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 .
Claims 1-20 are pending in this application.
Information Disclosure Statement
The information disclosure statement (IDS) submitted on 11/04/2024. The submission is in compliance with the provisions of 37 CFR 1.97. Accordingly, the information disclosure statement is being considered by the examiner.
Claim Objections
Claims 2, 5, 8, 12, 15 and 18 are objected to because of the following informalities:
As to claims 2 and 12, the limitation “associated with with the trained machine learning model” is indefinite due to duplication of “with”, making the scope unclear. It seems to be a typographical error. Duplicate “with” should be removed.
As to claims 5 and 15, the limitation “classificatiob” appears to be a typographical error; Examiner has interpreted as “classification”, for the purpose of examination.
As to clams 8 and 18, the limitation “uncertaintry” is indefinite. It also appears to be a typographical error; Examiner has interpreted as “uncertainty” for the purpose of examination.
Appropriate correction is required.
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.
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.
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 nonobviousness.
Claims 1-3, 5-7, 11-13 and 15-17 are rejected under 35 U.S.C. 103 as being unpatentable over Apostolescu et al. (US 10,708,296 B2) (hereinafter, “Apostolescu”) in view of Xie et al. (US 2015/0319136 A1) (hereinafter, “Xie”).
As to claim 1, Apostolescu discloses a method comprising steps of:
… analyzing the content with a trained machine learning model to provide an initial classification of benign or malicious (“… perform (220) anomaly detection on the extracted behaviors in the data set to classify the behaviors as anomalous or non-anomalous. This can be implemented by performing anomaly detection on all behaviors, aggregating all anomaly detection into a single “feature set” and collecting all anomalies into a large matrix.” -e.g., see, col. 6, lines 24-30; see also: “… training engine 605 can classify (620) behaviors of the multiple e-mails as anomalous or non-anomalous based on network data (e.g., network path information) of the e-mails, and iteratively generate (625) one or more classifiers using different feature sets of the classified behaviors per iteration until an accuracy threshold is met or exceeded. Once the accuracy threshold is met or exceeded, training engine 605 can output (630) the final one or more classifiers and the feature set of classified behaviors for classification of e-mails as a threat or non-threat.” -e.g., see, col. 8, lines 64-67 to col. 9, lines 1-6; herein, ML classifiers trained on execution-graph behaviors; anomalous or non-anomalous are equivalent to malicious or benign);
determining an uncertainty associated with the initial classification (“… in performing anomaly detection the system can plot ranges of values for metrics and calculate anomalous values (e.g., using algorithms such as “IQR”, “3-Sigma”) and detect anomalous patterns and activity detection (“IE Proxy changes in 80% of bad, 20% of good”).” -e.g., see, col. 4, lines 38-47; herein, “IQR”, “3-Sigma” =statistical distance-based uncertainty score); and
one of allowing the content, blocking the content, and sandboxing the content, based on the initial classification … (“In particular, training engine 105 can generate (200) an execution graph of each of the received files and extract (210) the behaviors from the execution graphs. This can be implemented in one embodiment by, for each of the received files (e.g., thousands): i) running the sample file in a sandbox and collecting its execution trace using known or custom tools” -e.g., see, col. 6, lines 6-12).
Apostolescu doesn’t explicitly disclose monitoring content inline between any of users, enterprises, and the Internet by a cloud-based system; one of allowing the content, blocking the content, …, based on … the uncertainty.
However, in an analogous art, Xie discloses monitoring content inline between any of users, enterprises, and the Internet by a cloud-based system (“… the network traffic is monitored using in-line monitoring techniques. In some embodiments, the network traffic is collected and/or monitored (e.g., some of the network traffic can be monitored using in-line monitoring techniques and/or some of the network traffic can be collected and analyzed for monitoring the network traffic offline, such as in logs of network traffic)” -e.g., see, [0034]; see also: “… integrated into a security appliance, firewall appliance, network/data appliance and/or executed on host device, such as a security server, network server or gateway, and/or client device (e.g., a personal computer, laptop, tablet, and/or other general purpose client device with sufficient processor and memory for executing a virtual machine). In some embodiments, the VM function for detecting malware for which preexisting signatures do not exist is provided be the security cloud service.” -e.g., see, [0046]; herein, Inline firewall monitoring of network traffic between users and enterprises and forwarded to VM/Cloud service);
one of allowing the content, blocking the content, and sandboxing the content, based on … the uncertainty (“Accordingly, a malware analysis system is disclosed. In particular, a malware analysis, using the various techniques described herein can provide zero-day protection, is provided, which protects against zero-day exploits by detecting new malware threats for which preexisting signatures do not exist and automatically generating new signatures in real-time for the new malware threats. For example, a virtual machine (VM) can be used to perform behavior profiling (e.g., in a VM sandbox environment) using various heurist based analysis techniques that can be performed in real-time during a file transfer (e.g., during a file download), and if the file being downloaded is determined to be malicious, then the firewall can automatically block the file download based on the analysis result, and a new signature can be generated and distributed to automatically block future file transfer requests to download the file determined to be malicious.” -e.g., see, Xie: [0026]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention was to modify the teaching of Apostolescu as taught by Xie in order to enable “zero-day protection…in real-time during a file transfer” (Xie: para. [0004]), improving Apostolescu’s offline training into scalable inline cloud security.
As to claim 11, it is rejected using the similar rationale as for the rejection of claim 1.
As to claim 2, Apostolescu in view of Xie discloses the method of claim 1, Apostolescu further discloses wherein the uncertainty is a distance between patterns in the content to patterns in a training dataset associated with the trained machine learning model (“… in performing anomaly detection the system can plot ranges of values for metrics and calculate anomalous values (e.g., using algorithms such as “IQR”, “3-Sigma”) and detect anomalous patterns and activity detection (“IE Proxy changes in 80% of bad, 20% of good”).” -e.g., see, col. 4, lines 38-47; herein, “IQR”, “3-Sigma” = distance from training-pattern medians).
As to claim 12, it is rejected using the similar rationale as for the rejection of claim 2.
As to claim 3, Apostolescu in view of Xie discloses the method of claim 2, Xie further discloses wherein the patterns are based on a plurality of features including raw byte ngram count and frequency, word frequency, entropy, and file size (“… a file size (e.g., file sizes forwarded for real-time VM emulation analysis can be limited to a maximum size, such as 1 GB, or another file size, to avoid too large of a file to emulate in real-time using VM techniques), obfuscated JavaScript (e.g., very long variables, long variables that include packed code using eval function, and/or other obfuscation techniques), and/or various other rules/policies for selection potentially suspicious malware appropriate for real-time VM emulation malware analysis.” -e.g., see, Xie: [0038]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention was to modify the teaching of Apostolescu as taught by Xie in order to provide richer malware signatures.
As to claim 13, it is rejected using the similar rationale as for the rejection of claim 3.
As to claim 5, Apostolescu in view of Xie discloses the method of claim 1, Xie further discloses wherein the sandboxing the content is based on the uncertainty being compared to thresholds for either the initial classification being benign or malicious (“… which protects against zero-day exploits by detecting new malware threats for which preexisting signatures do not exist and automatically generating new signatures in real-time for the new malware threats. For example, a virtual machine (VM) can be used to perform behavior profiling (e.g., in a VM sandbox environment) using various heurist based analysis techniques that can be performed in real-time during a file transfer (e.g., during a file download), and if the file being downloaded is determined to be malicious, then the firewall can automatically block the file download based on the analysis result” -e.g., see, Xie: [0026]; herein, sandbox triggered on uncertain score, see also, Xie: [0061]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention was to modify the teaching of Apostolescu as taught by Xie in order to enable “zero-day protection…in real-time during a file transfer” (Xie: para. [0004]), improving Apostolescu’s offline training into scalable inline cloud security.
As to claim 15, it is rejected using the similar rationale as for the rejection of claim 5.
As to claim 6, Apostolescu in view of Xie discloses the method of claim 1, Xie further discloses wherein the steps further include: providing a warning when the content is benign but the uncertainty is above a corresponding threshold (“… correlating the monitored and classified network traffic behaviors is performed, and a score (e.g., a severity score or malware score) is calculated based on the monitored/classified suspicious behaviors.” -e.g., see, Xie: [0061]; herein, severity score is referred to as score is above a threshold equal to warning for benign but suspicious).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention was to modify the teaching of Apostolescu as taught by Xie in order to enable “zero-day protection…in real-time during a file transfer” (Xie: para. [0004]), improving Apostolescu’s offline training into scalable inline cloud security.
As to claim 16, it is rejected using the similar rationale as for the rejection of claim 6.
As to claim 7, Apostolescu in view of Xie discloses the method of claim 1, Xie further discloses wherein the steps further include: sandboxing the content when the initial classification is malicious and the uncertainty is above a corresponding threshold (“… which protects against zero-day exploits by detecting new malware threats for which preexisting signatures do not exist and automatically generating new signatures in real-time for the new malware threats. For example, a virtual machine (VM) can be used to perform behavior profiling (e.g., in a VM sandbox environment) using various heurist based analysis techniques that can be performed in real-time during a file transfer (e.g., during a file download), and if the file being downloaded is determined to be malicious, then the firewall can automatically block the file download based on the analysis result” -e.g., see, Xie: [0026]; herein, sandbox triggered on uncertain score, see also, Xie: [0061]).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention was to modify the teaching of Apostolescu as taught by Xie in order to enable “zero-day protection…in real-time during a file transfer” (Xie: para. [0004]), improving Apostolescu’s offline training into scalable inline cloud security.
As to claim 17, it is rejected using the similar rationale as for the rejection of claim 7.
Claims 4 and 14 are rejected under 35 U.S.C. 103 as being unpatentable over Apostolescu in view of Xie as applied to claims 1 and 11 above, and further in view of Brown (US 2018/0121650 A1).
As to claim 4, Apostolescu in view of Xie discloses the method of claim 1, Although Apostolescu discloses a threshold-based decision on the initial classification (“Training engine 105 can generate (230) one or more classifiers based on a particular feature set of the classified behaviors. If the one or more classifiers do not meet or exceed (240) an accuracy threshold…” -e.g., see, col. 7, lines 15-19; herein, accuracy threshold=uncertainty threshold for allow (benign)/block (malicious)), Apostolescu in view of Xie doesn’t explicitly disclose wherein the allowing the content is based on the initial classification being benign and the uncertainty being compared to a corresponding threshold, and the blocking the content is based on the initial classification being malicious and the uncertainty being compared to another corresponding threshold.
However, in an analogous art Brown discloses wherein the allowing the content is based on the initial classification being benign and the uncertainty being compared to a corresponding threshold, and the blocking the content is based on the initial classification being malicious and the uncertainty being compared to another corresponding threshold (“… the event analysis module 136 can calculate a confidence score that a process is malicious based at least in part on one or more of the directory traversal event(s) 104b and/or the file access pattern event(s) 106b and/or the file access record 114. In that case, the event analysis module 136 can create a malicious process detection event 138 if a determined confidence metric exceeds a score threshold (e.g., the score threshold can be set by the remote security service 118 and/or the policy 130 in some examples). In additional or alternate examples, the event analysis module 136 can ascertain how well an observed process action 104a and/or 106 accords with directory traversal as specified by the policy 130 and/or a DAP § 132.” -e.g., see, Brown: [0055]; herein, Brown teaches dual-threshold allow/block logic based on classification and uncertainty (i.e. confidence metric)).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention was to modify the teaching of Apostolescu and Xie as taught by Brown in order to enable real-time, policy-driven security actions in a cloud/inline environment while reducing false positive/negatives.
As to claim 14, it is rejected using the similar rationale as for the rejection of claim 4.
Claims 8-10 and 18-20 are rejected under 35 U.S.C. 103 as being unpatentable over Apostolescu in view of Xie as applied to claims 1 and 11 above, and further in view of Miserendino et al. (US 10,121,108 B2) (hereinafter, “Miserendino”).
As to claim 8, Apostolescu in view of Xie discloses the method of claim 1, Apostolescu in view of Xie doesn’t explicitly disclose wherein the steps further include: monitoring the trained machine learning model in the cloud-based system over time including the determined uncertainty for corresponding initial classifications; and triggering retraining of the trained machine learning model based on the monitoring.
However, in an analogous art, Miserendino discloses wherein the steps further include: monitoring the trained machine learning model in the cloud-based system over time including the determined uncertainty for corresponding initial classifications; and triggering retraining of the trained machine learning model based on the monitoring (“With continuing reference to FIG. 1, the original third-party created base classifier becomes the first parent classifier in the repeatable in-situ process. Method 100 uses this base classifier to evaluate unknown content on the user network predicting a class (e.g., benign or malicious) for each sample, block 6. In an embodiment, a user may use system graphical user interface (GUI) to inspect some or all of the predicted classes and determines whether the sample is truly benign or malicious (e.g., confirms or corrects classifications), block 7. In another aspect of this disclosure, the in-situ retraining system may inspect, without human intervention, some or all of the predicted classes and determine whether the sample is benign or malicious (e.g., confirms or corrects classifications). The act of confirming or correcting the classification is called adjudication. In an embodiment, a retrain manager (e.g., instantiated as a retrain manager service) monitors the user's adjudication activity and determines when a sufficient number of in-situ samples have been adjudicated” -e.g. see, Miserendino: col. 5, lines 26-44; herein, monitors confidence/uncertainty over time; triggers retrain).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention was to modify the teaching of Apostolescu and Xie as taught by Miserendino in order to reduce false positives in a cloud pipeline.
As to claim 18, it is rejected using the similar rationale as for the rejection of claim 8.
As to claim 9, Apostolescu in view of Xie and Miserendino discloses the method of claim 8, Miserendino further discloses wherein the steps further include: identifying gaps in training data for the trained machine learning model based on the monitoring; and updating the training data based on the identified gaps prior to the retraining (“… the retrain manager creates a training and test set from the adjudicated in-situ samples, block 8. Alternatively, the in-situ retraining system may, without human intervention, initiate a retrain. The training and test set may be selected from a subset of the adjudicated samples. Retrain manager may also extract feature vectors from both retrain training and test sets, block 9.” -e.g., see, Miserendino: col. 5, lines 53-59; herein, adjudication identifies gaps; updates data before retrain).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention was to modify the teaching of Apostolescu and Xie as taught by Miserendino in order to reduce false positives in a cloud pipeline.
As to claim 19, it is rejected using the similar rationale as for the rejection of claim 9.
As to claim 10, Apostolescu in view of Xie discloses the method of claim 1, Apostolescu in view of Xie does not explicitly disclose wherein the steps further include: subsequent to sandboxing the content, determining whether the initial classification was correct or not for later processing with the trained machine learning model.
However, in an analogous art, Miserendino discloses wherein the steps further include: subsequent to sandboxing the content, determining whether the initial classification was correct or not for later processing with the trained machine learning model (“… In an embodiment, a user may use system graphical user interface (GUI) to inspect some or all of the predicted classes and determines whether the sample is truly benign or malicious (e.g., confirms or corrects classifications), block 7. In another aspect of this disclosure, the in-situ retraining system may inspect, without human intervention, some or all of the predicted classes and determine whether the sample is benign or malicious (e.g., confirms or corrects classifications). The act of confirming or correcting the classification is called adjudication. In an embodiment, a retrain manager (e.g., instantiated as a retrain manager service) monitors the user's adjudication activity and determines when a sufficient number of in-situ samples have been adjudicated.” -e.g., see, Miserendino: col. 5, lines 26-44; herein, review corrects for retraining).
Therefore, it would have been obvious to one of ordinary skill in the art before the effective filing date of the invention was to modify the teaching of Apostolescu and Xie as taught by Miserendino in order to reduce false positives in a cloud pipeline.
As to claim 20, it is rejected using the similar rationale as for the rejection of claim 10.
Conclusion
The prior art made of record and not relied upon is considered pertinent to applicant's disclosure.
Stokes et al. (US 2009/0099988 A1) – Stokes teaches entries are automatically classified into one of multiple classes. After classifying the entry, a generated model for the determined class is utilized to determine how well an entry corresponds to the model. Ambiguous classifications along with entries that do not fit the model well for the determined class are selected for labeling by a human analyst The selected entries are presented to a human analyst for labeling. These labels are used to further train the classifier and the models. During an evaluation phase, entries are automatically classified using the trained classifier and a policy associated with determined class is applied. -e.g., see, Fig. 9,10, Abstract, [0079] and [0080] of Stokes.
Any inquiry concerning this communication or earlier communications from the examiner should be directed to SUMAN DEBNATH whose telephone number is (571)270-1256. The examiner can normally be reached Mon-Fri; 9:00am-5:00pm.
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, Farid Homayounmehr can be reached at 571-272-3739. 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.
SUMAN DEBNATH
Patent Examiner
Art Unit 2495
/S.D/Examiner, Art Unit 2495
/JEFFERY L WILLIAMS/Primary Examiner, Art Unit 2495
