Jupyter Security - Jupyter Foundation Community Funding Reports

Status: In Progress · Proposal

September 2026¶

Summary of Work Completed¶

This month was focused on a triage catchup (20 advisories + 4 emails), mostly closing a lot of duplicates. We have a few vulnerabilities that are reported dozen of times due to AI.

The counterpart is that I had no time to write fixes, although I reached with two subprojects about open vulnerabilities.

I put together some statistics in JupyterLab: https://jupyter.zulipchat.com/#narrow/channel/538444-Security.5BPublic.5D/topic/Statistics/near/619738010 - the code has been pushed on the public security repository https://github.com/jupyter/security/tree/main/lab

I am drafting a new funding proposal with Carreau to continue this proposal after December

I was elected as the software steering security representative. Thanks to Jason Grout for handling the documentations updates consecutive to this!

Deliverables/Milestones¶

Documentation:

6 pull requests to group security policies to meta-repositories (org/.github/SECURITY.md) and remove the old email security@ipython.org (jupyter/security#125, jupyter/security#124)

Processed GHSAs:¶

Closed as duplicates:

Closed as not a vulnerability:

Accepted

Published

Statistics for August¶

Advisories in triage

Currently in triage: 11 (median age: 28 days)
New advisories opened this month: 20 (+ 4 by email)
Opened this month and still in triage: 7 (+ 0 by email)

Accepted advisories waiting for a fix

Advisories in draft: 56 (median age 135.5 days)
Advisories in draft older than 90 days: 34

Published advisories

All time: 84 (median time to publication 32 days)
This month: 4 (median time to publication 33 days)

Challenges or Risks¶

Backlog of accepted advisories waiting to be fixed is high and not decreasing.

Reports keep flowing in at high volumes.

Budget Update¶

Next Steps¶

Long term initiatives, not yet started:

Ownership initiative: map every repo to a subproject to be able to contact someone in charge when needed.
Labeling initiative: find a way to “flag” unmaintained projects with security risks.
Think about writing a blog post about security, because these reports probably don’t get read a lot by the users.

Appendix¶

Script used for statistics¶

See August 2026 report.

August 2026¶

Summary of Work Completed¶

11 reports were triaged including 3 by email, and fixes were written or reviewed for four of the accepted ones.

22 new advisories were opened across the Jupyter orgs in July + 8 emails received at security@jupyter.org.

Jupyterlab desktop got its first release in two years to address a security issue that got published in a blog article.

Ten advisories were published this month, most of them in jupyterlab.

Deliverables/Milestones¶

Reports accepted and turned into advisories:

https://github.com/jupyter-server/jupyter_server/security/advisories/GHSA-c3mw-737p-c7g2: fix written and merged
https://github.com/jupyterhub/binderhub/security/advisories/GHSA-q276-fxp7-xhx3: triaged, initially accepted then closed.
https://github.com/jupyterhub/jupyter-server-proxy/security/advisories/GHSA-569q-87m2-93rh: opened, patch has been opened by the maintainer and needs to be merged.
jupyterlab-contrib/jupyter-archive#163, fix written and merged

Reports triaged and closed:

Published :

Supply chain / tooling

Improvements to github-security-overview (tool to aggregate GHSAs across multiple orgs): https://github.com/Yann-P/github-security-overview and subsequently updated jupyter/cve#45

Statistics for July¶

Advisories in triage

Currently in triage: 17 (median age: 29 days)
Opened this month: 12
New advisories opened this month, all states: 22

Accepted advisories waiting for a fix

Advisories in draft: 50 (median age 186 days)
Advisories in draft older than 90 days: 30

Published advisories

All time: 80 (median time to publication 30 days)
This month: 10 (median time to publication 62.5 days)

Definitions

“age” : difference between current date and initial creation of the advisory. Does not take into account the date where the advisory was accepted (transition from triage to draft), because this information is not in the GH API.
“time to publication” = time between the initial creation of the advisory and publication date.

Challenges or Risks¶

Incoming volume is still the main risk: 22 new advisories opened this month. A large share of the reports still looks AI-generated and low signal.

The draft backlog is still very large.

Budget Update¶

Next Steps¶

Follow up on the fixes opened this month
Publish key indicators on jupyter.org for transparency (carried over from June).
Document the advisory publishing checklist in jupyter/security

Appendix¶

Script used for statistics¶

using https://github.com/Yann-P/github-security-overview @ 111a70467f8076740629a8c92fbafa233a9640df

JUPYTER_ORGS=(binder-examples binderhub-ci-repos ipython jupyter jupyter-attic jupyter-book jupyter-governance jupyter-incubator jupyter-resources jupyter-server jupyter-standard jupyter-standards jupyter-widgets jupyter-xeus jupytercon jupyterhealth jupyterhub jupyterlab jupyterlite pickleshare voila-dashboards voila-gallery)
FROM_DATE=2026-07-01
TO_DATE=2026-08-01

security-overview "${JUPYTER_ORGS[@]}" --state triage | wc -l
security-overview "${JUPYTER_ORGS[@]}" --state triage | datamash median 1
security-overview "${JUPYTER_ORGS[@]}" --state triage --opened-from $FROM_DATE --opened-to $TO_DATE | wc -l
# All states, to measure incoming volume
security-overview "${JUPYTER_ORGS[@]}" --opened-from $FROM_DATE --opened-to $TO_DATE | wc -l

security-overview "${JUPYTER_ORGS[@]}" --state draft | wc -l
security-overview "${JUPYTER_ORGS[@]}" --state draft | datamash median 1
# Open for more than 90 days
security-overview "${JUPYTER_ORGS[@]}" --state draft --opened-to $(date -v-90d +%Y-%m-%d) | wc -l

security-overview "${JUPYTER_ORGS[@]}" --state published | wc -l
security-overview "${JUPYTER_ORGS[@]}" --state published | datamash median 3 # 3 is the column for time to publication

security-overview "${JUPYTER_ORGS[@]}" --state published --published-from $FROM_DATE --published-to $TO_DATE | wc -l
security-overview "${JUPYTER_ORGS[@]}" --state published --published-from $FROM_DATE --published-to $TO_DATE | datamash median 3

July 2026¶

Summary of Work Completed¶

This month was focused on supply chain and measurements included below, and maintenance about the email addresses used to reach us and PGP keys. A few fixes. Issues in triage are stacking up and I need to prioritize this for the beginning of July.

Deliverables/Milestones¶

Automated inventories for supply chain security using GitHub actions: jupyter/cve#52, jupyter/cve#41, jupyter/cve#44
Removed PGP keys from all documentation since the private key is lost.
Initiated transition from security@ipython.org to security@jupyter.org, tracked in jupyter/security#124 and jupyter/security#125
Fixed and published https://github.com/jupyterlab/jupyterlab/security/advisories/GHSA-vmhf-c436-hxj4
Fixed and published https://github.com/jupyter-server/jupyter_server/security/advisories/GHSA-fcw5-x6j4-ccmp
Reviewed fix for https://github.com/jupyterlab/jupyterlab-git/security/advisories/GHSA-436q-jwfr-rm2h
Bi-weekly security meetings attended: June 2, 16, 30

Statistics for June¶

Advisories in triage

Currently in triage: 17 (median age: 21 days)
Opened this month: 15

Accepted advisories waiting for a fix

Advisories in draft: 51 (median age 275 days)
Advisories in draft older than 90 days: 31

Published advisories

All time: 70 (median time to publication 27 days)
This month: 7 (median time to publication 64 days)

Definitions

“age” : difference between current date and initial creation of the advisory. Does not take into account the date where the advisory was accepted (transition from triage to draft), because this information is not in the GH API.
“time to publication” = time between the initial creation of the advisory and publication date.

Challenges or Risks¶

Lots of incoming reports, lots of them seem to be AI.

Budget Update¶

Next Steps¶

Publish key indicators on jupyter.org for transparency

Catch up on triage (lots of reports lately).

Appendix¶

Script used for statistics¶

using https://github.com/Yann-P/github-security-overview @ 5bb654593f6bbbd700a108b59c6fef5873edfea3

JUPYTER_ORGS=(binder-examples binderhub-ci-repos ipython jupyter jupyter-attic jupyter-book jupyter-governance jupyter-incubator jupyter-resources jupyter-server jupyter-standard jupyter-standards jupyter-widgets jupyter-xeus jupytercon jupyterhealth jupyterhub jupyterlab jupyterlite pickleshare voila-dashboards voila-gallery)
FROM_DATE=2026-06-01
TO_DATE=2026-07-01

security-overview "${JUPYTER_ORGS[@]}" --state triage | wc -l
security-overview "${JUPYTER_ORGS[@]}" --state triage | datamash median 1

security-overview "${JUPYTER_ORGS[@]}" --state draft | wc -l
security-overview "${JUPYTER_ORGS[@]}" --state draft | datamash median 1
# Open for more than 90 days
security-overview "${JUPYTER_ORGS[@]}" --state draft --opened-to $(date -v-90d +%Y-%m-%d) | wc -l

security-overview "${JUPYTER_ORGS[@]}" --state published | wc -l
security-overview "${JUPYTER_ORGS[@]}" --state published | datamash median 3 # 3 is the column for time to publication

security-overview "${JUPYTER_ORGS[@]}" --state published --published-from $FROM_DATE --published-to $TO_DATE | wc -l
security-overview "${JUPYTER_ORGS[@]}" --state published --published-from $FROM_DATE --published-to $TO_DATE | datamash median 3

June 2026¶

Context: contract with Yann-P started on 2026-06-01 (2 days ago) for the next 6 months at 8h/week as part of jupyter-governance/funding-proposals#28.

Summary of Work Completed¶

As volunteer work before the contract started

Built a tool to oversee GHSAs over all the jupyter repositories: https://github.com/Yann-P/github-security-overview
Triaged ~20 GHSAs
Cleaned up spam in security@jupyter.org
Released jupyter_server v2.18.0 with security fixes

After start of contract

Tool to find supply chain security flaws across all jupyter repos (in jupyter/cve)

Deliverables/Milestones¶

As volunteer work before the contract started

Published advisories

as remediation developer:
- https://github.com/advisories/GHSA-37w4-hwhx-4rc4 / CVE-2026-42266
- https://github.com/advisories/GHSA-24qx-w28j-9m6p / CVE-2026-40110
- https://github.com/advisories/GHSA-5789-5fc7-67v3 / CVE-2026-35397
as remediation reviewer:
- https://github.com/advisories/GHSA-mqcg-5x36-vfcg / CVE-2026-42557
as coordinator:
- https://github.com/advisories/GHSA-qh7q-6qm3-653w / CVE-2025-61669
- https://github.com/advisories/GHSA-rch3-82jr-f9w9 / CVE-2026-40171
- https://github.com/advisories/GHSA-5mrq-x3x5-8v8f / CVE-2026-40934

And a few others that are not yet published.

After start of contract

Supply chain inventory scripts in jupyter/cve

Detect repos that need github actions security (zizmor, actionlint)
Detect repos that need trusted publishing

Challenges or Risks¶

Budget Update¶

Next Steps¶

Quantify everything (# of open GHSAs, # of repos that needs supply chain hardening, ...) and track progress over time.
More tools to oversee supply chain security across all repos
Triage new GHSAs as they are reported.
If I have time, work on fixes.