What is REDCap?
REDCap (Research Electronic Data Capture) is a PHP web application developed at Vanderbilt University for building and running online surveys and research databases. It is the default data collection platform for clinical and academic research: investigators use it to design case report forms, run longitudinal studies, collect participant-submitted survey responses, and store the resulting data. Vanderbilt licenses it at no cost to non-profit institutions through the REDCap Consortium, whose thousands of member institutions across more than 150 countries each run their own self-hosted instance.
That makes REDCap an unusually attractive target. Instances hold identifiable health information and research data that falls under HIPAA, IRB protocols, and institutional review obligations. They are run by individual universities, hospitals, and research institutes rather than by a central vendor, so patch cadence varies widely. And by design a REDCap server publishes public survey links that anyone on the internet can open without logging in, which guarantees an unauthenticated attack surface.
Overview
CVE-2026-90817 is an unauthenticated remote code execution vulnerability in REDCap. A crafted request against a public survey endpoint escapes the survey routing context and reaches an internal controller that should only be reachable by an authenticated user. From there the attacker supplies a crafted file-path or stream parameter to the data import handler, poisons a cached template file with PHP of their choosing, and then triggers a request that causes the server to include and execute it.
The only precondition is a valid public survey hash, which is not a secret: survey links are handed out to participants, posted on study recruitment pages, and embedded in public HTML. No account, no credentials, and no user interaction are needed. Securifera, acting as CNA, scored it 9.8 Critical under CVSS v3.1 (AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H); NVD carries the same score. VulDB's independent rating of 7.3 is an outlier.
Exploitation in the wild began within weeks of disclosure. Palo Alto Networks Unit 42 documented an XMRig cryptomining campaign using the flaw on 2026-10-07, and university operators observed further attacks through mid-October. CISA added the CVE to the KEV catalog on 2026-11-18 with a BOD 26-04 remediation deadline of 2026-11-21.
Affected Versions
| Product | Vulnerable | Fixed |
|---|---|---|
| REDCap LTS 16.x | 13.3.0 through 16.0.48 | 16.0.49 |
| REDCap LTS 17.3.x | Through 17.3.9 | 17.3.10 |
| REDCap Standard 17.4.x | Through 17.4.3 | 17.4.4 |
The vulnerable code was introduced in REDCap 13.3.0, so every release from that version until the fixes is affected regardless of branch. REDCap is distributed through the consortium's community site rather than a public download, and each institution applies upgrades itself.
Technical Details
The CVE record assigns CWE-94 (code injection) with CWE-73 (external control of file name or path) as the enabling weakness. The published chain has three stages.
First, a routing bypass. A double-encoded __passthru parameter on the public survey endpoint (/surveys/?s=<hash>) causes the request to be dispatched to a controller intended for authenticated sessions, skipping the authentication and CSRF checks that would normally guard it. Second, a file write. The data import controller's index action accepts a fname parameter that is passed to a PHP stream function without restricting the wrapper, so a php://filter chain can be used to turn attacker-controlled input into attacker-chosen bytes written into a cached template file under temp/cache/. Third, execution. A follow-up request to a controller route that renders from the template cache causes the poisoned file to be included and its PHP executed.
The filter-chain technique is what makes this practical: php://filter lets an attacker stack encoding conversions so that an innocuous-looking input string is transformed into arbitrary bytes on disk, which is how code gets into a file the application will later run. Reported exploit code also restores the original cache file after execution, leaving little on-disk evidence beyond the web server logs.
Attack complexity is low. The whole chain is a short sequence of unauthenticated HTTP requests to a single host, with no second vulnerability required and no race condition. Code runs as the web server user, with access to the REDCap database credentials in the application configuration and therefore to the study data itself.
Discovery
The vulnerability was found and reported by Securifera, which also acted as the CNA for the CVE and assigned its severity. Securifera published the CVE record on 2026-09-20 and followed with a technical write-up, "Survey Says.... Remote Code Execution: CVE-2026-90817," on 2026-09-27. Vanderbilt shipped the fixed builds (16.0.49, 17.3.10, and 17.4.4) to consortium members, who then have to apply them locally.
Independent tooling followed quickly. The CTS-IT group at the University of Florida published a passive scanner that checks a REDCap host's version, looks for the campaign's host-level artifacts, and greps web server logs for exploit attempts, and the ProjectDiscovery community added a Nuclei template reproducing the detection.
Exploitation Context
Unit 42's report (tracked as "2026-10-07-XMRig-campaign-through-CVE-2026-90817") describes attackers using the flaw to install XMRig, a Monero cryptocurrency miner, on Linux REDCap hosts. The documented post-exploitation behavior includes a dropper staged under temp/cache/, a miner binary masquerading as a syslog-ng process, a Monero wallet address visible on the command line, cron entries for persistence, an added SSH authorized_keys backdoor, and shell history wiping. Staging artifacts have been seen in /dev/shm/duet and /var/tmp/.bin, and the miner exposes its local API on 127.0.0.1:50001.
Public indicators include the attacker address 202.61.130.221 from the Unit 42 report, plus additional addresses first observed between 2026-10-14 and 2026-10-19 against a university REDCap instance. In web server logs the signature to hunt for is a request to the public survey endpoint carrying a double-encoded __passthru parameter, requests naming DataImportController, and any php://filter string in a query parameter.
Multiple public proof-of-concept scripts and a Nuclei template exist, so the barrier to opportunistic scanning is effectively zero. CISA has not named a state or criminal actor beyond the cryptomining activity, and the KEV entry records no known ransomware use. Note that cryptomining is the campaign that has been documented, not the ceiling: the same access yields the research database and any credentials stored with it.
Remediation
- Upgrade REDCap to 16.0.49 LTS, 17.3.10 LTS, or 17.4.4 Standard, or any later release on your branch. Downloads and release notes are on the REDCap Community site.
- Assume compromise if the instance was internet-reachable and unpatched between late September and the upgrade date. Patching does not evict an attacker who already has cron persistence and an SSH key.
- Run a compromise assessment on the host. Check
temp/cache/for unexpected files, look for an XMRig process or asyslog-ng-named binary that is not the real syslog-ng, inspect cron entries for every account, review~/.ssh/authorized_keysfor keys you did not add, and check for listeners on 127.0.0.1:50001. The CTS-IT scanner automates these checks read-only. - Search web server logs back to 2026-09-20 for requests to
/surveys/with a__passthruparameter, references toDataImportController, andphp://filterin any parameter. A log hit is evidence of an attempt; the host-state checks above confirm compromise. - Rotate credentials if there is any sign of execution: the REDCap database account, any API tokens issued to projects, SMTP credentials, and any LDAP or SSO service account in the configuration.
- Restrict network exposure where the research workflow allows it. Put the instance behind a VPN or an allowlist if participants do not need public survey access, and place a WAF rule in front of the survey endpoint otherwise. There is no vendor workaround short of upgrading.
- Involve your institution's privacy office. A confirmed execution on a REDCap host is potential unauthorized access to identifiable research and health data, which carries its own notification obligations independent of the technical cleanup.
- Federal civilian agencies must remediate by 2026-11-21 under BOD 26-04 and follow the accompanying Forensics Triage Requirements, or discontinue use of the product.
Key Details
| Property | Value |
|---|---|
| CVE ID | CVE-2026-90817 |
| Vendor / Product | Vanderbilt University - Research Electronic Data Capture (REDCap) |
| NVD Published | 2026-09-20 |
| NVD Last Modified | 2026-11-18 |
| CVSS 3.1 Score | 9.8 |
| CVSS 3.1 Vector | CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H |
| Severity | CRITICAL |
| CWE | CWE-73 find similar ↗ |
| CISA KEV Added | 2026-11-18 |
| CISA KEV Deadline | 2026-11-21 |
| Known Ransomware Use | No |
CVSS 3.1 Breakdown
Required Action
Timeline
| Date | Event |
|---|---|
| 2026-09-20 | CVE-2026-90817 published by Securifera as CNA |
| 2026-09-27 | Securifera publishes its technical write-up |
| 2026-10-07 | Palo Alto Unit 42 documents an XMRig cryptomining campaign exploiting the flaw |
| 2026-10-19 | Additional attacker IP activity observed against a university REDCap host |
| 2026-11-18 | Added to CISA Known Exploited Vulnerabilities catalog |
| 2026-11-21 | CISA BOD 26-04 remediation deadline |
References
| Resource | Type |
|---|---|
| NVD - CVE-2026-90817 | Vulnerability Database |
| CISA KEV Catalog Entry | US Government |
| Securifera - Survey Says.... Remote Code Execution: CVE-2026-90817 | Security Research |
| REDCap Community - release notes and upgrade downloads | Vendor Advisory |
| CTS-IT - passive REDCap exposure and compromise scanner for CVE-2026-90817 | Security Research |
| SecurityOnline - CISA adds four exploited flaws to KEV | News |
| CISA Alert - CISA Adds Four Known Exploited Vulnerabilities to Catalog | US Government |