A Survey Link Led to the Server: Inside REDCap CVE-2026-90817
REDCap is used to collect research data through public surveys. CVE-2026-90817 turns one of those public entry points into a route to server-side code execution when the deployment is unpatched and an attacker knows a valid survey hash.
The flaw is in survey passthrough routing and import processing. A crafted HTTP request can reach an unintended controller route from a public survey context. During import handling, a crafted file-path or stream parameter can influence where the code reads or executes. The result is unauthenticated remote code execution, scored 9.8 under CVSS 3.1.
Knowing a public survey hash is a real prerequisite, so not every REDCap installation is equally exposed. Public surveys are often intentionally shared, linked from email, or embedded in recruitment material. A survey hash is not a secret in the same way as an administrator password. Teams should not treat the URL as the primary security control.
Scope and fixed versions
The affected range begins with REDCap 13.3.0. Fixes are available in REDCap 16.0.49 LTS, 17.3.10 LTS, and 17.4.4 Standard Release, according to the vulnerability record. CISA added the issue to its Known Exploited Vulnerabilities catalog on November 18, 2026.
The KEV listing changes the priority. An internet-exposed research system may hold sensitive participant data, integrations, file storage, and credentials to downstream services. A web server process that can execute code may also be able to read application configuration, database connections, uploaded files, and scheduled jobs.
What to inspect
Patch first. If exploitation is possible or the host shows suspicious activity, preserve evidence before rebuilding. Review:
- Web access logs for requests that combine survey routes with import parameters, path traversal, or unexpected file streams.
- PHP-FPM, web server, and application logs around the first suspicious request.
- Upload directories and the application tree for files created or modified outside a deployment window.
- Child processes launched by the web server, especially shells, interpreters, download tools, and package managers.
- Outbound connections from the REDCap server during the suspected window.
- Application credentials, database users, mail settings, and integration secrets available to the compromised process.
The exact payload and post-exploitation chain can vary. The safe assumption after confirmed code execution is that every secret readable by the web application is exposed. Rotate credentials from a separate trusted host and invalidate sessions where the application supports it.
Reduce the next exposure
Set a release process that treats REDCap as an internet service with a research-data role. Track public surveys and their owners, remove surveys that are no longer needed, and review the routes that public survey traffic is allowed to reach. Put administrative and import functionality behind network or identity controls that do not depend on a survey URL.
Application allowlisting, egress filtering, and write restrictions on the web root raise the cost of exploitation. They do not replace the patch. They also provide evidence that helps distinguish a blocked attempt from a successful one.
Top comments (0)