CVE-2026-44180: Jupyter Enterprise Gateway: root privilege bypass in Kubernetes

GHSA-chq7-94j8-cj28 CRITICAL
Published June 3, 2026
CISO Take

Jupyter Enterprise Gateway versions 2.0.0rc1 through 3.2.x contain a trivially exploitable input validation flaw that completely defeats the prohibited UID/GID security control: appending a trailing space to the value '0 ' bypasses the string-match check, while the downstream Jinja2 template correctly parses it as integer 0, launching a Kubernetes kernel pod as root. With no authentication required (CVSS 9.8, AV:N/AC:L/PR:N/UI:N) and 1,872 downstream dependents, this affects a large share of enterprise ML platforms built on Jupyter — any multi-tenant AI/data science environment on Kubernetes is at full cluster compromise risk. Root access inside the container, combined with Kubernetes hostPath volume mounts (which Enterprise Gateway exposes through kernel pod templating), provides a documented and reproducible path to crontab injection on the worker node, from which lateral movement across all cluster workloads follows. Upgrade to version 3.3.0 immediately; if patching is not possible within 24 hours, restrict network access to the Enterprise Gateway API endpoint to trusted internal sources only and audit current kernel pod specs for unexpected UID 0 entries.

Sources: NVD GitHub Advisory ATLAS OpenSSF

What is the risk?

Severity is critical and exploitation is trivial — the bypass requires only appending a whitespace character to the KERNEL_UID parameter, no special tooling or AI/ML expertise needed. The CVSS 9.8 score reflects network-accessible exploitation with no privileges or user interaction required. In Kubernetes-deployed Enterprise Gateway environments (the primary deployment target), successful exploitation grants root inside a container with API-controlled volume mount capabilities, making container escape straightforward via hostPath mounts. The 1,872 downstream dependents and the broad enterprise adoption of Jupyter for AI/ML workloads amplifies the blast radius significantly. The OpenSSF Scorecard of 5.2/10 and 19 prior CVEs in the same package indicate a pattern of security hygiene concerns in this project. Not currently in CISA KEV, but the trivial PoC and high impact make near-term weaponization likely.

Attack Kill Chain

API Access
Attacker with network access to the Enterprise Gateway service sends a kernel launch POST request to /api/kernels with KERNEL_UID and KERNEL_GID set to '0 ' (trailing whitespace).
AML.T0049
Validation Bypass
The string '0 ' does not match '0' in the prohibited UIDs list (string comparison), so the security check passes; the Jinja2 template then casts '0 ' to integer 0, scheduling the pod as root.
Root Kernel Execution
Kubernetes schedules a kernel pod running as UID 0/GID 0 (root); attacker gains an interactive root shell via the Jupyter kernel API with full filesystem access inside the container.
AML.T0050
Container Escape & Cluster Compromise
Attacker mounts a hostPath volume pointing to the worker node filesystem, writes a root crontab entry to gain persistent code execution on the host, then pivots to the Kubernetes API using the node's service account to compromise the entire cluster.
AML.T0105

What systems are affected?

Package Ecosystem Vulnerable Range Patched
jupyter_enterprise_gateway pip >= 2.0.0rc1, < 3.3.0 3.3.0
13.2K OpenSSF 5.2 1.9K dependents Pushed 5d ago 77% patched ~12d to patch Full package profile →

Do you use jupyter_enterprise_gateway? You're affected.

Severity & Risk

CVSS 3.1
9.8 / 10
EPSS
N/A
Exploitation Status
No known exploitation
Sophistication
Trivial

Attack Surface

AV AC PR UI S C I A
AV Network
AC Low
PR None
UI None
S Unchanged
C High
I High
A High

What should I do?

5 steps
  1. PATCH

    Upgrade jupyter_enterprise_gateway to version 3.3.0 immediately — this is the only complete fix.

  2. DETECT

    Query your Kubernetes cluster for running pods with securityContext.runAsUser=0 or securityContext.runAsGroup=0 that were spawned by Enterprise Gateway: kubectl get pods -A -o jsonpath='{range .items[*]}{.metadata.name}{" "}{.spec.containers[*].securityContext.runAsUser}{"\n"}{end}' | grep ' 0$'.

  3. WORKAROUND (if patching is delayed): Restrict network access to the Enterprise Gateway service (port 8888) to trusted internal subnets only via NetworkPolicy or firewall rules. Add input sanitization at the ingress/proxy layer to strip whitespace from KERNEL_UID and KERNEL_GID values.

  4. HARDEN

    Enforce PodSecurityAdmission with baseline or restricted policy in namespaces where Enterprise Gateway schedules kernels to prevent root pod scheduling at the cluster level.

  5. AUDIT

    Review kernel pod spec templates for unexpected volume mounts, especially hostPath types.

Classification

Compliance Impact

This CVE is relevant to:

EU AI Act
Article 15 - Accuracy, robustness and cybersecurity
ISO 42001
A.6.2.6 - AI system security
NIST AI RMF
MANAGE 2.2 - Mechanisms are in place to respond to AI risks
OWASP LLM Top 10
LLM08 - Excessive Agency

Frequently Asked Questions

What is CVE-2026-44180?

Jupyter Enterprise Gateway versions 2.0.0rc1 through 3.2.x contain a trivially exploitable input validation flaw that completely defeats the prohibited UID/GID security control: appending a trailing space to the value '0 ' bypasses the string-match check, while the downstream Jinja2 template correctly parses it as integer 0, launching a Kubernetes kernel pod as root. With no authentication required (CVSS 9.8, AV:N/AC:L/PR:N/UI:N) and 1,872 downstream dependents, this affects a large share of enterprise ML platforms built on Jupyter — any multi-tenant AI/data science environment on Kubernetes is at full cluster compromise risk. Root access inside the container, combined with Kubernetes hostPath volume mounts (which Enterprise Gateway exposes through kernel pod templating), provides a documented and reproducible path to crontab injection on the worker node, from which lateral movement across all cluster workloads follows. Upgrade to version 3.3.0 immediately; if patching is not possible within 24 hours, restrict network access to the Enterprise Gateway API endpoint to trusted internal sources only and audit current kernel pod specs for unexpected UID 0 entries.

Is CVE-2026-44180 actively exploited?

No confirmed active exploitation of CVE-2026-44180 has been reported, but organizations should still patch proactively.

How to fix CVE-2026-44180?

1. PATCH: Upgrade jupyter_enterprise_gateway to version 3.3.0 immediately — this is the only complete fix. 2. DETECT: Query your Kubernetes cluster for running pods with securityContext.runAsUser=0 or securityContext.runAsGroup=0 that were spawned by Enterprise Gateway: `kubectl get pods -A -o jsonpath='{range .items[*]}{.metadata.name}{" "}{.spec.containers[*].securityContext.runAsUser}{"\n"}{end}' | grep ' 0$'`. 3. WORKAROUND (if patching is delayed): Restrict network access to the Enterprise Gateway service (port 8888) to trusted internal subnets only via NetworkPolicy or firewall rules. Add input sanitization at the ingress/proxy layer to strip whitespace from KERNEL_UID and KERNEL_GID values. 4. HARDEN: Enforce PodSecurityAdmission with baseline or restricted policy in namespaces where Enterprise Gateway schedules kernels to prevent root pod scheduling at the cluster level. 5. AUDIT: Review kernel pod spec templates for unexpected volume mounts, especially hostPath types.

What systems are affected by CVE-2026-44180?

This vulnerability affects the following AI/ML architecture patterns: Jupyter kernel deployments on Kubernetes, Multi-tenant ML development platforms, MLOps training pipelines with interactive kernels, Enterprise data science platforms.

What is the CVSS score for CVE-2026-44180?

CVE-2026-44180 has a CVSS v3.1 base score of 9.8 (CRITICAL).

AI Security Impact

Affected AI Architectures

Jupyter kernel deployments on KubernetesMulti-tenant ML development platformsMLOps training pipelines with interactive kernelsEnterprise data science platforms

MITRE ATLAS Techniques

AML.T0010.001 AI Software
AML.T0049 Exploit Public-Facing Application
AML.T0050 Command and Scripting Interpreter
AML.T0105 Escape to Host

Compliance Controls Affected

EU AI Act: Article 15
ISO 42001: A.6.2.6
NIST AI RMF: MANAGE 2.2
OWASP LLM Top 10: LLM08

Technical Details

Original Advisory

### Summary Jupyter Enterprise Gateway has a prohibited UID and GID feature that by default prevents launching kernels with UID or GID 0 (root). This can be bypassed. It is possible to launch kernels with a prohibited UID and/or GID by using a specially crafted `KERNEL_UID` or `KERNEL_GID` value. The feature is described in the documentation: https://github.com/jupyter-server/enterprise_gateway/blob/152c20f162f2fab700c04c8830ebf8c1e2e2217a/docs/source/operators/config-add-env.md?plain=1#L103-L107 https://github.com/jupyter-server/enterprise_gateway/blob/152c20f162f2fab700c04c8830ebf8c1e2e2217a/docs/source/operators/config-add-env.md?plain=1#L88-L92 https://github.com/jupyter-server/enterprise_gateway/blob/152c20f162f2fab700c04c8830ebf8c1e2e2217a/docs/source/operators/deploy-kubernetes.md?plain=1#L769 ### Details The `prohibited_uids` and `prohibited_uids` are set based of the OS env var `EG_PROHIBITED_UIDS` and `EG_PROHIBITED_GIDS`, and default to the string `0`. https://github.com/jupyter-server/enterprise_gateway/blob/152c20f162f2fab700c04c8830ebf8c1e2e2217a/enterprise_gateway/services/processproxies/container.py#L29-L30 The checks https://github.com/jupyter-server/enterprise_gateway/blob/152c20f162f2fab700c04c8830ebf8c1e2e2217a/enterprise_gateway/services/processproxies/container.py#L113 and https://github.com/jupyter-server/enterprise_gateway/blob/152c20f162f2fab700c04c8830ebf8c1e2e2217a/enterprise_gateway/services/processproxies/container.py#L119 look for the user supplied `KERNEL_UID` / `KERNEL_GID` string in the `prohibited_uids` / `prohibited_gids` strings. These checks can be bypassed by including whitespace, for example the string `0 ` (trailing space). The user supplied string is used in the Kubernetes manifest at https://github.com/jupyter-server/enterprise_gateway/blob/152c20f162f2fab700c04c8830ebf8c1e2e2217a/etc/kernel-launchers/kubernetes/scripts/kernel-pod.yaml.j2#L35 and https://github.com/jupyter-server/enterprise_gateway/blob/152c20f162f2fab700c04c8830ebf8c1e2e2217a/etc/kernel-launchers/kubernetes/scripts/kernel-pod.yaml.j2#L38 where they are parsed as an integer in the Jinja2 template - which will ignore the whitespace. ### PoC #### How it is meant to work Trying `0` gets denied, as expected. ```bash xh http://enterprise-gateway.bdawg.svc.cluster.local:8888/api/kernels name=python_kubernetes env:='{"KERNEL_POD_NAME":"bdawg", "KERNEL_UID": "0", "KERNEL_GID": "0"}' ``` ``` HTTP/1.1 403 Kernel's UID value of '0' has been denied via EG_PROHIBITED_UIDS! Content-Length: 94 Content-Type: application/json Date: Mon, 14 Jul 2025 12:57:09 GMT Server: TornadoServer/6.4.1 X-Content-Type-Options: nosniff ``` ```json { "reason": "Kernel's UID value of '0' has been denied via EG_PROHIBITED_UIDS!", "message": "" } ``` #### Exploit bypassing the checks Using `0 ` with a trailing space, bypasses the check. ```bash xh http://enterprise-gateway.bdawg.svc.cluster.local:8888/api/kernels name=python_kubernetes env:='{"KERNEL_POD_NAME":"bdawg", "KERNEL_UID": "0 ", "KERNEL_GID": "0 "}' ``` ``` HTTP/1.1 201 Created Content-Length: 172 Content-Type: application/json Date: Mon, 14 Jul 2025 14:15:19 GMT Location: /api/kernels/17eee032-994f-4dd2-8ade-87169c300a40 Server: TornadoServer/6.4.1 X-Content-Type-Options: nosniff ``` ``` { "id": "17eee032-994f-4dd2-8ade-87169c300a40", "name": "python_kubernetes", "last_activity": "2025-07-14T14:15:21.468155Z", "execution_state": "starting", "connections": 0 } ``` The pod is successfully scheduled. Inspecting the container we can see it is running as `root`: ```bash kubectl exec -it pod/bdawg -- bash ``` ``` (base) root@bdawg3:~# id uid=0(root) gid=0(root) groups=0(root),100(users) ``` If we had not supplied the `KERNEL_UID` / `KERNEL_GID` the container would have been running as UID:GID `1000:100` (`jovyan:users`). ### Impact This input validation vulnerability allows running Jupyter kernels as root, which can be dangerous as it allows more attack surface, and may lead to container escapes, compromising the worker node and all workloads running on it. Repeated exploitation can compromise all worker nodes, and thus the entire Kubernetes cluster. It is possible to specify volume mounts, so one vector for a container escape is to use a `hostPath` R/W volume mount, use this UID/GID bypass to run as root, and then gain code execution in the underlying worker node by creating a crontab entry in the mounted host file system. Organisations running Jupyter Enterprise Gateway to host Jupyter Kernels on at least Kubernetes clusters (I've tested this), and possibly on any other supported container orchestration systems or systems that utilise the `KERNEL_UID` and `KERNEL_GID` variables with the `EG_PROHIBITED_UIDS` and `EG_PROHIBITED_GIDS` feature.

Exploitation Scenario

An adversary with network access to an Enterprise Gateway deployment — whether an insider, a compromised CI/CD pipeline, or an attacker who enumerated the internal Kubernetes service endpoint — sends a single HTTP POST to `/api/kernels` with `KERNEL_UID: '0 '` and `KERNEL_GID: '0 '` (trailing space). The security check compares the string '0 ' against the prohibited list containing '0', finds no match, and proceeds. The Jinja2 template then casts '0 ' to integer 0 when writing the Kubernetes pod manifest, scheduling the kernel container as root. The attacker connects to the root shell via the kernel API, mounts a hostPath volume pointing to `/etc/cron.d` on the host, and writes a cron job that executes a reverse shell or downloads a persistence implant. This gives the attacker code execution on the underlying worker node with root privileges, from which they can read secrets mounted in other pods, pivot to the Kubernetes API server using the node's service account credentials, and ultimately compromise the entire cluster including all AI model artifacts, training data, and API keys stored as secrets.

CVSS Vector

CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H

Timeline

Published
June 3, 2026
Last Modified
June 3, 2026
First Seen
June 4, 2026

Related Vulnerabilities