Prosecution Insights
Last updated: April 19, 2026
Application No. 17/552,445

SYSTEM AND METHOD FOR CLUSTERING EMAILS IDENTIFIED AS SPAM

Non-Final OA §103
Filed
Dec 16, 2021
Examiner
SINGH, AMRESH
Art Unit
2159
Tech Center
2100 — Computer Architecture & Software
Assignee
Ao Kaspersky Lab
OA Round
3 (Non-Final)
76%
Grant Probability
Favorable
3-4
OA Rounds
3y 9m
To Grant
98%
With Interview

Examiner Intelligence

Grants 76% — above average
76%
Career Allow Rate
463 granted / 610 resolved
+20.9% vs TC avg
Strong +22% interview lift
Without
With
+22.0%
Interview Lift
resolved cases with interview
Typical timeline
3y 9m
Avg Prosecution
32 currently pending
Career history
642
Total Applications
across all art units

Statute-Specific Performance

§101
18.8%
-21.2% vs TC avg
§103
46.0%
+6.0% vs TC avg
§102
15.3%
-24.7% vs TC avg
§112
6.3%
-33.7% vs TC avg
Black line = Tech Center average estimate • Based on career data from 610 resolved cases

Office Action

§103
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 Claims 1-24 are presented for examination. Claims 1 and 17 are amended. Claim 26 is new. This is a non-Final Action. A request for continued examination under 37 CFR 1.114, including the fee set forth in 37 CFR 1.17(e), was filed in this application after final rejection. Since this application is eligible for continued examination under 37 CFR 1.114, and the fee set forth in 37 CFR 1.17(e) has been timely paid, the finality of the previous Office action has been withdrawn pursuant to 37 CFR 1.114. Applicant's submission filed on 03/06/2026 has been entered. Response to Arguments Applicant’s arguments with respect to claim(s) 1-26 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. 101 abstract idea has been obviated due to current amendment to the claims. Claims 4-8, 12-16 and 20-25 are still indicated as allowable and would move the prosecution forward if combined into independent claims with the interlinking claims. Claim Rejections - 35 USC § 103 The following is a quotation of 35 U.S.C. 103(a) which forms the basis for all obviousness rejections set forth in this Office action: (a) A patent may not be obtained though the invention is not identically disclosed or described as set forth in section 102 of this title, if the differences between the subject matter sought to be patented and the prior art are such that the subject matter as a whole would have been obvious at the time the invention was made to a person having ordinary skill in the art to which said subject matter pertains. Patentability shall not be negatived by the manner in which the invention was made. Claims 1-2, 9-10 and 17-18 are rejected under 35 U.S.C. 103(a) as being unpatentable over Cosoi et al. (US 8,131,655) in view of Salem et al. (US 10,762,200) further in view of Musat et al. (US 8,170,966) and Vadrevu et al. (US 20120016877) 1. Cosoi teaches, A method for clustering email messages identified as spam using a trained classifier, the method comprising (Abstract): selecting at least two characteristics from each received email message (Abstract – extracting multiple features from messages, Cosoi); for each received email message, using a classifier containing a neural network, determining whether or not the email message is a spam based on the at least two characteristics of the email message; (Abstract – spam filtering based on the features extracted from the message using a neural network to filter and classify, Cosoi); for each email message determined as being a spam email (Abstract – teaches spam detection and classification, Cosoi). Cosoi does not explicitly teach, calculating a feature vector, the feature vector being calculated at a final hidden layer of the neural network; and generating one or more clusters of the email messages identified as spam based on similarities of the feature vectors calculated at the final hidden layer of the neural network, wherein a cluster is generated when a cosine distance between two or more feature vectors is below a predetermined threshold value However, Salem teaches, calculating a feature vector, the feature vector being calculated at a final hidden layer of the neural network (Fig 7: 702, 704 and 708, Col 23: lines 20-63 – teaches extracting the output value of a final hidden layer and provide it as a feature, Salem). based on similarities of the feature vectors calculated at the final hidden layer of the neural network (Fig 7: col 23: 20-63 – teaches extracting feature representations from NN, including obtaining output values from internal layers and using those output as feature vectors. A POSITA would have recognized that such feature vectors are commonly obtained from a Hidden layer of a NN, Salem) It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which said subject matter pertains to allow Cosoi’s invention to be combined with Salem because both references are in the same field of endeavor of machine learning based message/content classification. Salem teaches extracting a feature vector from a final hidden layer of a neural network, a technique that would have been recognized as broadly applicable to any NN based classification system, including Cosoi’s spam classifier. The substitution or incorporation of Salem’s hidden layer vector extraction into Cosoi’s system would have been a predictable improvement yielding enhanced internal representations without altering Cosoi’s architecture. Musat teaches, wherein a width of the final hidden layer is restricted to reduce a dimensionality of the feature vector (Abstract, Figs. 4 and 10 – teaches clustering messages in a feature vector space and performing distance-based similarity computations between feature vectors, including determining distances between future vectors and cluster centroids, Musat; The features vectors utilized by Musat was taught by Salem’s extracting feature representations from neural networks (Fig 7: 702, 704, 708, Col 23: lines 20-63, Salem)) It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which said subject matter pertains to combine Cosoi with Salem and Musat because each reference addresses complementary aspects of processing electronic messages using feature-based machine learning techniques. A POSITA would have recognized that Cosoi’s neural network classifier produces internal representations of message content, and Salem teaches leveraging such internal representations as feature vectors. Musat teaches the clustering based on feature vector similarity enables identification of related messages. Applying Musat’s clustering techniques to feature vector derived from Cosoi’s classifiers (as taught by Salem) would allow grouping of semantically similar spam messages, thereby improving detection of coordinated spam patterns and reducing redundant processing. This represents a predictable integration of known techniques where feature representations generated by a NN are used as inputs to a clustering system. Vadrevu teaches, …wherein a cluster is generated when a cosine distance between two or more feature vectors is below a predetermined threshold value (Paragraphs 7-8, 34, 40-41 and 49 – teaches generating one or more clusters based on similarities of feature vectors, including clustering documents using similarity measures, further teaching computing cosine similarity between feature vectors and assigning items to clusters when a similarity or distance satisfies a predetermined threshold condition, Vadrevu). It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which said subject matter pertains to combine Vadrevu, Cosoi, Salem and Musat because each reference addresses complementary aspects of processing textual data using feature-based techniques. Cosoi teaches identifying email messages as spam using NN classifier, Salem teaches extracting feature representations from NN layers as feature vectors, and Musat teaches clustering messages in a feature vector space based on similarity. Vadrevu teaches clustering data using similarity measures, including cosin similarity, and assigning items to clusters when a similarity of distance satisfies a predetermined threshold condition. A POSITA would have been motivated to apply Vadrevu’s cosine-based similarity metric and threshold-based clustering criteria within the feature-vector clustering to the spam messages identified by Cosoi, in order to improve the accuracy and effectiveness of grouping similar messages (e.g. spam campaigns). This combination represents a predictable use or known techniques, NN feature extraction and similarity based clustering, yielding predictable results consistent with KSR. 2. The combination of Cosoi, Salem, Musat and Vadrevu teach, The method of claim 1, wherein a characteristic of the email message comprises at least one of a value of a header of the email message, and a sequence of parts of the header of the email message (Abstract – email characteristics such as header heuristics, Cosoi). Claims 9 and 17 are similar to claim 1 hence rejected similarly. Claims 10 and 18 are similar to claim 2 hence rejected similarly. Claims 3, 11 and 19 are rejected under 35 U.S.C. 103(a) as being unpatentable over Cosoi et al. (US 8,131,655) in view of Salem et al. (US 10,762,200) and Musat et al. (US 8,170,966) and Vadrevu et al. (US 20120016877) further in view of Jia et al. (“ORTHOGONAL DEEP NEURAL NETWORK”) All the limitations of claim 1 are taught above. 3. The combination of Cosoi, Salem and Zhevena do not explicitly teach, wherein the classifier is trained such that orthogonality of matrices of the neural network is preserved. However, Jia teaches wherein the classifier is trained such that orthogonality of matrices of the neural network is preserved (Abstract). It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which said subject matter pertains to allow Cosoi’s invention to be combined Jia . Because both Cosoi and Jia are in the same field of endeavor of NN design. Jia shows orthogonal weights improve gradient flow and generalization; a spam-filter engineer would immediately see the benefits of utilizing these teachings. Claims 11 and 19 are similar to claims 3 hence rejected similarly. Claim 25 is rejected under 35 U.S.C. 103(a) as being unpatentable over Cosoi et al. (US 8,131,655) in view of Salem et al. (US 10,762,200), Musat et al. (US 8,170,966) and Vadrevu et al. (US 20120016877) and further in view of Zheleva (US 8,825,769) All the limitation of claim 1 are taught above. 25. The combination of Cosoi, Salem Musat and Vadrevu teach, in response to determining that classifying all email messages within a cluster of the one or more clusters as spam requires application of multiple detection rules from a plurality of detection rules, consolidating the multiple detection rules into a single rule for performing spam detection (Col 8: lines 14-20 – teaches clustering engine 52 aggregates incoming messages 40 into message clusters… each cluster may contain message characterized by a subset of message features having similar values… “ an example message cluster may include a subset of spam messages such as … an individual spam wave” and Col 10: lines 39-49 – teaches a message is considered to belong to spam closer if the feature vector lies inside the area of influence of the cluster i.e hyperspace distance … is smaller than a threshold - this does not explicitly use the term “rule’ however, it teaches that each cluster represents a single characterization construct derived from many feature-based heuristics. A cluster effectively consolidates multiple detection heuristics into a single classification constructs (the cluster centroid / area of influence), used to determine whether messages belong to that spam type thus teaching the limitation of consolidating multiple detection-related heuristics into a single determination, Musat). The combination of Cosoi, Salem, Musat and Vadrevu do not explicitly teach, detecting spam in one or more new email messages using the single rule. However, Zheleva teaches, detecting spam in one or more new email messages using the single rule (Abstract – filtering the electronic email if the signature value exceeds the signature value threshold; Col 2: lines 29-31 – teaches the signature value represents aggregated reporter information used to determine whether the electronic message should be considered spam – thus teaching using one decision construction “signature value” compared to a threshold to determine whether new incoming messages are spam (i.e. applying a “single rule” to new messages”, Zheleva). It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which said subject matter pertains to apply Zheleva’s rule consolidation and reuse mechanism to cluster of similar messages identified using the techniques of Vadrevu and Musat and represent using Salem’s feature vectors in order to reduce redundant rule evaluation and improve efficiency when processing groups of similar spam messages (e.g., spam campaigns). This represents a predictable use of known techniques yielding predictable results consistent with KSR. Claim 26 is rejected under 35 U.S.C. 103(a) as being unpatentable over Cosoi et al. (US 8,131,655) in view of Salem et al. (US 10,762,200), Musat et al. (US 8,170,966) and Vadrevu et al. (US 20120016877) and further in view of Pettigrew et al. (US 7,797,443) All the limitations of claim 1 are taught above. 26. The combination of Cosoi, Salem Musat and Vadrevu teaches, …of the generated cluster (paragraphs 7-8 – teaches clustering the documents such that each document belongs to one of the clusters, Vadrevu). The combination of Cosoi, Salem Musat and Vadrevu do not explicitly teach, identifying one or more IP addresses of servers from which the email messages were sent to detect a bot-net; and updating a list of addresses of bot-nets based on the identified one or more IP addresses. However, Pettigrew teaches, identifying one or more IP addresses of servers from which the email messages (Abstract - teaches the packet sniffers extract originating IP addresses associated with e-mail messages, Pettigrew) were sent to detect a bot-net (Col 1: last paragraph, Col 2: first paragraph - determine whether traffic from an originating IP address has exceeded a threshold value… detect and stop spam messages from an IP address, Pettigrew); and updating a list of addresses of bot-nets based on the identified one or more IP addresses (Col 8: lines 24-31 - teaches blocking future email messages from the IP address, and add the ip address to the blacklist, Pettigrew). It would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which said subject matter pertains to combine Pettigrew with Cosoi, Salem, Vadrevu and Musat because each references address complementary aspects of email processing and spam detection. A POSITA would have been motivated to apply Pettigrew’s IP based source identification and blacklist updating techniques to the clustered spam messages identified using the techniques of Vadrevu and Musat and represented using Salem’s feature vectors in the classification framework of Cosoi, in order to improve detection of coordinated spam campaigns and prevent future messages from known malicious sources. This combination represents a predictable use of known techniques, neural network based classification, feature based clustering, and IP based source tracking and blacklisting, yielding predictable result consistent with KSR. Allowable Subject Matter Claim 4-8, 12-16, 20-24 objected to as being dependent upon a rejected base claim, but would be allowable if rewritten in independent form including all of the limitations of the base claim and any intervening claims. Conclusion Any inquiry concerning this communication or earlier communications from the examiner should be directed to AMRESH SINGH whose telephone number is (571)270-3560. The examiner can normally be reached Monday-Friday 8am-5pm. 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, Ann J. Lo can be reached at (571) 272-9767. 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. /AMRESH SINGH/Primary Examiner, Art Unit 2159
Read full office action

Prosecution Timeline

Dec 16, 2021
Application Filed
May 31, 2025
Non-Final Rejection — §103
Sep 02, 2025
Response Filed
Nov 24, 2025
Examiner Interview (Telephonic)
Dec 13, 2025
Final Rejection — §103
Mar 06, 2026
Request for Continued Examination
Mar 14, 2026
Response after Non-Final Action
Mar 21, 2026
Non-Final Rejection — §103 (current)

Precedent Cases

Applications granted by this same examiner with similar technology

Patent 12591804
SYSTEMS AND METHODS FOR DISTRIBUTED LEARNING FOR WIRELESS EDGE DYNAMICS
2y 5m to grant Granted Mar 31, 2026
Patent 12585549
BACKING UP DATABASE FILES IN A DISTRIBUTED SYSTEM
2y 5m to grant Granted Mar 24, 2026
Patent 12585715
SYSTEMS AND METHODS FOR INDEPENDENT AUDIT AND ASSESSMENT FRAMEWORK FOR AI SYSTEMS
2y 5m to grant Granted Mar 24, 2026
Patent 12561572
METHOD FOR CALIBRATING PARAMETERS OF HYDROLOGY FORECASTING MODEL BASED ON DEEP REINFORCEMENT LEARNING
2y 5m to grant Granted Feb 24, 2026
Patent 12554774
GRAPH DATA LOADING
2y 5m to grant Granted Feb 17, 2026
Study what changed to get past this examiner. Based on 5 most recent grants.

AI Strategy Recommendation

Get an AI-powered prosecution strategy using examiner precedents, rejection analysis, and claim mapping.
Powered by AI — typically takes 5-10 seconds

Prosecution Projections

3-4
Expected OA Rounds
76%
Grant Probability
98%
With Interview (+22.0%)
3y 9m
Median Time to Grant
High
PTA Risk
Based on 610 resolved cases by this examiner. Grant probability derived from career allow rate.

Sign in with your work email

Enter your email to receive a magic link. No password needed.

Personal email addresses (Gmail, Yahoo, etc.) are not accepted.

Free tier: 3 strategy analyses per month