US 11,750,629 B2
Classification based anomaly detection
Sergey Serebryakov, Milpitas, CA (US); Tahir Cader, Liberty Lake, WA (US); and Nanjundaiah Deepak, Bangalore Karnataka (IN)
Assigned to Hewlett Packard Enterprise Development LP, Spring, TX (US)
Filed by Hewlett Packard Enterprise Development LP, Spring, TX (US)
Filed on Sep. 18, 2020, as Appl. No. 17/24,884.
Claims priority of application No. 201941047553 (IN), filed on Nov. 21, 2019.
Prior Publication US 2021/0160267 A1, May 27, 2021
Int. Cl. H04L 9/40 (2022.01); H04L 43/0817 (2022.01); H04L 43/065 (2022.01); H04L 43/062 (2022.01); H04L 43/04 (2022.01)
CPC H04L 63/1425 (2013.01) [H04L 43/04 (2013.01); H04L 43/062 (2013.01); H04L 43/065 (2013.01); H04L 43/0817 (2013.01); H04L 63/20 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A device, comprising:
processing circuitry; and
a memory including instructions that, when executed on the processing circuitry, cause the device to:
receive a plurality of datastreams from a plurality of sensors of a high performance computing system;
classify, based on a plurality of time-series observations of each sensor of the plurality of sensors, each datastream to one of a plurality of datastream models;
select, for each datastream and based on a classification of each datastream, an anomaly detection algorithm from a plurality of anomaly detection algorithms;
determine parameters of each anomaly detection algorithm of the plurality of anomaly detection algorithms, based on at least one of: characteristics of each datastream and characteristics of the high performance computing system;
receive an anomaly score from each of the plurality of anomaly detection algorithms;
determine, for each datastream an anomaly threshold based on each anomaly score;
generate, when an anomaly score of the plurality of anomaly scores exceeds its respective anomaly threshold, an indication that the sensor associated with the respective datastream is acting anomalously.