Using categorization tags for rule generation and update in a rules-based security system

Patent metadata

Field	Value
Publication number	US-12495075-B2
Application number	US-202318107729-A
Country	US
Kind code	B2
Filing date	Feb 9, 2023
Priority date	Feb 9, 2023
Publication date	Dec 9, 2025
Grant date	Dec 9, 2025

Abstract

A technique for classifying and handling threat data in a rules-based security system. For each rule in the set, a set of one or more first tags are generated. The tags categorize the rule according to a hierarchical scheme. In response to receipt of a new threat, the system automatically determines whether the existing set of rules provide an acceptable coverage for the new threat. This determination is made by generating a set of one or more second tags that categorize the new threat, and then comparing the set of one or more second tags with the set of one or more first tags according to given match criteria. Upon a determination that the set of rules do not provide an adequate coverage for the new threat, a recommendation is output from the system. The rules-based security system is then adjusted according to the recommendation for subsequent handling of the new threat.

First claim

The invention claimed is:

1 . A method of classifying and handling threat data in a rules-based security system that applies a set of rules to search for and detect anomalies, comprising:
for each of one or more rules in the set, generating a set of one or more first tags that categorize the rule according to a hierarchy using a natural lage processing (NLP) machine learning model; enabling a reviewer to accept, update or add tags to the set of one or more first tags: responsive to receipt of a new threat, determining whether the set of rules provide an acceptable coverage for the new threat by generating a set of one or more second tags with a multi-label classifier that categorize the new threat according to an attack type, and comparing the set of one or more second tags with the set of one or more first tags to generate confidence score indicating a level of coverage for the new threat; based on the confidence score gen d by the comparison of the set of one or more second tags with the set of one or more first tags, outputting a recommendation, the recommendation including at least one selected from the group consisting of maintaining the current set of rules without change, identifying a list of groups of rules in the set of rules that should be combined to increase a percentage of coverage, identifying a new rule to be added to the set of rules, and identifying an update to at least one existing rule in the set of rules; and adjusting the rules-based security system according to the recommendation for subsequent handling of the new threat.

2 . The method as described in claim 1 wherein adjusting the rules-based security system includes one of: creating a new rule and adding the new rule to the set of rules: and identifying an existing rule for update and updating the existing rule.

3 . The method as described in claim 1 wherein the confidence score identifies a percent coverage for the new threat provided by the set of rules.

4 . The method as described in claim 1 wherein each of the first and second tags are generated using the natural language processing (NLP) machine learning model.

5 . The method as described in claim 4 further including retraining the NLP machine learning model based at least in part on feedback collected from one or more users.

6 . The method as described in claim 1 wherein comparing the set of one or more second tags with the set of one or more first tags includes identifying any partial or complete tag matches, and identifying the one or more rules of the set of rules that correspond to the partial or complete tag matches.

7 . The method as described in claim 1 wherein the set of one or more first tags includes one of: tags from a pre-defined list, and one or more custom tags.

8 . An apparatus, comprising:
a hardware processor; computer memory holding computer program instructions executed by the hardware processor for classifying and handling threat data in a rules-based security system that applies a set of rules to search for and detect anomalies, the computer program instructions comprising program configured to:
for each of one or more rules in the set, generate a set of one or more first tags that categorize the rule according to a hierarchy using a natural language pro cessing (NLP) machine learning model;
enable a reviewer to accept, update or add tags to the set of one or more first tags;
responsive to receipt of a new threat, determine whether the set of rules provide an acceptable coverage for the new threat by generating a set of one or more second tags with a multi-label classifier that categorize the new threat according to an attack type, and comparing the set of one or more second tags with the set of one or more first tags to generate a confidence score indicating a level of coverage for the new threat;
based on the confidence score generated by the comparison of the set of one or more second tags with the set of one or more first tags, output a recommendation, the recommendation including at least one selected from the group consisting of maintaining the current set of rules without change, identifying a list of groups of rules in the set of rules that should be combined to increase a percentage of coverage, identifying a new rule to be added to the set of rules, and identifying an update to at least one existing rule in the set of rules; and
adjust the rules-based security system according to the recommendation for subsequent handling of the new threat.

9 . The apparatus as described in claim 8 wherein the program code configured to adjust the rules-based security system includes program code configured to perform one of: creating a new rule and adding the new rule to the set of rules: and identifying an existing rule for update and updating the existing rule.

10 . The apparatus as described in claim 8 wherein the confidence score identifies a percent coverage for the new threat provided by the set of rules, and an identification of a list of groups of the set of rules that should be combined to increase the percent coverage.

11 . The apparatus as described in claim 8 wherein the program code configured to generate each of the first and second tags comprises the language processing (NLP) machine learning model.

12 . The apparatus as described in claim 11 wherein the program code is further configured to retrain the NLP machine learning model based at least in part on feedback collected from one or more users.

13 . The apparatus as described in claim 8 wherein the program code configured to compare the set of one or more second tags with the set of one or more first tags includes program code configured to identify any partial or complete tag matches, and to identify the one or more rules of the set of rules that correspond to the partial or complete tag matches.

14 . The apparatus as described in claim 8 wherein the set of one or more first tags includes one of: tags from a pre-defined list, and one or more custom tags.

15 . A computer program product in a non-transitory computer readable medium, the computer program product holding computer program instructions executed by a hardware processor for classifying and handling threat data in a rules-based security system that applies a set of rules to search for and detect anomalies, the computer program instructions comprising program configured to:
for each of one or more rules in the set, generate a set of one or more first tags that categorize the rule according to a hierarchy using a natural language p ocessing (NLP) machine earning model; enable a reviewer to accept, update or add tags to the set of one or more first tags; responsive to receipt of a new threat, determine whether the set of rules provide an acceptable coverage for the new threat by generating a set of one or more second tags with a multi-label classifier that categorize the new threat according to an attack type, and comparing the set of one or more second tags with the set of one or more first tags to generate a confidence score indicating a level of coverage for the new threat; based on the confidence score generated by the comparison of the set of one or more second tags with the set of one or more first tags, output a recommendation, the recommendation including at least one selected from the group consisting of maintaining the current set of rules without change, identifying a list of groups of rules in the set of rules that should be combined to increase a cen age of coverage, identifying a new nule to be added to the set of rules, and identifying an update to at least one existing rule in the set of

Assignees

• IBM

Inventors

• Grout Juliet
• Arunkumar Saritha
• Nagar Seema
• Yuceer Ilgen Banu

Classifications

• G06F16/285

  Clustering or classification · CPC title

• H04L63/1425 Primary

  Traffic logging, e.g. anomaly detection · CPC title

• H04L63/20 Primary

  for managing network security; network security policies in general (filtering policies H04L63/0227) · CPC title