Network monitoring of time synchronization protocols using convolutional neural networks

Abstract

A device and method monitor integrity of a communication network. A local clock maintains a local time. A network interface receives time synchronization packets and associates with each the local time of receipt at the network interface. A processing system implements a neural network for classifying whether integrity of the communication network is compromised from originate, receive, and transmit values determined from the time synchronization packets. The originate value is a difference between the local time of the receipt of the request packet and a transmission timestamp of the request packet. The receive value is a difference between a reception timestamp of the request packet or the reply packet and the local time of the receipt of the request packet or the reply packet. The transmit value is a difference between the local time of the receipt of the reply packet and a transmission timestamp of the reply packet.