Kollaborative Internet-Der-Dinge-Anomaliedetektion

Details

Veröffentlichungs-Nr.

EP3528457

Aktenzeichen

EP19157787

Anmeldetag

18. Februar 2019

Veröffentlichung

23. Oktober 2019

Rechtsraum

EP

IPC

H04L29/06

Abstract

A method for collaboration of remote participants in a distributed manner with no central server, on training an anomaly detection model, based on each of the local observations of the participants, wherein, a block-chain protocol is used where the blocks are the current anomaly detection models of a subset of participants. An agent configured to maintain a local model of each participant's application behavior is introduced to each of the participants where a first agent is use to observe the behavior of a participant application. The first agent records new intelligence by updating the local model with new observations and adds the updated local model to a partial block in a blockchain. The blockchain is sent to the agents of the other participants and each agent of the other participants verifies against its own partial blocks, that the received partial block is longer than its own partial block, and that the received partial block is safe. Then the local model of each agent of the other participants is replaced with the updated local model from the verified partial block.

Anmelder

Firma

Deutsche Telekom AG

Land

🇩🇪 Deutschland

Fachgebiete