US 12,278,832 B2
Detecting anomalous resources and events in social data using a trained anomaly detector
Neil Shah, Los Angeles, CA (US)
Assigned to Snap Inc., Santa Monica, CA (US)
Filed by Snap Inc., Santa Monica, CA (US)
Filed on Dec. 13, 2021, as Appl. No. 17/549,193.
Application 17/549,193 is a continuation of application No. 16/235,990, filed on Dec. 28, 2018, granted, now 11,212,303.
Prior Publication US 2022/0279001 A1, Sep. 1, 2022
Int. Cl. G06F 21/71 (2013.01); H04L 9/40 (2022.01); G06F 18/2411 (2023.01); G06F 21/00 (2013.01); G06Q 50/00 (2012.01)
CPC H04L 63/1433 (2013.01) [H04L 63/1425 (2013.01); G06F 18/2411 (2023.01); G06Q 50/01 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
receiving, by a processor, a plurality of events associated with a plurality of resources;
computing, in real-time, statistical representations for each resource associated with at least one of the plurality;
generating a reference window comprising a first portion of the plurality of events and a current window comprising a second portion of the plurality of events;
training an anomaly detector on the reference window using the statistical representations of resources in the first portion of the plurality of events;
evaluating, by the trained anomaly detector, the current window using the statistical representations of resources in the second portion of the plurality of events to identify a set of anomalous resources;
identifying a set of anomalous events associated with the set of anomalous resources; and
causing display, on a computing device, of an interface comprising the set of anomalous resources and the set of anomalous events.