US 12,470,569 B2
Anomaly detection relating to communications using information embedding
Edmond J. Abrahamian, Richmond Heights, MO (US); Andrew Campbell, Pleasant Hill, CA (US); Ana Armenta, San Jose, CA (US); and Prince Paulraj, Coppell, TX (US)
Assigned to AT&T Intellectual Property I, L.P., Atlanta, GA (US)
Filed by AT&T Intellectual Property I, L.P., Atlanta, GA (US)
Filed on Nov. 24, 2021, as Appl. No. 17/456,520.
Prior Publication US 2023/0164150 A1, May 25, 2023
Int. Cl. H04L 9/40 (2022.01)
CPC H04L 63/1416 (2013.01) 20 Claims
OG exemplary drawing
 
1. A method, comprising:
embedding, by a system comprising a processor, respective properties associated with a user identity of a first user and respective relationships between the respective properties to generate an embedded array comprising bits of data representative of the respective properties and the respective relationships between the respective properties, wherein each cell of a plurality of cells in the embedded array is for representing one of the respective properties or one of the respective relationships, wherein a first cell of the plurality of cells has a first position in the embedded array that is reserved to refer to a first property of the respective properties, wherein a second cell of the plurality of cells has a second position in the embedded array that is reserved to refer to a second property of the respective properties, and wherein a third cell of the plurality of cells has a third position in the embedded array that is reserved to refer to a first relationship between the first property and the second property, wherein the first relationship comprises a spatial relationship, and wherein a value in the third cell is associated with a measure of the spatial relationship;
determining, by the system, a pattern associated with the respective properties and the respective relationships between the respective properties based on a first analysis of the embedded array; and
detecting, by the system, an anomaly in the pattern based on a second analysis of the pattern, wherein the anomaly relates to an event associated with the user identity, and wherein the second analysis is performed by a trained model that was trained on respective embedded arrays associated with other users that do not include the first user.