DETAILED ACTION
Notice of Pre-AIA or AIA Status
The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA .
Information Disclosure Statement
The information disclosure statement(s) (IDS) submitted on 03/21/2025 and 03/20/2026 were filed before the mailing date of this office action. The submission is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statements are being considered by the examiner.
Claim Rejections - 35 USC § 103
The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action:
A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made.
Claims 1-2, 4 and 7-8 are rejected under 35 U.S.C. 103 as being unpatentable over US-PGPUB No. 2023/0385548 A1 to Tully, and further in view of US-PGPUB No. 2013/0114864 A1 to Garcia et al. (hereinafter “Garcia”)
Regarding claim 1:
Tully discloses:
(Currently Amended) A classification apparatus (see Fig. 2, First compute device 202) comprising:
a memory (see Fig., Memory 206);
and processing circuitry (see Fig. 2, Processor 204) configured to:
extract feature values of text […] included in a post related to a security threat of a Social Networking Service (SNS) from the post (¶50: “At 442, text is extracted from the post data, and the extracted text is optionally evaluated, at 444, to identify a subset of the post data belonging to one or more IO campaigns of interest … At 446, features are extracted from the extracted post text … and the extracted features are tokenized to produce a set of tokens,”);
perform learning using the feature values for teaching data to which a correct answer label indicating whether each post is a post related to the security threat is attached to thereby learn a machine learning model (¶52: “At 449, the tokens generated at 446 are used to “fine tune” the pre-trained neural network language model, to produce the fine-tuned neural network IO campaign prediction model 450 (e.g., a “trained neural network language model” 206G in FIG. 2). As used herein, “fine tuning” refers to one or more of training, updating, and retraining, and may include transfer learning.”) for classifying whether an input post is a post related to the security threat (¶52: “The fine-tuned neural network IO campaign prediction model 450 can then be used to directly predict, in response to an input including social media post data, whether the social media post associated with the social media post data is part of an IO campaign.”);
classify whether an input post is a post related to the security threat by using the learned machine learning model (¶54-55: “At 564, a fine-tuned neural network language model (e.g., the fine-tuned neural network IO campaign prediction model 450 of FIG. 4 or the trained neural network language model 206G of FIG. 2) is used to predict a score for the social media post, using the tokens generated at 562. The score is compared, at 566, to a threshold value to determine whether the prediction score exceeds the threshold value. [0055] If the prediction score does not exceed the threshold value, the classifier 206P may discard (at 574) or otherwise disregard (e.g., whitelist and/or classify as “benign” or “not IO activity”) the social media post. If the prediction score exceeds the threshold value, a threat warning is generated (at 568).”); and
output a result of classification (¶55: “The threat warning (optionally in combination with the social media post, the extracted tokens and/or data associated with the social media post) are incorporated into a threat report at 569.”).
However, Tully does not explicitly teach the following limitation taught by Garcia:
[extract feature values of text] […] and an image (Garcia, ¶38: “Extracted features 210 displays a list of features or characteristics 210a-210n extracted by API 140 from photo 200.”, see Fig 2, Extracted Features 210).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to modify the teachings of Tully to incorporate the functionality of the API to extract information from metadata or EXIF data attached to an uploaded image, as disclosed by Garcia, such modification would enable the system to determine an engagement metric, and apply one or more policies to an image based on the engagement metric.
Regarding claim 2:
The combination of Tully and Garcia discloses:
(Original) The classification apparatus according to claim 1, wherein the feature values of the image included in the post include a feature value of the image (Garcia, ¶37: “API 140 may extract the resolution and color depth from the image, whether the image was modified and by what software, the metering and autofocus settings, or any other metadata of the image. ”) and a feature value of a character string obtained by optical character recognition of the image (Garcia, ¶35: “a bottle of "Coca-Cola", may be detected through optical character recognition or other computer vision techniques.”, see Fig. 2, Extracted Features, 210i=COCA-COLA).
The same motivation which is applied to claim 1 with respect to Garcia applies to claim 2.
Regarding claim 4:
The combination of Tully and Garcia discloses:
(Original) The classification apparatus according to claim 1, wherein the feature value is at least any of a feature value of an account of a poster of the post, a feature value of a content of the post, a feature value of an URL or a domain name extracted from the text or the image of the post, a feature value of a character string obtained by optical recognition of the image included in the post (Garcia, ¶35: “a bottle of "Coca-Cola", may be detected through optical character recognition or other computer vision techniques.”, see Fig. 2, Extracted Features, 210i = COCA-COLA), a feature value of the image included in the post, and a feature value of context of the text included in the post.
The same motivation which is applied to claim 1 with respect to Garcia applies to claim 4.
Regarding claim 7:
Claim 7 recites substantially the same limitations as claim 1 in the form of a method implementing the corresponding functionality. Therefore, it is rejected by the same rationale.
Regarding claim 8:
Claim 8 recites substantially the same limitations as claim 1 in the form of a recording medium for storing the classification program. Therefore, it is rejected by the same rationale.
Claim 3 is rejected under 35 U.S.C. 103 as being unpatentable over Tully, Garcia, and further in view of US-PGPUB No. 2014/0096242 A1 to Dong et al. (hereinafter “Dong”)
Regarding claim 3:
The combination of Tully and Garcia discloses the classification apparatus according to claim 1, but fails to explicitly disclose the following limitation taught by Dong:
wherein the feature values include a feature value of an URL or a domain name extracted from the text or the image of the post (Dong, ¶91 : “… features of the microblog post are extracted from the information related to the microblog post, and credibility of the URL of the website contained in the microblog post is calculated according to the extracted features of the microblog post, so as to determine whether the URL of the website is a URL of a phishing website.”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to modify the teachings of the combination of Tully and Garcia to incorporate the functionality of the method to extract features of a microblog post containing a url and determine the credibility of the URL of the website contained in the microblog post, as disclosed by Dong, such modification would enable the system to determine whether a URL of a website contained in a microblog post is a URL of a phishing website, thereby providing convenience for users.
Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over Tully, Garcia, and further in view of US-PGPUB No. 2018/0330258 A1 to Harris et al. (hereinafter “Harris”)
Regarding claim 5:
The combination of Tully and Garcia discloses the classification apparatus according to claim 1, but fails to explicitly disclose the following limitation taught by Harris:
wherein the processing circuitry is further configured to select a feature value effective for classification of whether a post is related to the security threat from among the feature values extracted (Harris, ¶140-141: “VII. USING SELECTED FEATURE SET TO TRAIN MODEL … the updated feature set may be used to retrain the AI model.”, ¶41: “predictive output 150 may include a classification of a threat,”) and learn the machine learning model by using the selected feature value (Harris, ¶141: “the training data may be refreshed, and the AI model may continue learning; this time with new features that better reflect the current state of the information space from which the AI is predicting outcomes for.”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to modify the teachings of the combination of Tully and Garcia to incorporate the functionality of the method to determine appropriate responses for the chat bot by looking at the relevance of the message to topics or products being discussed by other users in other chat bot conversations, as disclosed by Harris, such modification would enable the system to determine a response that may recommend or initiate an action which is relevant to the user and other users that share the same features at any given moment in time based on current trends and conversations.
Claim 5 is rejected under 35 U.S.C. 103 as being unpatentable over Tully, Garcia, Harris, and further in view of US-PGPUB No. 2022/0164248 A1 to Stein et al. (hereinafter “Stein”)
Regarding claim 6:
The combination of Tully, Garcia and Harris discloses the classification apparatus according to claim 5, but fails to explicitly disclose the following limitation taught by Stein:
wherein the processing circuitry is further configured to select a feature value effective for classification of whether a post is related to the security threat by Boruta-SHAP (Stein, ¶228: “The classification modeling is finalized with a feature selection process (e.g. a Boruta based feature selection process) in order to include only the statistically significant features from amongst all of the candidate features of the features matrix.”).
It would have been obvious to one of ordinary skill in the art before the effective filing date of the invention to modify the teachings of the combination of Tully, Garcia and Harris to implement a Boruta based feature selection process to include statistically significant features from candidate features, as disclosed by Harris, such modification would enable the system to selectively apply significant features that affect a classification model the most.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Batchu (US 11399035 B1)- disclosed a method comprising: extracting, from a link contained in an electronic message received from an upstream device on a network, first unit-level input data of a first semantic type and second unit-level input data of a second semantic type; in response to inputting the first and second unit-level input data into first and second deep learning models, respectively, outputting, by the first and second deep learning models, first and second unit-level classification data that corresponds to the first and second unit-level input data, respectively.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to MATTHIAS HABTEGEORGIS whose telephone number is (571)272-1916. The examiner can normally be reached M-F 8am-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, William R. Korzuch can be reached at (571)272-7589. 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.
/MATTHIAS HABTEGEORGIS/ Examiner, Art Unit 2491
Prosecution Insights