SciTokens¶

SciTokens is a capability-based authorisation system whereby tokens are issued that grant access to perform specific actions on specific services (for example to query for segment information from the Segment Database).

What is a SciToken?¶

A SciToken is a special implementation of a JSON Web Token designed for distributed scientific computing. For full details on the SciTokens project and the implementation of the tokens, see

https://scitokens.org/

Key SciTokens concepts¶

Key to the IGWN usage of SciTokens are the following concepts:

Issuer¶

The Issuer of a token (encoded in the iss claim) is the address of the service that issued the token. Most services that accept tokens are configured to only accept tokens issued by specific issuers.

For all IGWN usage, the token issuer should be

https://cilogon.org/igwn

Audience¶

The Audience of a token (the aud claim) is the service that a token is authorised to acccess.

This should typically be the fully-qualified URI of the target service, e.g. https://gracedb.ligo.org.

The Audience can also have the special value ANY, which should allow it to be used with any service. Usage of ANY tokens is discouraged for security reasons, except in the case of authorising with CVMFS.

Scope¶

The Scope of a token (the scope claim) is a space-separated list of capabilities that the token should authorise. The valid scopes are typically specific to a given service. See Token-aware services below for details on the IGWN-specific scopes.

Installing the SciTokens tools¶

There are a few different tools to install based on usage:

• htgettoken - for generating SciTokens
• igwn-auth-utils - for discovering/using SciTokens with HTTP requests

Installing htgettoken¶

htgettoken is a tool for generating SciTokens via the IGWN Vault server.

The htgettoken tool can be installed using your preferred package manager on a number of systems:

conda install --channel conda-forge htgettoken

apt-get install htgettoken

yum install htgettoken

Installing igwn-auth-utils¶

igwn-auth-utils is a Python library for discovering SciTokens and using them with HTTP requests.

igwn-auth-utils can be installed using your preferred package manager on a number of systems:

conda install --channel conda-forge igwn-auth-utils

apt-get install python3-igwn-auth-utils

python -m pip install igwn-auth-utils

yum install python3-igwn-auth-utils

How to generate a SciToken¶

Generating a default token¶

LIGO.ORG and KAGRA identity holders can generate tokens using htgettoken:

htgettoken -a vault.ligo.org -i igwn

The token generated by htgettoken can be inspected using the httokendecode utility provided by the same software package:

$ httokendecode
{
  "sub": "marie.curie@ligo.org",
  "aud": "ANY",
  "ver": "scitoken:2.0",
  "nbf": 1681817886,
  "scope": "read:/frames gwdatafind.read gracedb.read dqsegdb.read",
  "iss": "https://cilogon.org/igwn",
  "exp": 1681828691,
  "iat": 1681817891,
  "jti": "https://cilogon.org/oauth2/adfghy549ij7gfd599453718c902af46?type=accessToken&ts=1681817891006&version=v2.0&lifetime=10800000"
}

By default this command will generate a token with the following claims

Generating a secure token¶

To generate a secure token, you should specify the --audience and --scopes as narrowly as possible to generate a token that with a single capability accepted by a single service.

For example, to generate a token that allows the bearer to read data from https://datafind.igwn.org only:

htgettoken \
  -a vault.ligo.org \
  --audience https://datafind.igwn.org \
  --scopes gwdatafind.read

This generates a token with only the required claims, and nothing else.

$ httokendecode -H
{
  "sub": "marie.curie@ligo.org",
  "aud": "https://datafind.igwn.org",
  "ver": "scitoken:2.0",
  "nbf": "Tue 18 Apr 2023 12:38:06 PM BST",
  "scope": "gwdatafind.read",
  "iss": "https://cilogon.org/igwn",
  "exp": "Tue 18 Apr 2023 03:38:11 PM BST",
  "iat": "Tue 18 Apr 2023 12:38:11 PM BST",
  "jti": "https://cilogon.org/oauth2/adfghy549ij7gfd599453718c902af46?type=accessToken&ts=1681817891006&version=v2.0&lifetime=10800000"
}

Generating a token using a robot Kerberos keytab¶

To support automated analyses, the IGWN Computing group can issue a robot credential that can be used to generate SciTokens without requiring interactive authorisation, or use of a personal identity.

To requests a robot credential to support SciTokens, please go to https://robots.ligo.org and select option 4. Apply for a SciToken keytab.

Once the keytab has been issued, you can use it to generate a SciToken by calling htgettoken with some extra arguments specific to your robot:

htgettoken \
  -a vault.ligo.org \
  --audience https://gracedb.ligo.org \
  --scopes gracedb.read \
  --role <my-robot-name> \
  --credkey <my-robot-credkey>

Here the values for the --role and --credkey arguments will be given to when the robot keytab is issued.

Setting default htgettoken options¶

To simplify the default operations for htgettoken, you can set the HTGETTOKENOPTS environment variable to encode the options you always give:

export HTGETTOKENOPTS="--vaultserver vault.ligo.org --issuer igwn"

setenv HTGETTOKENOPTS "--vaultserver vault.ligo.org --issuer igwn"

$Env:HTGETTOKENOPTS = "--vaultserver vault.ligo.org --issuer igwn"

Token-aware services¶

Proprietary IGWN data¶

Prorietary IGWN data are made available using the Open Science Data Federation with one 'origin' for each collaboration's h(t) data and one for shared derived/auxiliary data.

These data are then available over HTTP using the Pelican client or via CVMFS.

IGWN members can access these data using authorisation tokens with the appropriate scope according to which collaboration's data are being accessed:

DQSegDB¶

DQSegDB accepts the following claims:

GraceDB¶

GraceDB understands the following token scopes:

The gracedb.read scope is used as a proxy for an identification token - GraceDB uses its own internal authorisation system to assign capabilities to the subject (owner) of the token.

GWDataFind¶

GWDataFind accepts the following tokens:

Using tokens¶

Interactive use¶

For all interactive usage, tokens should be generated using htgettoken.

Once a token has been generated, the client tools for most token-aware services (including all of those listed above) will be able to automatically discover it by following the WLCG Bearer Token Discovery specification.

htgettoken --audience https://datafind.igwn.org --scope gwdatafind.read
python3 -m gwdatafind -r datafind.igwn.org -o H -t H1_HOFT_C00 --latest

HTCondor¶

For details on using SciTokens with HTCondor workflows, please see Using IGWN credentials with HTCondor.