CVE-2026-48026: lakeFS Web UI XSS Vulnerability

CVE-2026-48026 Overview

CVE-2026-48026 is a stored cross-site scripting (XSS) vulnerability in lakeFS, an open-source tool that transforms object storage into Git-like repositories. The lakeFS Web UI renders Markdown files stored as repository objects without sanitizing the resulting HTML. Any user with write access to a repository branch can commit a .md object that contains arbitrary HTML or JavaScript. When another authenticated user opens that object or navigates to a repository containing a malicious README.md, the attacker-supplied script executes in the victim's authenticated session. The issue affects lakeFS open source prior to v1.81.1 and lakeFS Enterprise prior to v1.84.0.

Critical Impact
Attackers with write access can hijack authenticated sessions of any user who views a malicious Markdown file, enabling repository takeover and data theft.

Affected Products

lakeFS open source edition prior to v1.81.1
lakeFS Enterprise edition prior to v1.84.0
lakeFS Web UI Markdown rendering component (useMarkdownProcessor.tsx)

Discovery Timeline

2026-08-07 - CVE-2026-48026 published to the National Vulnerability Database (NVD)
2026-08-12 - Last updated in NVD database

Technical Details for CVE-2026-48026

Vulnerability Analysis

The vulnerability is a stored cross-site scripting flaw [CWE-79] in the lakeFS Web UI. The frontend component responsible for rendering Markdown files, implemented in useMarkdownProcessor.tsx, converts Markdown into HTML but does not sanitize the resulting output before inserting it into the DOM. Markdown supports inline HTML by design, so a committed .md object can contain <script> tags, event handlers such as onerror, or <iframe> payloads. Because rendering occurs inside an authenticated session, injected JavaScript runs with the victim's privileges against the lakeFS API. An attacker with write access to any branch can plant a malicious README.md that automatically executes when other users browse the repository or directory.

Root Cause

The root cause is missing output sanitization in the Markdown-to-HTML pipeline. The processor passes rendered HTML directly to the browser without applying an allowlist-based sanitizer such as DOMPurify. User-supplied Markdown content is therefore treated as trusted, violating the principle of context-aware output encoding.

Attack Vector

Exploitation requires an authenticated attacker with write privileges to at least one branch and user interaction from the victim, who must view the object or navigate to a directory containing the malicious file. The scope is changed because scripts execute in the security context of any viewing user, allowing the attacker to escalate access beyond their own permissions. See the GitHub Security Advisory GHSA-c2r5-5j4w-2xff and the GitHub Markdown Processor Code for technical details.

// No verified exploit code is published. The vulnerability manifests when
// a committed .md object contains inline HTML such as <script>, <img onerror=...>,
// or <iframe srcdoc=...> payloads that the Web UI renders without sanitization.

Detection Methods for CVE-2026-48026

Indicators of Compromise

Markdown objects (.md, README.md) in lakeFS repositories containing inline <script>, <iframe>, <object>, or event-handler attributes such as onerror, onload, or onclick.
Unexpected outbound HTTP requests from browsers to attacker-controlled domains initiated shortly after users viewed a lakeFS repository page.
lakeFS API calls originating from browser sessions that do not match the acting user's normal workflow, such as unexpected token generation or permission changes.

Detection Strategies

Scan repository object stores for Markdown files containing HTML tags or JavaScript event handlers using regex-based content inspection.
Review lakeFS audit logs for commits that add or modify .md files, especially from accounts with recently granted write access.
Correlate Web UI page loads with anomalous API activity from the same session to identify script-driven abuse.

Monitoring Recommendations

Forward lakeFS access logs and object commit events to a centralized analytics platform for continuous inspection.
Monitor for creation of new access tokens, credential exports, or role changes initiated through the Web UI shortly after Markdown views.
Alert on browser telemetry showing script execution from lakeFS origins to unrelated third-party endpoints.

How to Mitigate CVE-2026-48026

Immediate Actions Required

Upgrade lakeFS open source to v1.81.1 or later, and lakeFS Enterprise to v1.84.0 or later.
Audit repositories for existing .md files that contain embedded HTML or JavaScript and remove or rewrite malicious content.
Review recent write-permission grants and revoke access for accounts that should not modify repository objects.

Patch Information

The maintainers fixed the issue by sanitizing rendered Markdown output. lakeFS open source addresses the vulnerability in v1.81.1, and lakeFS Enterprise addresses it in v1.84.0. Refer to the GitHub Security Advisory GHSA-c2r5-5j4w-2xff for release details.

Workarounds

Enterprise customers on older versions can temporarily disable Markdown rendering by adding the appropriate directive to their lakeFS YAML configuration, per the vendor advisory.
No workaround exists for the open-source edition; upgrading is the only supported remediation.
Restrict branch write permissions to trusted users until the upgrade is completed to reduce the pool of potential attackers.

bash

# Enterprise-only temporary mitigation: disable Markdown rendering in lakeFS config
# Refer to the vendor advisory for the exact YAML keys supported by your version
# Example structure (consult GHSA-c2r5-5j4w-2xff for the authoritative directive):
# ui:
#   snippets:
#     markdown_rendering: disabled