公开(公告)号：US12238119B1

公开(公告)日：2025-02-25

申请号：US17544538

申请日：2021-12-07

Applicant: Amazon Technologies, Inc.

Inventor： Radu Stefan Chivu ,  Daniel Lee Moor ,  Saad Ali Rana

IPC: H04L9/40 ,  H04L41/16

Abstract: Threats can be determined from anomalous events based on artificial intelligence (AI) models. For example, a computer system stores, based on an output of a first AI model, first information indicating that a first event cluster of the first event clusters is associated with a threat classification and that a second event cluster of the first event clusters is associated with a non-threat classification. The computer system receives a first dataset representing first events and generate a first input to the first AI model based on the first dataset. The computer system determines, based on the first input to the first AI model, second event clusters and that a third event cluster of the second event clusters has no correspondence in the first event clusters and is associated with an unknown classification. The computer system generates second information indicating that the third event cluster is associated with the unknown classification.