Overview

The purpose of this project is to make educated guesses about the types of hosts behind a gateway where NAT (Network Address Translation) is in use. We plan to apply to some machine learning model to accomplish this.

What is Network Address Translation?

NAT is a process by which one IP address (typically that of your network gateway) represents all the hosts in some network. NAT was devised to address the IPv4 address shortage, but it has the ancillary benefit of hiding network hosts from would be attackers.

How Was Data Collected?

tcpdump was run on OpenWRT devices deployed at various locations, chief among them WAN ports. This data consisted of packet headers, timestamps, etc. This data was collected using a Shell script deployed to these devices.

Parts of this project

• inf folder contains the terraform and ansible files for a blueprint of our AWS network that we used for data collection. Please refer to the README of it for detailed usage.
• data_util folder contains the tools that we used for data process before they can be used by machine learning models. Please refer to the README of it for detailed usage.
• dl folder contains the model-related file, including the model itself and tools for training and prediction. Please refer to the README of it for detailed usage.