stealthiness thresholds. (Updated to maximize the difference, Cao, para [0012]. The classical generator is updated based on the classifier’s response to the generated noisy example. Maximizing the difference between the original label and an incorrect label is an optimization objective and therefore corresponds to training with a loss function)
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao in order to improve automated vulnerability assessment and network security (See Cao, Abstract)
As per claim 5, Verma and Cao disclose the computer-implemented method of claim 2, wherein
Furthermore, Cao discloses:
the generator comprises a deep neural network (Classical noise generator, Cao, para [0012]. Here the subject matter is classified under neural networks, generative networks and adversarial learning and uses Boltzmann-machine neural network generative models. This is analogous to a deep neural network)
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao in order to improve automated vulnerability assessment and network security (See Cao, Abstract)
As per claim 6, Verma and Cao discloses the computer-implemented method of claim 5, wherein
Furthermore, Cao discloses:
the generator is trained with the loss function:
PNG
media_image1.png
44
416
media_image1.png
Greyscale
(Quantum Boltzmann machines offer a potential method to ameliorate the exponential training time of such models by encoding the energy function in a quantum system Hamiltonian on a quantum device (i.e. a quantum computer or quantum annealer) and minimizing a loss function through repeated measurements on the device.
PNG
media_image2.png
50
263
media_image2.png
Greyscale
, Cao, para [0015]. Here, a minimizing loss function for a quantum Boltzmann machine and optimizing a generator to cause misclassification is disclosed).
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao in order to improve automated vulnerability assessment and network security (See Cao, Abstract)
As per claim 7, Verma discloses a computer-implemented method for training a machine learning model for vulnerability assessment of a networked cyber-physical system, the method comprising:
receiving a generative model parameter vector; (G (z; Θg) and D (x; θd), Verma, para [0027]. Here, Θg is a generative model parameter vector used by the system)
generating sample points from a prior distribution of the generative model parameter vector; (Sample mini-batch from noisy data generating distribution pdata(z), Verma, para [0038]. Pdata(z) functions as a prior distribution).
However, Verma does not explicitly disclose the limitations:
generating attack policy parameters from the sample points, the attack policy parameters defining a start time of attack injection, a duration of the attack injection, and an attack profile;
simulating of the networked cyber-physical system, an attack represented by the attack policy parameters;
evaluating an effectiveness metric and a stealthiness metric of the simulated attack to define a vulnerability set having effectiveness and stealthiness thresholds; and
training a generative model using the simulated attack policy parameters and a loss function based on the effectiveness metric and the stealthiness metric to generate attack policy parameters within the vulnerability set.
Cao discloses:
generating attack policy parameters from the sample points, the attack policy parameters defining a start time of attack injection, a duration of the attack injection, and an attack profile; (Generating a set of data that can fool a classical classifier, Cao, para [0019]. The generated collection of adversarial examples corresponds to a generated attack dataset. Each adversarial example is a sample modified by generator produced noise)
simulating of the networked cyber-physical system, an attack represented by the attack policy parameters; (This generator provides the source of randomness for adversarial noise, which is added to an image and fed into the substitute classifier. Using gradient ascent, or some other equivalent method, the classical generator can then be optimized so that misclassification occurs at the output of the substitute classifier, Cao, para [0022]. Here, the quantum generator output corresponds to the random parameter vector or random latent input. The classical generator uses that random input to produce adversarial noise. Adding noise to an image corresponds to injecting an attack signal into a system input. Processing the resulting image through the classifier can be characterized as a system experiment)
evaluating an effectiveness metric and a stealthiness metric of the simulated attack to define a vulnerability set having effectiveness and stealthiness thresholds; and (This generator provides the source of randomness for adversarial noise, which is added to an image and fed into the substitute classifier. Using gradient ascent, or some other equivalent method, the classical generator can then be optimized so that misclassification occurs at the output of the substitute classifier, Cao, para [0022]. Here, the quantum generator output corresponds to the random parameter vector or random latent input. The classical generator uses that random input to produce adversarial noise. Adding noise to an image corresponds to injecting an attack signal into a system input. Processing the resulting image through the classifier can be characterized as a system experiment)
training a generative model using the simulated attack policy parameters and a loss function based on the effectiveness metric and the stealthiness metric to generate attack policy parameters within the vulnerability set. (Updated to maximize the difference, Cao, para [0012]. The classical generator is updated based on the classifier’s response to the generated noisy example. Maximizing the difference between the original label and an incorrect label is an optimization objective and therefore corresponds to training with a loss function)
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao in order to improve automated vulnerability assessment and network security (See Cao, Abstract).
As per claim 8, Verma and Cao disclose the computer-implemented method of claim 7, wherein
Furthermore, Cao discloses:
training the generative model comprises a deep neural network (Classical noise generator, Cao, para [0012]. Here the subject matter is classified under neural networks, generative networks and adversarial learning and uses Boltzmann-machine neural network generative models. This is analogous to a deep neural network)
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao in order to improve automated vulnerability assessment and network security (See Cao, Abstract)
As per claim 13, Verma discloses a cybersecurity system comprising:
a network configured for communications and/or control of the physical plant; and (The system includes one or more computing devices configured to perform classification and decision-making tasks in an operational environment, Verma, para [0021]. Here, an operational environment encompasses the physical plant or any physical system/ environment whose behavior is monitored or protected)
a cybersecurity controller operably connected to the network, wherein the cybersecurity controller comprises a processor and memory with instructions stored thereon, that, when executed by the processor, cause the processor to: (The generator G (z; Θg) and discriminator D (x; ed) are trained and then used during operation, Verma, para [0027]. A generator is trained and then used, implies that the system receives or accesses a trained generative model for operational use. Because the generator is trained adversarial to produce outputs that challenge another model, it is analogous to trained attack model)
simulate, by the trained attack generative model in a system simulation of the physical plant, a plurality of attacks on network devices of the physical plant; (Sample min-batch of m noise samples {z(1)...z(m)} and generate corresponding outputs using the generator, Verma, para [0038]. The generator produces multiple generated outputs from sampled inputs, each being a synthetic, modeled instance. In adversarial learning, such synthetic outputs simulate attack like behaviors intended to stress or deceive the system. Generating a mini-batch constitutes simulating a plurality of attacks).
However, Verma does not explicitly disclose the limitations:
a physical plant;
receive a trained attack generative model;
determine an attack effectiveness of each of the plurality of attacks;
simulate a vulnerability of the physical plant based on the attack effectiveness of the plurality of attacks; and
control access to the network based on the simulated vulnerability by securing at least one of the network devices.
Cao discloses:
a physical plant; (A physical qubit, Cao, para [0026]. The quantum computer includes physical qubits and those maybe controlled by a physical plant)
receive a trained attack generative model; (The QBM is fed into a classical generator, Cao, para [0022]. Here, a trained quantum generator supplies output to the classical generator. A system receiving or using the trained generator arrangement is interpreted to receiving a trained adversarial or attack generative model. Training and operating the model is disclosed)
determine an attack effectiveness of each of the plurality of attacks; (Compare the classification accuracy, Cao, para [0023]. Multiple adversarial examples are generated and their classification performance is compared. Whether each example fools the classifier is treated as the effectiveness of that attack. Here, aggregate classification accuracy is emphasized)
simulate a vulnerability of the physical plant based on the attack effectiveness of the plurality of attacks; and (Evaluating adversarial examples reveals vulnerability of a classifier to adversarial perturbations, Cao, para [0003]. This is equivalent to testing or assessing classifier vulnerability)
control access to the network based on the simulated vulnerability by securing at least one of the network devices. (Quantum computers offer a potential alternative method for securing AI against adversarial attack, Cao, para [0013]. Here, securing AI against adversarial attacks is identified as a potential application which means it secures a network device, change a network device configuration and patch a device or control access)
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao in order to improve automated vulnerability assessment and network security (See Cao, Abstract)
Claims 2-4,10-12 and 14-20 is/are rejected under 35 U.S.C 103 as being unpatentable over Verma et al. (US 20220180190 A1) hereinafter referred to as Verma, in view of Cao et al. (US 20210256351 A1), hereinafter referred to as Cao in further view of Rivera et al. (US 20210112090 A1), hereinafter referred to as Rivera.
As per claim 2, Verma, Cao disclose the computer-implemented method of claim 1, wherein
However, Verma in view of Cao does not explicitly disclose the limitation:
the generated attack dataset comprises an attack policy
Rivera discloses:
the generated attack dataset comprises an attack policy (Attack vectors may be defined as parameterized attack strategies that determine how an attack signal is generated and applied to the system over time, Rivera, para [0071]. Attack strategies are similar to attack policies which are rules governing how attacks are generated)
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao with Rivera by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao) with network visualization, intrusion detection (Rivera). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao with Rivera in order to effectively identify abnormalities in systems (See Rivera, para [0071])
As per claim 3, Verma, Cao disclose the computer-implemented method of claim 2, wherein
However, Verma in view of Cao does not explicitly disclose the limitation:
the attack policy comprises a ramp attack
Rivera discloses:
the attack policy comprises a ramp attack (Ramp attack: This attack vector involves adding a time varying ramp signal to the input control signal based on a ramp signal parameter, y.sub.ramp, Rivera, para [0075]. Ramp attack is listed as an attack vector within the attack strategy framework, it is a part of the attack policy)
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao with Rivera by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao) with network visualization, intrusion detection (Rivera). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao with Rivera in order to effectively identify abnormalities in systems (See Rivera, para [0071])
As per claim 4, Verma, Cao disclose the computer-implemented method of claim 2, wherein
However, Verma in view of Cao does not explicitly disclose the limitation:
the attack policy comprises a sensor attack
Rivera discloses:
the attack policy comprises a sensor attack (Single cyber-attacks consist of isolated attacks that can be performed on measurements, Rivera, para [0071]. Sensor attacks are those that corrupt sensor measurements)
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao with Rivera by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao) with network visualization, intrusion detection (Rivera). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao with Rivera in order to effectively identify abnormalities in systems (See Rivera, para [0071])
As per claim 10, Verma, Cao disclose the computer-implemented method of claim 7, wherein
However, Verma in view of Cao does not explicitly disclose the limitation:
the attack policy parameters comprise a ramp attack policy.
Rivera discloses:
the attack policy parameters comprise a ramp attack policy (Ramp attack: This attack vector involves adding a time varying ramp signal to the input control signal based on a ramp signal parameter, y.sub.ramp, Rivera, para [0075]. Ramp attack is listed as an attack vector within the attack strategy framework, it is a part of the attack policy)
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao with Rivera by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao) with network visualization, intrusion detection (Rivera). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao with Rivera in order to effectively identify abnormalities in systems (See Rivera, para [0071])
As per claim 11, Verma, Cao discloses the computer-implemented method of claim 7, wherein
However, Verma in view of Cao does not explicitly disclose the limitation:
the attack policy parameters comprise a sine attack policy.
Rivera discloses:
the attack policy parameters comprise a sine attack policy (Malicious tripping attack vector involves malicious tripping of a physical relay. During the attack, false tripping command packets are injected to disconnect the power system components by tripping a circuit breaker, Rivera, para [0073]. This is analogous to a sine attack vector as the malicious attack vector takes a tripping signal input based on a parameter)
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao with Rivera by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao) with network visualization, intrusion detection (Rivera). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao with Rivera in order to effectively identify abnormalities in systems (See Rivera, para [0071])
As per claim 12, Verma, Cao disclose the computer-implemented method of claim 7, wherein
However, Verma in view of Cao does not explicitly disclose the limitation:
the attack policy parameters comprise a pulse attack policy
Rivera discloses:
the attack policy parameters comprise a pulse attack policy (Pulse attack: This attack vector involves periodically changing an input control signal by adding the pulse attack parameter, y.sub. pulse, for a small-time interval, (t1). It retains back the original input for a remaining interval, (T-t1), for the given time period, (T), Rivera, para [0074])
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao with Rivera by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao) with network visualization, intrusion detection (Rivera). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao with Rivera in order to effectively identify abnormalities in systems (See Rivera, para [0071])
As per claim 14, Verma, Cao disclose the system of claim 13, wherein
However, Verma in view of Cao does not explicitly disclose the limitation:
the physical plant comprises a networked pipeline system.
Rivera discloses:
the physical plant comprises a networked pipeline system (The main components of the electrical or power grid are generating stations, electrical substations, and transmission lines, Rivera, para [0016])
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao with Rivera by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao) with network visualization, intrusion detection (Rivera). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao with Rivera in order to effectively identify abnormalities in systems (See Rivera, para [0071])
As per claim 15, Verma, Cao disclose the system of claim 14, wherein
However, Verma in view of Cao does not explicitly disclose the limitation:
the networked pipeline system comprises a plurality of pressure sensors operably coupled to the network, and controlling access to the network comprises securing at least one of the plurality of pressure sensors
Rivera discloses:
the networked pipeline system comprises a plurality of pressure sensors operably coupled to the network, and controlling access to the network comprises securing at least one of the plurality of pressure sensors (IDS develops comprehensive solutions for monitoring possible intrusions upon a power system network 100. In the exemplary system of FIG. 3, a network-based IDS, a model-based IDS, machine learning IDS, and synchrophasor data are integrated into the cyber-security system 300 to detect unknown, coordinated, and stealthy cyber-attacks targeting the power system network 101, Rivera, para [0050], [0068])
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao with Rivera by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao) with network visualization, intrusion detection (Rivera). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao with Rivera in order to effectively identify abnormalities in systems (See Rivera, para [0071])
As per claim 16, Verma, Cao discloses the system of claim 13, wherein
However, Verma in view of Cao does not explicitly disclose the limitation:
the physical plant comprises a power grid
Rivera discloses:
the physical plant comprises a power grid (Cyber physical systems such as power grids, Rivera, para [0016])
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao with Rivera by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao) with network visualization, intrusion detection (Rivera). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao with Rivera in order to effectively identify abnormalities in systems (See Rivera, para [0071])
As per claim 17, Verma, Cao discloses the system of claim 16, wherein
However, Verma in view of Cao does not explicitly disclose the limitation:
the power grid comprises a plurality of meters configured to measure the power and frequency of the power grid, and controlling access to the network comprises securing at least one of the plurality of meters
Rivera discloses:
the power grid comprises a plurality of meters configured to measure the power and frequency of the power grid, and controlling access to the network comprises securing at least one of the plurality of meters (Cyber security system receives a first data set from a power system network. The first data set includes frequency data, Rivera, para [0003] and [0004]).
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao with Rivera by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao) with network visualization, intrusion detection (Rivera). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao with Rivera in order to effectively identify abnormalities in systems (See Rivera, para [0071])
As per claim 18, Verma, Cao disclose the system of claim 13, wherein
However, Verma in view of Cao does not explicitly disclose the limitation:
the physical plant comprises a cyber-physical system
Rivera discloses:
the physical plant comprises a cyber-physical system (The supervisory control and data acquisition (SCADA) network is a cyber-security system, Rivera, para [0003])
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao with Rivera by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao) with network visualization, intrusion detection (Rivera). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao with Rivera in order to effectively identify abnormalities in systems (See Rivera, para [0071])
As per claim 19, Verma, Cao disclose the system of claim 13, wherein
However, Verma in view of Cao does not explicitly disclose the limitation:
the physical plant comprises an industrial facility
Rivera discloses:
the physical plant comprises an industrial facility (SCADA systems are industry- controlled systems which are monitored, Rivera, para [0054])
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao with Rivera by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao) with network visualization, intrusion detection (Rivera). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao with Rivera in order to effectively identify abnormalities in systems (See Rivera, para [0071])
As per claim 20, Verma, Cao disclose the system of claim 13, wherein
However, Verma in view of Cao does not explicitly disclose the limitation:
the controller further contains instructions that cause the processor to simulate a vulnerability of the physical plant and control the physical plant based on the vulnerability
Rivera discloses:
the controller further contains instructions that cause the processor to simulate a vulnerability of the physical plant and control the physical plant based on the vulnerability (An intrusion detection system may use synchrophasor measurements and cyber logs to learn patterns of different scenarios based on spatio-temporal behaviors of power system networks. Such a system may include three layers: Layer 1 includes a model-based IDS that uses a set of specific rules that may be developed based on the spatio-temporal behavior of power system networks during cyber-attacks and normal operation, Rivera, para [0048]).
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao with Rivera by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao) with network visualization, intrusion detection (Rivera). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao with Rivera in order to effectively identify abnormalities in systems (See Rivera, para [0071])
Claims 9 is/are rejected under 35 U.S.C 103 as being unpatentable over Verma et al. (US 20220180190 A1) hereinafter referred to as Verma, in view of Cao et al. (US 20210256351 A1) in further view of Shanbhag et al. (US 20210406681 A1), hereinafter referred to as Shanbhag
As per claim 9, Verma, Cao disclose the computer-implemented method of claim 7, wherein
However, Verma in view of Cao does not explicitly disclose the limitation:
training the generative model comprises selecting a best loss value by
Shanbhag discloses:
training the generative model comprises selecting a best loss value by
(A loss function metric value of a loss value is computed. The method employs the first deep learning network to predict the loss function metric value in association with training a second deep learning network to perform a defined deep learning task, Shanbhag, para [0007]. Claim intents to evaluate loss values and adapt training based on minimizing loss over iterations. Here, the system trains a deep learning model using loss evaluations. A first deep network learns a loss function, the output is used to train another network based on the predicted loss metric. This corresponds to tracking loss values and using them to optimize training).
A person of ordinary skill in the art before the effective filing date of the claimed invention would have combined Verma and Cao by incorporating the method of generative adversarial network for classification (Verma) and hybrid quantum classical adversarial generator (Cao) with learning loss function using deep learning network (Shanbhag). It would have been obvious to a person of ordinary skill in the art before the effective filing date of the claimed invention to combine Verma and Cao with Shanbhag in order to effectively in order to effectively train models using loss functions (See Shanbhag, para [0007]).
Conclusion
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 RAGHAVENDER CHOLLETI whose telephone number is (703) 756-1065. The examiner can normally be reached M-F 9am-5pm ET.
Examiner interviews are available via telephone, in-person, and video conferencing using a USPTO supplied web-based collaboration tool. To schedule an interview, applicant is encouraged to use the USPTO Automated Interview Request (AIR) at http://www.uspto.gov/interviewpractice.
If attempts to reach the examiner by telephone are unsuccessful, the examiner’s supervisor, RUPAL DHARIA can be reached on (571) 272-3880. 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.
Respectfully submitted,
/RAGHAVENDER NMN CHOLLETI/Examiner, Art Unit 2492
/RUPAL DHARIA/ Supervisory Patent Examiner, Art Unit 2492
Prosecution Insights