Skip to content
/ revtrvp Public
forked from NEU-SNS/revtrvp

Code for reverse traceroute vantage points running on M-Lab

Notifications You must be signed in to change notification settings

m-lab/revtrvp

 
 

Repository files navigation

Introduction

This is the documenation for using the REVTR 2.0 system. You can either run reverse traceroutes from destinations back to a M-Lab source using the API, or back to your own source by adding it to the REVTR 2.0 system.

API

Currently, we give access to our RESTful API provided that you contacted us at [email protected] and we gave you an API key.

curl -X POST -k -H "Revtr-Key: <your-api-key>" https://revtr.ccs.neu.edu/api/v1/revtr --data '{"revtrs":[{"src":"<source-ip-address>", "dst":"<destination-ip-address>", "label":"<label>"}]}'

The label will server to retrieve your measurements, so put something like a unique ID and/or your name.

To find the available sources, you can use this query:

curl -X GET -k -H "Revtr-Key: <your-api-key>" https://revtr.ccs.neu.edu/api/v1/sources

Fetching the results of your reverse traceroutes

We are uploading all the finished reverse traceroutes to M-Lab archive every 10 minutes, so you should wait about 10-15 minutes after you started your last reverse traceroute measurement before it gets uploaded. Then, you can download your results here on this M-Lab public archive. You can postfilter your reverse traceroutes using the filter that you provided.

Interpreting the results of your reverse traceroutes

The results in the M-Lab archive are the jsonl extension, with one reverse traceroute per line. An example is:

{"id": 20525538, "stop_reason": "REACHES", "fail_reason": "", "src": "193.142.125.51", "dst": "91.224.181.97", "runtime": 1113501997, "date": 1669835097, "label": "bgp_survey_no_timestamp_test_load", "revtr_hops": [{"hop_number": 0, "hop_ip": "91.224.181.97", "hop_type": 1, "measurement_id": 0, "rtt": 18365, "rtt_measurement_id": 1774719956, "cidr": null, "asn": null}, {"hop_number": 1, "hop_ip": "87.245.249.121", "hop_type": 5, "measurement_id": 1774718488, "rtt": 18251, "rtt_measurement_id": 1774719952, "cidr": null, "asn": null}, {"hop_number": 2, "hop_ip": "87.245.225.176", "hop_type": 5, "measurement_id": 1774718488, "rtt": 50144, "rtt_measurement_id": 1774719944, "cidr": null, "asn": null}, {"hop_number": 3, "hop_ip": "80.231.65.2", "hop_type": 5, "measurement_id": 1774718488, "rtt": 870, "rtt_measurement_id": 1774719949, "cidr": null, "asn": null}, {"hop_number": 4, "hop_ip": "185.11.76.45", "hop_type": 4, "measurement_id": 2905588, "rtt": 23231, "rtt_measurement_id": 1774719951, "cidr": null, "asn": null}, {"hop_number": 5, "hop_ip": "185.11.76.76", "hop_type": 4, "measurement_id": 2905588, "rtt": 22508, "rtt_measurement_id": 1774719957, "cidr": null, "asn": null}, {"hop_number": 6, "hop_ip": "213.242.112.49", "hop_type": 4, "measurement_id": 2905588, "rtt": 18089, "rtt_measurement_id": 1774719946, "cidr": null, "asn": null}, {"hop_number": 7, "hop_ip": "4.69.159.46", "hop_type": 4, "measurement_id": 2905588, "rtt": 22129, "rtt_measurement_id": 1774719945, "cidr": null, "asn": null}, {"hop_number": 8, "hop_ip": "4.68.74.110", "hop_type": 4, "measurement_id": 2905588, "rtt": 18752, "rtt_measurement_id": 1774719953, "cidr": null, "asn": null}, {"hop_number": 9, "hop_ip": "5.23.30.17", "hop_type": 4, "measurement_id": 2905588, "rtt": 296, "rtt_measurement_id": 1774719955, "cidr": null, "asn": null}, {"hop_number": 10, "hop_ip": "193.142.125.51", "hop_type": 4, "measurement_id": 2905588, "rtt": 22, "rtt_measurement_id": 1774719954, "cidr": null, "asn": null}]}
{"id": 20525506, "stop_reason": "REACHES", "fail_reason": "", "src": "173.205.3.25", "dst": "182.75.124.57", "runtime": 1095979593, "date": 1669835097, "label": "bgp_survey_no_timestamp_test_load", "revtr_hops": [{"hop_number": 0, "hop_ip": "182.75.124.57", "hop_type": 1, "measurement_id": 0, "rtt": 274242, "rtt_measurement_id": 1774720375, "cidr": null, "asn": null}, {"hop_number": 1, "hop_ip": "203.101.87.155", "hop_type": 5, "measurement_id": 1774718425, "rtt": 263928, "rtt_measurement_id": 1774720373, "cidr": null, "asn": null}, {"hop_number": 2, "hop_ip": "173.205.3.25", "hop_type": 5, "measurement_id": 1774718425, "rtt": 34, "rtt_measurement_id": 1774720530, "cidr": null, "asn": null}]}

Fields of a reverse traceroute measurement

id id of the reverse traceroute measurement
src source of the reverse traceroute measurement (e.g., an M-Lab source or your own source)
dst destination of the reverse traceroute measurement
stop_reason stopping reason of the measurement (REACHES, or FAILED)
runtime time to measure the reverse path (in ns)
date starting date of the measurement (UNIX timestamp)
label label of the measurement

Hop types

The hop types and number are described in the next table. For more details, please refer to the corresponding sections of our REVTR 2.0 IMC 2022 paper

Type Name Description
1 Destination The hop of the destination
2 Assume symmetry This hop was found by running a forward traceroute to the current previous hop and assumed symmetry on the penultimate hop, i.e. the penultimate hop was the next reverse hop
3 Intersected traceroute This hop was found in a traceroute that was intersected by the last hop of type != 3. The intersection was exactly the last hop or an alias of the last hop (Sec.2, Intersecting a traceroute)
4 Intersected Record Route atlas This hop was found in a Record Route hop revealed by our new technique to reveal Record Route interfaces from the traceroute atlas (Sec. 4.1, Q2 and Sec. 4.2). See more details about how to process these hops here
5 Record Route This hop was found using Record Route (Sec.2, Record Route)
6 Spoofed Record Route This hop was found using spoofed Record Route (Sec.2, Record Route)

Details about how to interpret hops of type 2, assumed symmetric

We found in our paper that assuming symmetry on the penultimate hop of a forward traceroute is correct in 57% of the cases if the link on which we assume symmetry is interdomain and 90% for intradomain (Sec 4.4). To let you the possibility to consider or filter out these measurements, we still return the paths measured by REVTR 2.0 with these symmetry assumptions. What we suggest though, is that you run your IP to AS mapping on the path and only keep the paths that consider trustworthy. In the paper, we only kept those with intradomain assumptions of symmetry.

Details about how to interpret hops of type 4, Intersected Record Route atlas

When a reverse traceroute intersects the Record Route atlas, as we have incomplete alias information, we might not know exactly where the reverse traceroute intersected the traceroute. For instance, if the traceroute in atlas was VP -> T1 -> T2 -> T3 -> T4 -> S, where VP and S are our vantage point and our source, then our technique issued Record Route pings from S (or spoofed as S) to T1, T2, T3, and T4. Let us say the ping to T1 reveals R1, and other reveal nothing. If we know that T1 is an alias of R1, then, when a reverse traceroute intersects R1, we can say the rest of the path is T1 (or R1) -> T2 -> T3 -> T4 -> S. But if we do not have this information, it could be that R1 is an alias of T2, or even T3, if T1 and T2 do not stamp Record Route packets. In that case, we can only say that a future reverse traceroute intersecting in R1 intersected somewhere between T1 and T4. What you will have in the data in a list of hops of type 4, corresponding to the segment of the traceroute where the reverse traceroute could have intersected. A trick that we used to narrow down the size of the potential intersected segment is to map IP addresses to their AS, and remove any AS loop from the AS path.

Setting up a Reverse Traceroute source

This is the documentation for adding a source to the Reverse Traceroute system. Hosting a source is required to measure a reverse traceroute, as we need to receive measurement probes sent toward the source to splice together hops into a complete route. Once set-up, a source can be used to measure reverse traceroutes from arbitrary destinations.

Prerequisites

First of all, by joining the system, you are allowing our vantage points to use your source IP address to send packets such that the responses go back to your source. Your source will listen for responses to the spoofed ICMP packets (spoofed as your source IP address).

The source should either have a public IP address or, if it is behind a NAT, should be forwarded all ICMP packets arriving at the NAT. The ICMP packets received by a Reverse Traceroute source contain IP Record Route options, so the source's hosting network (and upstream networks) should not filter out or drop packets with IP options.

Build the Docker container

The Reverse Traceroute software is packaged as a Docker container. You will need to install Docker on your machine following instructions for your operating system and distribution.

The source's configuration is in the plvp.config YAML file. You should not to need to change anything, but we discuss two parameters:

  • local.interface is the name of the interface inside the container where packets will be received. The Reverse Traceroute source will listen for ICMP packets arriving on this interface and report them to the central controller. You may need to change this to the name of the interface on the machine that will run the container. (More on this below.)
  • scamper.rate controls the maximum probing rate for the source. The default is 100 pps, which should be enough for running tens of concurrent reverse traceroutes. Large measurement campaigns running a larger number of concurrent measurements require increasing the probing rate. (Of course, check with your upstream networks and notify the Reverse Traceroute operators before increasing the probing rate.)
  • scamper.port is the local port used by the source to receive commands from the central controller. You can keep it at the default value. This port needs to be forwarded into the container when running it.
  • local.host, local.pprofaddr, scamper.host point to the central Reverse Traceroute controller. These values can be kept fixed, but would need to be updated when connecting a source to an alternate Reverse Traceroute deployment.

You can build the Docker container running docker build -t <tag> . inside the repository's root directory. Set a tag to more easily launch the container later.

Add the source to the system

The Reverse Traceroute controller maintains a list of authorized sources. To add your source to the system, you will need to email [email protected]. Provide the public IP address, a description of where that IP is hosted, your name, and your email, and a brief statement of why you want to add your source to the system. Prefer to use an institutional e-mail, which helps operators clear new sources.

After your source is cleared, Reverse Traceroute operators will e-mail you informing that your source's IP address has been added to the set of authorized sources. The e-mail will also contain an API key to launch reverse traceroute measurements.

Run the Reverse Traceroute source

As explained under "Prerequisites" above, a Reverse Traceroute source needs to receive responses to spoofed ICMP probes sent by other vantage points in the Reverse Traceroute system. However, Docker runs containers behind a NAT by default, which will interfere with receiving the responses to spoofed ICMP probes. You can work around this issue in different ways, the easiest is to just run the Docker container with --net=host, which will attach the container on the same network namespace as the host's. This will allow the Docker container to access the host's interface directly and capture any ICMP responses. (You may need to change the local.interface configuration parameter in plvp.config to point the host's interface, as explained under "Build the Docker container" above.)

Run the following command to start the container on the host's namespace. The <tag> parameter is the one you chose when building the container, and 4381 is the value of scamper.port in the plvp.config configuration file.

docker run --name=<name> --net=host --restart=unless-stopped \
        --detach --log-opt max-size=1g --log-opt max-file=1 \
        --publish 4381:4381
        <tag> /root.crt /plvp.config -loglevel debug

This will launch your container, which will connect to the controller on startup. Whenever your source connects to the system, the controller tests if your source is capable of receiving packets containing the IP Record Route option as well as sending pings and traceroutes. An e-mail will be sent to you after the tests complete with a report on whether your sources can receive reverse traceroutes. If your source can receive reverse traceroutes, you will also receive a second e-mail stating that your source is ready after it's been integrated into the system.

Bootstrapping the traceroute atlas to your source

To benefit from the full functionality of Reverse Traceroute, we need to bootstrap your source by building an atlas of forward traceroutes towards it from public vantage points, then running Record Route pings to the traceroute hops revealed by traceroutes toward your source to support IP aliasing.

If this is the first time your source is being added to the system, we provide the RIPE Atlas credits for the traceroutes. You can use the following REST request to run measurements to update the atlas towards your source:

curl -X POST \
        -H 'Revtr-Key: <your-api-key>' \
        -H source:<your-source-ip-address> \
        https://<controller-hostname>/api/v1/atlas/run

Where <your-api-key> is the key you received over e-mail from the Reverse Traceroute operators. The <controller-hostname> is just the hostname of the server running the central controller, but anonymized for double-blind review.

For subsequent refreshes of the atlas to your source, you need to provide a RIPE Atlas API key. This key should have permissions to create traceroute requests. Refreshing the atlas to a source uses around 60K RIPE Atlas credits.

curl -X POST \
        -H 'Revtr-Key: <your-api-key>' \
        -H source:<your-source-ip-address> \
        -H 'RIPE-Key: <your-ripe-atlas-key>' \
        https://<controller-hostname>/api/v1/atlas/run

Refreshing the atlas involves waiting for RIPE Atlas traceroutes and Record Route pings. It currently takes around 20 minutes.

Running reverse traceroutes toward your source

You can send a REST API request to the /api/v1/revtr endpoint to issue a reverse traceroute toward your source. You should pass your API key and a JSON object specifying the measurements. The payload should be a JSON object with a revtrs key containing a list of source-destination pairs. Each source-destination pair should be specified as a JSON object with src and dst keys. In the example below, a single reverse traceroute would be issued from 1.1.1.1 towards your source:

curl -X POST -k -H "Revtr-Key: <your-api-key>" https://<controller-hostname>/api/v1/revtr --data '{"revtrs":[{"src":"<your-source-ip-address>", "dst":"1.1.1.1"}]}'

About

Code for reverse traceroute vantage points running on M-Lab

Resources

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Go 98.2%
  • Other 1.8%