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 .
DETAILED ACTION
This office action is in response to the claims filed on 02/18/2026.
Claims 1-24 are presented for examination.
Response to Arguments
In reference to applicant’s argument regrading rejections under 35 U.S.C. § 103:
Applicant’s Argument:
Assignee further respectfully submits that Fukushima appears to show a system for estimating a state of an observation target using existing time-series data. See, for example, However, Assignee respectfully submits that Fukushima does not state that it shows "generating, utilizing the processor of the computing device, signals and/or states representative of a set of parameters for adaptive system training at least in part via performing a mixinq operation on the plurality of candidates for time series mixing including maintaininq one or more causal relationships between and/or among the signals and/or states representative of the plurality of candidates for time series mixing in accordance with the causal graph, wherein the mixing operation aligns events based at least in part on dependency paths of the causal graph" as recited in claim 1, as amended. For example, Fukushima discloses "an estimator configured to estimate a state of an observation target on a first time, based on data on the first time included in time- series data obtained." See, for example, paragraph [0029] of Fukushima. It is respectfully submitted that Fukushima therefore appears to teach performing state estimation using observed data, rather than construction of new time-series content,
for example. More particularly, it is respectfully submitted that Fukushima does not teach or suggest constructing training parameters by mixing time series while aligning
events along dependency paths of a causal graph, for example. Therefore, for at least the reasons discussed above, Assignee respectfully submits that Fukushima does not make up for the deficiencies of Abbaszadeh in failing to teach the aforementioned elements of claim 1, as amended. It is respectfully submitted that any purported combination of Abbaszadeh and Fukushima, proper or otherwise, would not yield all of the elements and limitations of claim 1, as amended. Amended claim 1 is therefore patentably distinguished over any combination, proper or otherwise, of Abbaszadeh and Fukushima.
Examiner’s Response:
Examiner respectfully reminds applicant that Fukushima is only brought to cure the specific deficiencies of Abbaszadeh regarding their respective claims. As Abbaszadeh teaches generating, utilizing the processor of the computing device, signals and/or states representative of a set of parameters for adaptive system training, as it can be seen at (Abbaszadeh , [Par.0087]). Furthermore, Fukushima teaches generating, utilizing a processor of the computing device, signals and/or states representative of a plurality of candidates for time series mixing based at least in part on the signals and/or states representative of the plurality of time series, as Fukushima teaches the time-series data shown in FIG. 2B has an incorrect waveform as a whole but includes a waveform same as or similar to the time-series data of the correct action as well. For example, the waveform shown in FIG. 2B is different from a waveform shown in FIG. 2A at P-th second from an action start to Q-th second and R-th second to S-th second. On the other hand, the waveform shown in FIG. 2B is the same as the waveform shown in FIG. 2A until P-th second from an action start, at Q-th second to R-th second and S-th second and subsequent seconds. The P-th second, the Q-th second, the R-th second, and the S-th second are boundaries between the waveform of the correct action and the waveform of the wrong action, therefore, the actions are performed though the different time series (P-th second, Q-th second to R-th second and S-th second and subsequent seconds), that is corresponding to the mixing of the time series, as it can be seen at (FUKUSHIMA, [Par.0032-0034, Fig. 2A, 2B), therefore, the applicant argument is not persuasive, the rejection is still maintained.
training at least in part via performing a mixing operation on the plurality of candidates for time series mixing, including maintaining one or more causal relationships between and/or among the signals and/or states representative of the plurality of candidates for time series mixing in accordance with the causal graph, as Fukushima teaches even when it is determined that entire time-series data acquired from the observation target doing a certain action is wrong, a correction action portion and a wrong action portion are mixed. That is even the time series data of the action is determined as wrong but the action corresponding to a part of the time-series data is sometimes correct, therefore, the mixing of the wrong action and the correct action are considered as the mixing operation, as it can be seen at (FUKUSHIMA, [0032, Fig.2A, 2B]).
including maintaining one or more causal relationships between and/or among the signals and/or states representative of the plurality of candidates for time series mixing in accordance with the causal graph, as FUKUSHIMA teaches the P-th second, the Q-th second, the R-th second, and the S-th second are boundaries between the waveform of the correct action and the waveform of the wrong action, therefore, the fig.2 shows the relationship between the plurality of time series, as it can be seen at (FUKUSHIMA, [Par.0032-0034, Fig. 2A, 2B]).
wherein the mixing operation aligns events based at least in part on dependency paths of the causal graph (FUKUSHIMA, [Par.0032-0034, Fig. 2A, 2B], … [0033] FIGS. 2A and 2B show an example of time-series data of a correct action and time-series data of an incorrect action. More specifically, FIG. 2A shows time-series data of an action determined as correct by a user of the information processing apparatus 101 (for example, an observer of the observation target). FIG. 2B shows time-series data of an action determined as incorrect by the user of the information processing apparatus 101. The time-series data shown in FIG. 2B has an incorrect waveform as a whole but includes a waveform same as or similar to the time-series data of the correct action as well. For example, the waveform shown in FIG. 2B is different from a waveform shown in FIG. 2A at P-th second from an action start to Q-th second and R-th second to S-th second. On the other hand, the waveform shown in FIG. 2B is the same as the waveform shown in FIG. 2A until P-th second from an action start, at Q-th second to R-th second and S-th second and subsequent seconds. The P-th second, the Q-th second, the R-th second, and the S-th second are boundaries between the waveform of the correct action and the waveform of the wrong action.” Examiner’s note, the both (mixed) correct and wrong actions are performed at the different time series are iteratively performed on the graph, such as the P-th second, Q-th second to R-th second and S-th second and subsequent seconds, as shows in the Fir. 2A and 2B.)
Therefore, the applicant’s argument is not persuasive, the rejection is still maintained.
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.
The factual inquiries for establishing a background for determining obviousness under 35 U.S.C. 103 are summarized as follows:
1. Determining the scope and contents of the prior art.
2. Ascertaining the differences between the prior art and the claims at issue.
3. Resolving the level of ordinary skill in the pertinent art.
4. Considering objective evidence present in the application indicating obviousness or non-obviousness.
This application currently names joint inventors. In considering patentability of the claims the examiner presumes that the subject matter of the various claims was commonly owned as of the effective filing date of the claimed invention(s) absent any evidence to the contrary. Applicant is advised of the obligation under 37 CFR 1.56 to point out the inventor and effective filing dates of each claim that was not commonly owned as of the effective filing date of the later invention in order for the examiner to consider the applicability of 35 U.S.C. 102(b)(2)(C) for any potential 35 U.S.C. 102(a)(2) prior art against the later invention.
Claims 1-4, 6-9, 13-16, and 18-21 are rejected under 35 U.S.C. 103 as being unpatentable over Abbaszadeh (U.S. Patent Application Publication No. US 20210081270 A1) in view of FUKUSHIMA (U.S. Patent Application Publication No. US 20210209132 A1).
Regarding claim 1, Abbaszadeh teaches a method, comprising: obtaining and/or generating at a computing device signals and/or states representative of a plurality of time series (Abbaszadeh: Paragraph [0050], Lines 1-3, “time-series data may be received from a collection of monitoring nodes (e.g., sensor, actuator, and/or controller nodes)”);
obtaining and/or generating at the computing device signals and/or states representative of a causal graph (Abbaszadeh: Paragraph [0092], Lines 1-13, “at S1630 a propagation paths map may be used to determine if the current attack potentially propagated from a previous attack … The anomaly propagation paths might also be defined by domain knowledge and pre-stored in the localization system”: The propagation map shows causal relationships between nodes which can be used to determine anomaly propagation.);
the plurality of time series including preserving temporal dependencies between and/or among the signals and/or states representative of the plurality of time series based at least in part on the signals and/or states representative of the causal graph, (Abbaszadeh, [Par.0098-0106], “FIG. 17 illustrates a feature time series 1700 of a first attack example comparing the real-time feature of a monitoring node to the modeled feature of a monitoring node via a graph 1710 according to some embodiments. In particular, the examples described with respect to FIGS. 17 through 16 involve the following parameters for a gas power turbine (similar to those values described with respect to FIGS. 4 through 6): [0099] Compressor Discharge Pressure (“CPD”), [0100] Compressor Discharge Temperature (“CTD”), [0101] Compressor Inlet Temperature (“CTIM”), [0102] Turbine Fuel Flow (“FQG”), [0103] Generator Electrical Power Output (“DWATT”), and [0104] Turbine Exhaust Temperature (“TTXM”). [0105] Consider, for example, an attack on TTXM. In this single attack scenario, the system may want to verify whether it can detect and localize the attacked node. As illustrated in FIG. 17, the attack is detected at t=11 sec. Using the embodiments described herein, the attack is detected within 1 sec and correctly localized to TTXM. FIG. 17 shows the measured feature time series of the detected and localized attack 1730 along with the generated features 1720 estimated using stochastic model-based estimation.[0106] FIG. 18 illustrates a feature time series 1800 via a graph 1810 of a second (stealthy) attack comparing the real-time feature of a monitoring ode to the modeled feature of a monitoring node in accordance with some embodiments. That is, this is again an attack on TTXM but this time the attack simulates a stealthy attack in which the sensor is tampered with slowly over time and/or elaborately. Such stealthy attacks are designed to pass the existing fault diagnosis system and can remain in the control system for a long time without being detected. In this simulation, the attack was applied at t=40 sec. Using the localization methods described herein, the attack was detected at t=105 sec, and is correctly localized to TTXM. FIG. 18 shows the measured feature time series of the detected and localized attack 1830 along with the expected features 1820 estimated using the stochastic model-based estimation.”).
generating, utilizing the processor of the computing device, signals and/or states representative of a set of parameters for adaptive system training (Abbaszadeh , [Par.0087], “Embodiments might be tested using various simulations of a gas turbine. For example, an asset may have 20 monitoring nodes, each having 5 local base features, including, median, standard deviation, kurtosis, range, and a moving average. The features may be extracted, for example, over a sliding window of batch data of node measurements of size 50 second, sliding by one sampling time (Ts=1 sec) at each sampling time. In addition, one transient capturing feature might be added, namely the first derivative of the time-domain node values (rate features) as the 6-th local feature for each node. The transient capturing feature for each node might then pass through a 5-degree smoothing filter. At the global level, there may be two interactive features as the correlation of two monitoring nodes used in those features. The global feature vector might be comprised of 122 features (6 local per node plus 2 global interactive). Then a classification decision boundary could be trained based on ELM neural networks using “normal” and “abnormal” data sets collected by simulating a high-fidelity model of the asset. The “normal” data set might be created by Pseudo-Random Binary Sequence (“PRBS”) excitation to resemble different operational conditions, and the “abnormal” data set might be created by DoE. The features can then be extracted over a sliding window of the time-series data. The training data set might comprise, for example, over 2 million data points, each being a vector of size 122, in the feature space. The ELM training code is implemented efficiently, using sparse matrix manipulations, to be able to handle the big data. To resemble real operations, the simulations may be done in close-loop with the gas turbine controller in the loop. The results may then be compared with the results of another classification decision boundary (using the same classification methods and same data sets) in which only base features are used (no transient capturing features included) as the base-line. The performance of the reliable cyber-threat detection system in various test scenarios may not create false alarms during rapid normal transients and DLN mode transfers, while still detecting attacks even faster than the base-line classifier. The transient capturing features may improve both sensitivity and accuracy of the detection system. Moreover, the reliable system may be computationally low-cost and not add a noticeable demand to real-time computational needs.”)
However, generating, utilizing a processor of the computing device, signals and/or states representative of a plurality of candidates for time series mixing based at least in part on the signals and/or states representative of the plurality of time series, training at least in part via performing a mixing operation on the plurality of candidates for time series mixing, including maintaining one or more causal relationships between and/or among the signals and/or states representative of the plurality of candidates for time series mixing in accordance with the causal graph.
On the other hand, FUKUSHIMA teaches generating, utilizing a processor of the computing device, signals and/or states representative of a plurality of candidates for time series mixing based at least in part on the signals and/or states representative of the plurality of time series (FUKUSHIMA, [Par.0032-0034, Fig. 2A, 2B], “In this embodiment, a class for the entire time-series data obtained from the observation target is not determined. A class is determined for any time of the time-series data. In some cases, even when it is determined that entire time-series data acquired from the observation target doing a certain action is wrong, a correction action portion and a wrong action portion are mixed. That is, even in time-series data of an action determined as wrong as a whole, an action corresponding to a part of the time-series data is sometimes correct. For example, it is conceived that an action for rehabilitation or technical training is performed. A track considered to be correct (a track serving as a model) is often present in the action of the rehabilitation. In the rehabilitation, functional recovery is achieved by repeatedly carrying out the correct action. In a situation without an instructor, a rehabilitation practitioner sometimes performs a wrong action. However, when an action wrong as a whole is partially seen, a correct action is sometimes performed. This embodiment makes it possible to highly accurately determine whether for time-series data of an action serving as an evaluation target, the action is a correct action (a correct state) or a wrong action (an wrong state) at every time. [0033] FIGS. 2A and 2B show an example of time-series data of a correct action and time-series data of an incorrect action. More specifically, FIG. 2A shows time-series data of an action determined as correct by a user of the information processing apparatus 101 (for example, an observer of the observation target). FIG. 2B shows time-series data of an action determined as incorrect by the user of the information processing apparatus 101. The time-series data shown in FIG. 2B has an incorrect waveform as a whole but includes a waveform same as or similar to the time-series data of the correct action as well. For example, the waveform shown in FIG. 2B is different from a waveform shown in FIG. 2A at P-th second from an action start to Q-th second and R-th second to S-th second. On the other hand, the waveform shown in FIG. 2B is the same as the waveform shown in FIG. 2A until P-th second from an action start, at Q-th second to R-th second and S-th second and subsequent seconds. The P-th second, the Q-th second, the R-th second, and the S-th second are boundaries between the waveform of the correct action and the waveform of the wrong action.” Examiner’s note, the time-series data shown in FIG. 2B has an incorrect waveform as a whole but includes a waveform same as or similar to the time-series data of the correct action as well. For example, the waveform shown in FIG. 2B is different from a waveform shown in FIG. 2A at P-th second from an action start to Q-th second and R-th second to S-th second. On the other hand, the waveform shown in FIG. 2B is the same as the waveform shown in FIG. 2A until P-th second from an action start, at Q-th second to R-th second and S-th second and subsequent seconds. The P-th second, the Q-th second, the R-th second, and the S-th second are boundaries between the waveform of the correct action and the waveform of the wrong action, therefore, the actions are performed though the different time series (P-th second, Q-th second to R-th second and S-th second and subsequent seconds), that is corresponding to the mixing of the time series, as the time series is mixing based on correct action time series data and the wrong action time series data.).
training at least in part via performing a mixing operation on the plurality of candidates for time series mixing, including maintaining one or more causal relationships between and/or among the signals and/or states representative of the plurality of candidates for time series mixing in accordance with the causal graph (FUKUSHIMA, [0032, Fig.2A, 2B], “In this embodiment, a class for the entire time-series data obtained from the observation target is not determined. A class is determined for any time of the time-series data. In some cases, even when it is determined that entire time-series data acquired from the observation target doing a certain action is wrong, a correction action portion and a wrong action portion are mixed. That is, even in time-series data of an action determined as wrong as a whole, an action corresponding to a part of the time-series data is sometimes correct. For example, it is conceived that an action for rehabilitation or technical training is performed. A track considered to be correct (a track serving as a model) is often present in the action of the rehabilitation. In the rehabilitation, functional recovery is achieved by repeatedly carrying out the correct action. In a situation without an instructor, a rehabilitation practitioner sometimes performs a wrong action. However, when an action wrong as a whole is partially seen, a correct action is sometimes performed. This embodiment makes it possible to highly accurately determine whether for time-series data of an action serving as an evaluation target, the action is a correct action (a correct state) or a wrong action (an wrong state) at every time.” Examiner’s note, even when it is determined that entire time-series data acquired from the observation target doing a certain action is wrong, a correction action portion and a wrong action portion are mixed. That is even the time series data of the action is determined as wrong but the action corresponding to a part of the time-series data is sometimes correct, therefore, the mixing of the wrong action and the correct action are considered as the mixing operation, it is representing the relationship between the wrong and right action is trained by the system.).
including maintaining one or more causal relationships between and/or among the signals and/or states representative of the plurality of candidates for time series mixing in accordance with the causal graph (FUKUSHIMA, [Par.0032-0034, Fig. 2A, 2B], “In this embodiment, a class for the entire time-series data obtained from the observation target is not determined. A class is determined for any time of the time-series data. In some cases, even when it is determined that entire time-series data acquired from the observation target doing a certain action is wrong, a correction action portion and a wrong action portion are mixed. That is, even in time-series data of an action determined as wrong as a whole, an action corresponding to a part of the time-series data is sometimes correct. For example, it is conceived that an action for rehabilitation or technical training is performed. A track considered to be correct (a track serving as a model) is often present in the action of the rehabilitation. In the rehabilitation, functional recovery is achieved by repeatedly carrying out the correct action. In a situation without an instructor, a rehabilitation practitioner sometimes performs a wrong action. However, when an action wrong as a whole is partially seen, a correct action is sometimes performed. This embodiment makes it possible to highly accurately determine whether for time-series data of an action serving as an evaluation target, the action is a correct action (a correct state) or a wrong action (an wrong state) at every time. [0033] FIGS. 2A and 2B show an example of time-series data of a correct action and time-series data of an incorrect action. More specifically, FIG. 2A shows time-series data of an action determined as correct by a user of the information processing apparatus 101 (for example, an observer of the observation target). FIG. 2B shows time-series data of an action determined as incorrect by the user of the information processing apparatus 101. The time-series data shown in FIG. 2B has an incorrect waveform as a whole but includes a waveform same as or similar to the time-series data of the correct action as well. For example, the waveform shown in FIG. 2B is different from a waveform shown in FIG. 2A at P-th second from an action start to Q-th second and R-th second to S-th second. On the other hand, the waveform shown in FIG. 2B is the same as the waveform shown in FIG. 2A until P-th second from an action start, at Q-th second to R-th second and S-th second and subsequent seconds. The P-th second, the Q-th second, the R-th second, and the S-th second are boundaries between the waveform of the correct action and the waveform of the wrong action.” Examiner’s note, the P-th second, the Q-th second, the R-th second, and the S-th second are boundaries between the waveform of the correct action and the waveform of the wrong action, therefore, the fig.2 shows the relationship between the plurality of time series.).
wherein the mixing operation aligns events based at least in part on dependency paths of the causal graph (FUKUSHIMA, [Par.0032-0034, Fig. 2A, 2B], … [0033] FIGS. 2A and 2B show an example of time-series data of a correct action and time-series data of an incorrect action. More specifically, FIG. 2A shows time-series data of an action determined as correct by a user of the information processing apparatus 101 (for example, an observer of the observation target). FIG. 2B shows time-series data of an action determined as incorrect by the user of the information processing apparatus 101. The time-series data shown in FIG. 2B has an incorrect waveform as a whole but includes a waveform same as or similar to the time-series data of the correct action as well. For example, the waveform shown in FIG. 2B is different from a waveform shown in FIG. 2A at P-th second from an action start to Q-th second and R-th second to S-th second. On the other hand, the waveform shown in FIG. 2B is the same as the waveform shown in FIG. 2A until P-th second from an action start, at Q-th second to R-th second and S-th second and subsequent seconds. The P-th second, the Q-th second, the R-th second, and the S-th second are boundaries between the waveform of the correct action and the waveform of the wrong action.” Examiner’s note, the both (mixed) correct and wrong actions are performed at the different time series are iteratively performed on the graph, such as the P-th second, Q-th second to R-th second and S-th second and subsequent seconds, as shows in the Fir. 2A and 2B.).
Abbaszadeh and FUKUSHIMA are considered to be of the same field of endeavor of the training the machine learning model based on the time series data.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of obtaining and/or generating at the computing device signals and/or states representative of a causal graph; the plurality of time series, including preserving temporal dependencies between and/or among the signals and/or states representative of the plurality of time series based at least in part on the signals and/or states representative of the causal graph; and generating, utilizing the processor of the computing device, signals and/or states representative of a set of parameters for adaptive system training, as taught by Abbaszadeh, to include the generating, utilizing a processor of the computing device, signals and/or states representative of a plurality of candidates for time series mixing based at least in part on the signals and/or states representative of the plurality of time series, training at least in part via performing a mixing operation on the plurality of candidates for time series mixing including maintaining one or more causal relationships between and/or among the signals and/or states representative of the plurality of candidates for time series mixing in accordance with the causal graph, wherein the mixing operation aligns events based at least in part on dependency paths of the causal graph, as taught by FUKUSHIMA . The modification would have been obvious because one of the ordinary skills in art would be motivated to highly accurately determine whether for the time series data of action, ([Par.0032], “as a whole is partially seen, a correct action is sometimes performed. This embodiment makes it possible to highly accurately determine whether for time-series data of an action serving as an evaluation target, the action is a correct action (a correct state) or a wrong action (an wrong state) at every time.”).
Regarding claim 2, Abbaszadeh further teaches storing the signals and/or states representative of the set of parameters for adaptive system training in a memory of the computing device (Abbaszadeh: Paragraph [0117], Lines 4-15, “The table may include, for example, entries identifying industrial assets to be protected … The fields 2402, 2404, 2406, 2408, 2410, 2412, 2414 may, according to some embodiments, specify: an industrial asset identifier 2402, an industrial asset description 2404, a virtual sensor identifier 2406, a matrix 2408, description 2410, a status 2412, and a neutralization level 2414. The virtual sensor database 2400 may be created and updated, for example, when a new physical system is monitored or modeled, an attack is detected, etc.”: The database stores virtual sensor statuses and neutralization levels, which are some of the parameters used to update the virtual sensing model is adapted in reaction to an event or instruction. The matrix 2408 additionally includes propagation path information.).
Regarding claim 3, Abbaszadeh further teaches training the adaptive system at least in part by providing the signals and/or states representative of the set of parameters for adaptive system training as inputs to the adaptive system (Abbaszadeh: Paragraph [0062], Lines 11-17, “The system may then down-select the significant sensors, which are desirable for virtual modeling of each particular on-line sensor estimator. Then, using the aforementioned ANOVA or correlation analysis, the list of the factors to be used in each virtual model may pre-stored into the system, while the virtual sensing model is learnt and adapted online”: the parameters for the adaptive virtual sensing model are determined by propagation analysis of time series data).
Regarding claim 4, Abbaszadeh teaches signals and/or states representative of the causal graph comprise signals and/or states representative at least in part of one or more dependencies between and/or among one or more state variables of at least a first time series of the plurality of time series and at least a second time series of the plurality of time series (Abbaszadeh, [par.0091—0105], “. This time separation test may utilize the fact that if the attacked monitoring under investigation is an artifact of the closed-loop feedback system, then the effect should arise within a time window between the rise time and the settling time of the control loop corresponding to the monitoring node. However, since the system uses a dynamic estimator, a propagation time may need to be added throughout the estimator. Using n features, and p lags in the models, the dynamic estimator will have n*p states, and therefore adds n*p sampling times delay into the system. Therefore, the expected time window for a dependent attack to occur might be defined by… Two sensors are attacked at the same time, namely CPD and CTD, and both attacks are applied at t=15 sec. Using embodiments described herein, both attacks are truly detected and localized within seconds. Out of the other 4 sensors, 3 are correctly not detected at all. One is detected (DWATT) at a later time, which is dependent attack.” Examiner’s note, multiple attacks are detected at the multiple time series.).
Regarding claim 6, Abbaszadeh fails to teach the performing the mixing operation on the signals and/or states representative of the plurality of candidates for time series mixing comprises time-shifting one or more parameters of the at least a first candidate of the plurality of candidates for time series mixing in relation to one or more parameters of the at least a second candidate of the plurality of candidates for time series mixing.
On the other hand, FUKUSHIMA teaches the performing the mixing operation on the signals and/or states representative of the plurality of candidates for time series mixing comprises time-shifting one or more parameters of the at least a first candidate of the plurality of candidates for time series mixing in relation to one or more parameters of the at least a second candidate of the plurality of candidates for time series mixing (FUKUSHIMA, [Par.0032, 0033, 0152-0153], “ FIG. 17 is a diagram showing an example of a result of determination on the actions in FIGS. 14 to 16 for the time-series data including multiple times 1 to T. For each time, “RIGHT” or “WRONG” is indicated as the determination result. Result data in FIG. 17 may be displayed on a display apparatus to allow a user of this apparatus to confirm the determination result. The result data in FIG. 17 includes the section, time, and determination result. The column of section stores the values of classification performance of the classification model used in the time section to which the time belongs. In a case of “HIGH CLASSIFICATION PERFORMANCE”, “RIGHT” or “WRONG” is determined using the classification model, and the determination result is stored on the column of the determination result. In the case of “HIGH CLASSIFICATION PERFORMANCE”, “RIGHT” is stored uniformly on the column of the determination result. [0153] As described above, according to this embodiment, based on a plurality of time-series data for learning to which the training label of a correct or incorrect answer is allocated, the classification model is generated for each time section. In a case of performing the right or wrong determination for each time in the time-series data serving as an evaluation target, the classification model is used to perform the right or wrong determination in the time sections where classification models having high classification performances are generated.”
Abbaszadeh and FUKUSHIMA are considered to be of the same field of endeavor of the training the machine learning model based on the time series data.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the generating the plurality time series data, as taught by Abbaszadeh, to include the performing the mixing operation on the signals and/or states representative of the plurality of candidates for time series mixing comprises time-shifting one or more parameters of the at least a first candidate of the plurality of candidates for time series mixing in relation to one or more parameters of the at least a second candidate of the plurality of candidates for time series mixing, as taught by FUKUSHIMA . The modification would have been obvious because one of the ordinary skills in art would be motivated to highly accurately determine whether for the time series data of action , (FUKUSHIMA, [Par.0032], “as a whole is partially seen, a correct action is sometimes performed. This embodiment makes it possible to highly accurately determine whether for time-series data of an action serving as an evaluation target, the action is a correct action (a correct state) or a wrong action (an wrong state) at every time.”).
Regarding claim 7, Abbaszadeh fails to teach the performing the mixing operation on the signals and/or states representative of the plurality of candidates for time series mixing comprises offsetting in time the signals and/or states representative of the at least a first candidate of the plurality of candidates for time series mixing in relation to the signals and/or states representative of the at least a second candidate of the plurality of candidates for time series mixing.
On the other hand, FUKUSHIMA teaches the performing the mixing operation on the signals and/or states representative of the plurality of candidates for time series mixing comprises offsetting in time the signals and/or states representative of the at least a first candidate of the plurality of candidates for time series mixing in relation to the signals and/or states representative of the at least a second candidate of the plurality of candidates for time series mixing (FUKUSHIMA, [Par.0032, 0033, 0152-0153], “ FIG. 17 is a diagram showing an example of a result of determination on the actions in FIGS. 14 to 16 for the time-series data including multiple times 1 to T. For each time, “RIGHT” or “WRONG” is indicated as the determination result. Result data in FIG. 17 may be displayed on a display apparatus to allow a user of this apparatus to confirm the determination result. The result data in FIG. 17 includes the section, time, and determination result. The column of section stores the values of classification performance of the classification model used in the time section to which the time belongs. In a case of “HIGH CLASSIFICATION PERFORMANCE”, “RIGHT” or “WRONG” is determined using the classification model, and the determination result is stored on the column of the determination result. In the case of “HIGH CLASSIFICATION PERFORMANCE”, “RIGHT” is stored uniformly on the column of the determination result. [0153] As described above, according to this embodiment, based on a plurality of time-series data for learning to which the training label of a correct or incorrect answer is allocated, the classification model is generated for each time section. In a case of performing the right or wrong determination for each time in the time-series data serving as an evaluation target, the classification model is used to perform the right or wrong determination in the time sections where classification models having high classification performances are generated.”).
Abbaszadeh and FUKUSHIMA are considered to be of the same field of endeavor of the training the machine learning model based on the time series data.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the generating the plurality time series data, as taught by Abbaszadeh, to include the the performing the mixing operation on the signals and/or states representative of the plurality of candidates for time series mixing comprises offsetting in time the signals and/or states representative of the at least a first candidate of the plurality of candidates for time series mixing in relation to the signals and/or states representative of the at least a second candidate of the plurality of candidates for time series mixing, as taught by FUKUSHIMA . The modification would have been obvious because one of the ordinary skills in art would be motivated to highly accurately determine whether for the time series data of action , (FUKUSHIMA, [Par.0032], “as a whole is partially seen, a correct action is sometimes performed. This embodiment makes it possible to highly accurately determine whether for time-series data of an action serving as an evaluation target, the action is a correct action (a correct state) or a wrong action (an wrong state) at every time.”).
Regarding claim 8, Abbaszadeh further teaches the performing the mixing operation on the signals and/or states representative of the plurality of candidates for time series mixing comprises: determining whether a temporal attribute of a particular tuple of the at least a first candidate of the plurality of candidates for time series mixing corresponds to a specified reference point in time; determining whether a respective particular tuple of the at least a second candidate of the plurality of candidates for time series mixing comprises a null value; and responsive at least in part to a determination that the temporal attribute of the particular tuple of the at least the first candidate of the plurality of candidates for time series mixing corresponds to the specified reference point in time and at least in part to a determination that the respective particular tuple of the at least the second candidate of the plurality of candidates for time series mixing comprises the null value, setting a temporal attribute of the respective particular tuple of the at least the second time series to the specified reference point in time (Abbaszadeh: Paragraph [0111-0121], “[0111], FIG. 21 shows an architecture for an autonomous reconfigurable virtual sensing system 2100. The system 2100 receives time-series measurements 2120 of the sensors as inputs. The measurements are pre-filtered 2120 for de-noising and outlier removal. Denoising may be done, for example, by low pass filtering using law pass filters whose individual cut-off frequencies may be turned based on the individual bandwidths of each sensor... For example, the system 2100 may have N sensors, of which p sensors are normal and q sensors are independently abnormal. Note that both p and q are time-varying but p[k]+q[k]=N at each time instant k. The p normal sensors are specified by the conformance matrix logic 2160 and down-selected via the indexed selector 2130 to be inputted to the autonomous, resilient estimator 2180., and “[0121], When an anomaly is detected, the switch 2630 is closed and the virtual healthy estimated of the abnormal sensors are passed to the control feedback loop”: At the occurrence of an event within a time series (reference point) from a given node/sensor, the time series output of abnormal nodes effected by the event is turned off (set to zero) at the time of the event and is replaced by virtual healthy node estimated time series, derived from the time series of healthy nodes).
Regarding claim 9, Abbaszadeh teaches the signals and/or states representative at least in part of one or more dependencies between and/or among the one or more state variables of the at least the first time series and the at least the second time series comprise signals and/or states representative of one or more parameters specifying a cause before effect relationship between and/or among at least one of the one or more state variables of the at least the first time series and at least one of the one or more state variables of the at least the second time series (Abbaszadeh, [par.0091—0105], “. This time separation test may utilize the fact that if the attacked monitoring under investigation is an artifact of the closed-loop feedback system, then the effect should arise within a time window between the rise time and the settling time of the control loop corresponding to the monitoring node. However, since the system uses a dynamic estimator, a propagation time may need to be added throughout the estimator. Using n features, and p lags in the models, the dynamic estimator will have n*p states, and therefore adds n*p sampling times delay into the system. Therefore, the expected time window for a dependent attack to occur might be defined by… Two sensors are attacked at the same time, namely CPD and CTD, and both attacks are applied at t=15 sec. Using embodiments described herein, both attacks are truly detected and localized within seconds. Out of the other 4 sensors, 3 are correctly not detected at all. One is detected (DWATT) at a later time, which is dependent attack.” Examiner’s note, multiple attacks are detected at the multiple time series.).
Claims 13-16 and 18-21 recite the same limitations as claims 1-4 and 6-9. Regarding the additional limitations, Abbaszadeh teaches an apparatus, comprising: at least one processor of at least one computing device (Abbaszadeh: Paragraph [0113], Lines 6-13, “The storage device 2330 stores a program 2312 and/or a virtual sensor model 2314 for controlling the processor 2310. The processor 2310 performs instructions of the programs 2312, 2314, and thereby operates in accordance with any of the embodiments described herein. For example, the processor 2310 may determine that at least one abnormal monitoring node is currently being attacked or experiencing a fault”).
Claims 5, 17 are rejected under 35 U.S.C. 103 as being unpatentable over Abbaszadeh (U.S. Patent Application Publication No. US 20210081270 A1) in view of FUKUSHIMA (U.S. Patent Application Publication No. US 20210209132 A1) and further in view of f Xu (U.S. Patent Application Publication No. US 20190097865 A1).
Regarding claim 5, the combination of Abbaszadeh in view of FUKUSHIMA fails to teach the performing the mixing operation on the signals and/or states representative of the plurality of candidates for time series mixing comprises: determining whether one or more respective tuples of the at least a first candidate of the plurality of candidates for time series mixing and the at least a second candidate of the plurality of candidates for time series are independent;
and responsive at least in part to a determination that the one or more respective tuples of the at least the first candidate of the plurality of candidates for time series mixing and the at least the second candidate of the plurality of candidates are independent, performing a superposition operation on the one or more respective tuples of the at least the first candidate and the at least the second candite.
On the other hand, Xu teaches the performing the mixing operation on the signals and/or states representative of the plurality of candidates for time series mixing comprises: determining whether one or more respective tuples of the at least a first candidate of the plurality of candidates for time series mixing and the at least a second candidate of the plurality of candidates for time series are independent (Paragraph [0147], Lines 8-9, “The multiple Type 1 devices/Type 2 devices may operate independently and/or collaboratively”; Paragraph [0364], Lines 16-36, “obtaining, asynchronously by each of the at least one first receiver based on the training wireless signal, at least one time series of training channel information (training CI time series) of the wireless multipath channel between the first receiver and the first transmitter in the training time period associated with the known event … obtaining, asynchronously by each of the at least one second receiver based on the current wireless signal, at least one time series of current channel information (current CI time series) of the wireless multipath channel between the second receiver and the second transmitter in the current time period associated with the current event”: the two time series obtained (training and CI time series) are independent of each other.);
and responsive at least in part to a determination that the one or more respective tuples of the at least the first candidate of the plurality of candidates for time series mixing and the at least the second candidate of the plurality of candidates are independent, performing a superposition operation on the one or more respective tuples of the at least the first candidate and the at least the second candidate (Paragraph [0368], Lines 1-6, “the method comprises aligning a first section of a first time duration of a first CI time series (e.g. the training CI time series) and a second section of a second time duration of a second CI time series (e.g. the current CI time series), and computing a map between items of the first section and items of the second section”).
Abbaszadeh and Xu are considered to be of the same field of endeavor as both are pertinent to time series analysis, particularly where it relates to event recognition.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Abbaszadeh to incorporate the teachings of Xu by mixing/merging time series gathered from different nodes/sensors by, upon determining independence between two items or tuples of two time series, superimposing the two time series by computing a mapping between the two at the particular item or tuple of the time series. Doing so would allow the method to implement event recognition by feeding the information determined by the superposition operation to a classifier, which could use it to associate the particular tuple with a particular kind of event (Xu: Paragraph [0379]).
Claim 17 recites the same limitations as claim 5. Regarding the additional limitations, Abbaszadeh teaches an apparatus, comprising: at least one processor of at least one computing device (Abbaszadeh: Paragraph [0113], Lines 6-13, “The storage device 2330 stores a program 2312 and/or a virtual sensor model 2314 for controlling the processor 2310. The processor 2310 performs instructions of the programs 2312, 2314, and thereby operates in accordance with any of the embodiments described herein. For example, the processor 2310 may determine that at least one abnormal monitoring node is currently being attacked or experiencing a fault”).
Claims 10, 22 are rejected under 35 U.S.C. 103 as being unpatentable over Abbaszadeh (U.S. Patent Application Publication No. US 20210081270 A1) in view of FUKUSHIMA (U.S. Patent Application Publication No. US 20210209132 A1) and further in view of Liu (Foreign Patent Application Publication No. CN 109254865 A, see attached English translation of description).
Regarding claim 10, Abbaszadeh fails to teach generating the signals and/or states representative of the plurality of candidates for time series mixing comprises: generating one or more sets of parameters representative of one or more subgraphs, wherein the one or more subgraphs respectively comprise parameters representative of one or more paths between particular nodes of a plurality of nodes of the causal graph; and for respective subgraphs of the one or more subgraphs, sorting the parameters representative of the one or more paths between the particular nodes of the plurality of nodes of the causal graph in accordance with one or more topological dependencies.
On the other hand, Liu teaches generating the signals and/or states representative of the plurality of candidates for time series mixing comprises: generating one or more sets of parameters representative of one or more subgraphs, wherein the one or more subgraphs respectively comprise parameters representative of one or more paths between particular nodes of a plurality of nodes of the causal graph (Liu: Lines 363-371, “Assuming the abnormal node list set is
V
*
∈
V
, extract the abnormal subgraph
G
*
∈
G
through depth-first search traversal and construct a path, which includes abnormal nodes and physical nodes reachable from
V
*
. Therefore, the abnormal subgraph is a possible abnormal propagation path starting from or arriving at
V
*
.”);
and for respective subgraphs of the one or more subgraphs, sorting the parameters representative of the one or more paths between the particular nodes of the plurality of nodes of the causal graph in accordance with one or more topological dependencies (Liu: Lines 372-379, “Assume the set of all subgraphs is
C
*
∈
C
, the influence count
I
C
*
is the number of times
C
*
appears in the anomaly subgraph, and they are sorted according to
I
C
*
. Nodes with high influence values will affect or be affected by more anomalies.”).
Abbaszadeh and Liu are considered to be of the same field of endeavor as both are pertinent to event/incident propagation analysis and response.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Abbaszadeh to incorporate the teachings of Liu by generating, from a larger causal graph, subgraphs representing possible propagation paths of abnormalities/events and then sorting them based on topological dependencies or influence. Doing so would allow the method to better analyze causal relationships between different features within the time series data and isolate the root cause of an event/abnormality. It would additionally enable the method to detect which events/abnormalities are more important, thus improving the effectiveness of corrective actions (Liu: Lines 341-350).
Claim 22 recites the same limitations as claim 10. Regarding the additional limitations, Abbaszadeh teaches an apparatus, comprising: at least one processor of at least one computing device (Abbaszadeh: Paragraph [0113], Lines 6-13, “The storage device 2330 stores a program 2312 and/or a virtual sensor model 2314 for controlling the processor 2310. The processor 2310 performs instructions of the programs 2312, 2314, and thereby operates in accordance with any of the embodiments described herein. For example, the processor 2310 may determine that at least one abnormal monitoring node is currently being attacked or experiencing a fault”).
Claims 11, 12, 23, 24 are rejected under 35 U.S.C. 103 as being unpatentable over Abbaszadeh (U.S. Patent Application Publication No. US 20210081270 A1) in view of FUKUSHIMA (U.S. Patent Application Publication No. US 20210209132 A1) and further in view of Liu (Foreign Patent Application Publication No. CN 109254865 A, see attached English translation of description).
Regarding claim 11, Abbaszadeh fails to teach the obtaining and/or generating the signals and/or states representative of a causal graph comprises obtaining and/or generating signals and/or states representative of at least one virtual node representative of a plurality of nodes.
On the other hand, Bjørke teaches the obtaining and/or generating the signals and/or states representative of a causal graph comprises obtaining and/or generating signals and/or states representative of at least one virtual node representative of a plurality of nodes (Bjørke: Page 279, Paragraph 1, “A certain hypernode represents a collapse from a set of nodes into one single node”; see network of hypernodes in Figure 4).
Abbaszadeh and Bjørke are considered to be of the same field of endeavor as both are pertinent to using directed graphs/networks to illustrate dependencies/relationships between nodes.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Abbaszadeh to incorporate the teachings of Bjørke by including nodes (referred to as hypernodes by Bjørke) within the causal graph which were representative of a plurality of other nodes. Doing so would allow the method create simplified representations of significantly complex causal graphs, which would in turn make analysis of event propagation easier (Bjørke: Page 275, Paragraph 2).
Regarding claim 12, Abbaszadeh fails to teach the obtaining and/or generating the signals and/or states representative of the at least one virtual node comprises inductively generating signals and/or states representative of a plurality of virtual nodes individually representative of respective pluralities of nodes.
On the other hand, Bjørke teaches the obtaining and/or generating the signals and/or states representative of the at least one virtual node comprises inductively generating signals and/or states representative of a plurality of virtual nodes individually representative of respective pluralities of nodes (Bjørke: Page 279, Paragraph 1, “A certain hypernode represents a collapse from a set of nodes into one single node”; Page 279, Paragraph 3, “From the hypernodes a generalized graph, i.e., a coarser graph, G1 of graph G0 can be constructed. This procedure can be repeated in a recursive manner”, see Figure 6).
Abbaszadeh and Bjørke are considered to be of the same field of endeavor as both are pertinent to using directed graphs/networks to illustrate dependencies/relationships between nodes.
It would have been obvious to one of ordinary skill in the art before the effective filing date of the claimed invention to have modified the method of Abbaszadeh to incorporate the teachings of Bjørke by including virtual nodes (referred to as hypernodes by Bjørke) within the causal graph by inductively generating virtual nodes representative of pluralities of nodes. Doing so would allow the method create simplified representations of significantly complex causal graphs, which would in turn make analysis of event propagation easier (Bjørke: Page 275, Paragraph 2).
Claims 23-24 recites the same limitations as claims 11-12. Regarding the additional limitations, Abbaszadeh teaches an apparatus, comprising: at least one processor of at least one computing device (Abbaszadeh: Paragraph [0113], Lines 6-13, “The storage device 2330 stores a program 2312 and/or a virtual sensor model 2314 for controlling the processor 2310. The processor 2310 performs instructions of the programs 2312, 2314, and thereby operates in accordance with any of the embodiments described herein. For example, the processor 2310 may determine that at least one abnormal monitoring node is currently being attacked or experiencing a fault”).
Conclusion
THIS ACTION IS MADE FINAL. 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 EM N TRIEU whose telephone number is (571)272-5747. The examiner can normally be reached on Mon-Fri from 9:00-5:00.
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, Omar Fernandez Rivas can be reached on (571) 272-2589. 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.
/E.T./Examiner, Art Unit 2128
/OMAR F FERNANDEZ RIVAS/Supervisory Patent Examiner, Art Unit 2128
Prosecution Insights