US 12,192,243 B2
Security policy selection based on calculated uncertainty and predicted resource consumption
Robson Pereira, Hortolândia (BR); Leandro Cesar Fida, Campinas (BR); Edson Jose Montanhini, Hortolândia (BR); Sergio Varga, Campinas (BR); and Daniele Jaqueline Marchiori, São Paulo (BR)
Assigned to Kyndryl, Inc., New York, NY (US)
Filed by Kyndryl, Inc., New York, NY (US)
Filed on Nov. 18, 2022, as Appl. No. 17/990,511.
Prior Publication US 2024/0171613 A1, May 23, 2024
Int. Cl. G06F 17/00 (2019.01); G06N 20/00 (2019.01); H04L 9/40 (2022.01); H04L 29/06 (2006.01)
CPC H04L 63/20 (2013.01) [G06N 20/00 (2019.01); H04L 63/105 (2013.01)] 17 Claims
OG exemplary drawing
 
1. A computer-implemented method, comprising:
receiving a request to perform a security policy implementation analysis for a first deployment associated with a first client in an IT environment;
collecting IT information associated with the first deployment;
applying trained machine learning models to analyze the IT information of the first client to compute a security policy for the first deployment,
wherein the security policy is computed based on a calculated uncertainty of effects that applying the security policy to the first deployment is capable of causing, and a predicted amount of resources of the first deployment that applying the security policy to the first deployment would consume;
outputting an indication of the security policy for display in a dashboard on a display of a user device of the first client;
training the machine learning models to analyze IT information; and
storing the trained machine learning models to a predetermined database, wherein training the machine learning models includes: retrieving IT information associated with a second deployment of a training IT environment; computing risk level errors from components of the second deployment and applications of the second deployment; transforming the IT information into training datasets for the machine learning models; training a first of the machine learning models using a first of the training datasets, wherein the first of the machine learning models is trained for calculating uncertainty; and training a second of the machine learning models using a second of the training datasets, wherein the second of the machine learning models is trained for predicting resource consumption.