US 11,924,049 B2
Network performance metrics anomaly detection
Abdolreza Shirvani, Ottawa (CA); Elizabeth Keddy, Ottawa (CA); Glenda Ann Leonard, Carp (CA); and Christopher Daniel Fridgen, Kanata (CA)
Assigned to ACCEDIAN NETWORKS INC., Saint-Laurent (CA)
Filed by Accedian Networks Inc., Saint-Laurent (CA)
Filed on Nov. 25, 2022, as Appl. No. 17/994,097.
Application 17/994,097 is a continuation of application No. 17/384,195, filed on Jul. 23, 2021, granted, now 11,539,573.
Application 17/384,195 is a continuation of application No. 15/929,956, filed on May 29, 2020, granted, now 11,108,621, issued on Aug. 31, 2021.
Prior Publication US 2023/0092829 A1, Mar. 23, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. H04L 41/142 (2022.01); G06F 18/22 (2023.01); H04L 9/40 (2022.01); H04L 41/0631 (2022.01)
CPC H04L 41/142 (2013.01) [G06F 18/22 (2023.01); H04L 41/064 (2013.01); H04L 41/065 (2013.01); H04L 63/1425 (2013.01)] 16 Claims
OG exemplary drawing
 
1. A method for detecting anomalies in one or more time series relating to one or more performance measures for one or more monitored objects in one or more networks comprising:
selecting, by a processor, a discrete window on one of said one or more time series to extract a first motif for a first performance measure of said one or more performance measures for a first monitored object of said one or more monitored objects;
maintaining, by the processor, an abnormal cluster center and a normal cluster center, from a binary clustering of one or more historical time series for said first performance measure for said first monitored object;
classifying, by the processor, said first motif based on a distance between said first motif and said abnormal cluster center and said normal cluster center; and
determining, by the processor, whether an anomaly for said first performance measure for said first monitored object occurred based on said distance and a predetermined decision boundary.