Free and open-source medical device discovery and identification
Clone or download

README.md

Tapirx logo

Build Status

Tapirx ("taper ecks") passively discovers and identifies networked medical devices.

Tapirx is written in cross-platform Go and runs on Linux, macOS, and Windows. The following products already support Tapirx discovery out of the box:

Quick Start

Install Go, then install Tapirx:

$ go get github.com/virtalabs/tapirx

(See Windows instructions below if you run into trouble on Windows.)

Read a sample pcap file (included with this package) and output JSON data:

$ tapirx -pcap testdata/HL7-ADT-UDI-PRT.pcap -verbose

List your network interfaces, then sniff live traffic and display discovered device data:

$ tapirx -interfaces   # show a list of interfaces available for capture
$ tapirx -iface <interface_name> -verbose

Read on to understand how to share discovered asset information with other tools to fit your workflow.

Detailed Instructions

Finding Devices via a SPAN Port (a/k/a Port Mirroring)

In this configuration, a single network switchport will receive a copy of all traffic arriving at a set of switchports you want to monitor. A typical "managed" switch can copy traffic from all switchports on the switch, or just a subset of them. Ask your network administrator to configure a SPAN port (a/k/a a mirror port) on a switch or router, then connect your machine to that port with an Ethernet cable. If you run Wireshark or tcpdump on your machine's Ethernet interface, you should start to see traffic from the other switchports.

On the machine you've connected to the SPAN port, you should begin to see device information as the other connected devices on the switch generate HL7 or DICOM traffic:

$ tapirx -iface <your_ethernet_interface> -verbose

(On Windows, you may need to run tapirx -interfaces to find the appropriate interface name to pass to -iface, which will look like "\Device\NPF_{XXXXXXXX-XXXX-XXXX-XXXX-XXXXXXXXXXXX}".)

If you are looking for an inexpensive switch to experiment with, we suggest Netgear's inexpensive managed switches, such as the GS108E, which offers eight gigabit Ethernet ports and supports port mirroring via its web interface.

Finding Devices via Prerecorded Pcap Files

In this configuration, you will record and analyze a pcap file, which is the predominant format for network traffic captures.

To generate a Pcap file, run Wireshark on a machine that is connected to a SPAN port, then save the resulting capture(s) to pcap file(s). Instead of Wireshark, you can instead run tcpdump -w myfile.pcap. Or ask your network administrator to capture traffic on your behalf and share a pcap file with you.

You can then pass the pcap file directly to Tapirx instead of specifying a network interface:

$ tapirx -pcap myfile.pcap

Connecting Tapirx to Other Systems

Tapirx can share data about discovered devices with other systems. For example, you can:

  • Leave an instance running for a fixed amount of time to collect a snapshot of active inventory;
  • Continuously catalogue devices' MAC and IP addresses along with identification data to maintain a high level of preparedness for security incidents; or
  • Automatically highlight ePHI devices for HIPAA record keeping.

Device data can be automatically shared with any tool that offers a REST API. As it discovers devices and identifiers, Tapirx will issue POST requests that look like this (in this example, Tapirx spotted an infusion pump on 10.0.0.155 communicating with an EHR via HL7 on port 2575):

$ tapirx -apiurl https://my-asset-management.example.com/assets [-apitoken MY_SECRET]

The following is POSTed to https://my-asset-management.example.com/assets:

{
  "ipv4_address": "10.0.0.155",
  "ipv6_address": "",
  "open_port_tcp": "",
  "connect_port_tcp": "2575",
  "mac_address": "00:03:b1:b5:b6:48",
  "identifier": "Infuse-O-Matic Peach B+",
  "provenance": "HL7 PRT-10",
  "last_seen": "2019-01-02T12:37:22.938687-08:00",
  "client_id": "mymachine.example.com"
}

Alternatively, you can stream CSV output to a file using the -csv command-line option.

Run tapirx -help to see more usage information.

Building on Windows

Building Tapirx on Windows is somewhat more involved than on Linux or macOS, but it's within reach with a tiny bit of elbow grease:

(This procedure will become easier once gopacket supports npcap; see @google/gopacket#568.)

Building on Linux

You will need to install libpcap development headers, which fortunately are in most distributions' main repositories. After you have done so, the quick start instructions should work properly.

Ubuntu (or other Debian-based distributions):

$ sudo apt install libpcap-dev

RHEL7 or CentOS:

$ sudo yum install libpcap-devel

Tests and Benchmarking

Run go test github.com/virtalabs/tapirx to run the test suite.

Run go test -bench=. github.com/virtalabs/tapirx to run tests and performance benchmarks.

Frequently Asked Questions

How do I select specific network traffic to monitor?

Incoming data, whether live or from a file, can be filtered using BPF expressions. For example, if we want to sniff port 2575 (common for HL7 traffic):

$ tapirx -iface <interface_name> -bpf 'port 2575' -verbose

What does "discover" mean?

Tapirx inspects network traffic, either live or prerecorded, and extracts device information. Essentially, you can think of this tool as emitting a continuous stream of clues along the lines of "this device was spotted using the network at this moment." You can collect these clues, and over time you will discover a population of devices that use your network.

Tapirx does not discover devices that don't appear to be medical devices. Put another way, if a device doesn't use HL7, DICOM, or another protocol that Tapirx understands, it won't appear in Tapirx's announcements. This is by design, as medical devices merit special attention and there are plenty of other tools that can simply regurgitate MAC-to-IP mappings. (For example, if you run tcpdump arp for long enough, you'll get a sense of a network segment's population over time but you won't see information extracted from medical device protocols.)

To risk stating the obvious, Tapirx won't discover devices that never use the network.

What does "identify" mean?

Tapirx can help you answer the following questions about many devices:

  • Is this a medical device?
  • Does this device connect to an EHR or DICOM system?
  • What is this device's manufacturer and model?
  • What is this device's GUDID?

But it may not help with the following questions, which require information that is not generally found in network traffic:

  • What software is this device running?
  • What is this device's serial number?

What protocols can Tapirx currently inspect?

HL7 and DICOM. Support for more discovery methods and protocol fields is on the way. See the Contributing section to see how you can help.

What can't Tapirx do?

Tapirx can extract device identifiers from medical devices' network traffic, but it will not attempt to guess a device's identity if it cannot find relevant information.

One simple way to assign (somewhat reliable) manufacturer information to discovered devices is to connect Tapirx to a system that looks up MAC addresses in the public OUI database. Plenty of free and commercial tools fit the bill.

There are many, many medical devices out there. Contact the maintainers if you would like to see support for a specific kind of medical device that Tapirx does not adequately discover or identify.

Are there prebuilt releases so I don't have to build Tapirx myself?

Not yet, but building it yourself requires only one go get command; see above.

Why is Tapirx free?

Three reasons:

  • Discovery and identification of networked devices are crucial building blocks for cybersecurity, but they're not the most interesting parts. Democratizing this important function means organizations can move past initial stumbling blocks.
  • You deserve complete control over your data (such as ePHI). You shouldn't have to give a company or a proprietary data collector all of your proprietary traffic in order to catalogue your own devices.
  • Free tools scale better. Given the variety of devices in the world and the prohibitive cost of scaling proprietary tools in modern healthcare environments, the community needs free, open-source tools that aren't locked into a single vendor's product line.

Contributing

This is a community project, and contributions from the community are welcome! We have written a guide for contributing.

Authors

The original authors of this project are Andrew DeOrio (@awdeorio), Henrik Holm (@ukrutt), and Ben Ransford (@ransford) of Virta Labs.

This project would be far more complex if it weren't for the following open-source software: