Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB
CVE-2026-94220 Overview
CVE-2026-94220 is a Cross-Site Request Forgery (CSRF) vulnerability [CWE-352] in the feishu-auth and dingtalk-auth plugins of Apache APISIX. The flaw affects Apache APISIX versions 3.17.0 through 3.18.0. An attacker who tricks a user into clicking a crafted link can bind that user's authenticated browser session to the attacker's identity on a protected route. Any subsequent user activity on that route, including uploads, form submissions, and account bindings, is attributed to the attacker's account. The Apache APISIX project has released version 3.19.0 to address the issue.
Critical Impact
Successful exploitation allows an attacker to hijack the identity binding of a victim's authenticated session, redirecting the victim's uploads, submissions, and account bindings to the attacker-controlled account.
Affected Products
- Apache APISIX 3.17.0
- Apache APISIX versions between 3.17.0 and 3.18.0
- Apache APISIX 3.18.0
Discovery Timeline
- 2026-10-01 - CVE CVE-2026-94220 published to NVD
- 2026-10-01 - Last updated in NVD database
Technical Details for CVE-2026-94220
Vulnerability Analysis
Apache APISIX is an open-source API gateway that routes and secures traffic to backend services. The feishu-auth and dingtalk-auth plugins integrate the gateway with Feishu (Lark) and DingTalk OAuth-style identity providers. The vulnerability resides in the authentication callback flow used by these two plugins.
The plugins fail to validate that an authentication response initiated in the user's browser was also started by that same browser. As a result, a crafted link can cause the victim's browser to complete an authentication exchange that the attacker started. The protected route then establishes the session under the attacker's identity rather than the victim's. Work performed in that session, including file uploads, form submissions, and third-party account bindings, is applied to the attacker's account.
Root Cause
The root cause is the absence of anti-CSRF protection on the authentication callback endpoints exposed by the feishu-auth and dingtalk-auth plugins. Specifically, the plugins do not bind the authentication state to the originating browser through a cryptographic state parameter or an equivalent per-session token. Without this binding, the gateway cannot distinguish a user-initiated login from a session-fixation attempt driven by an attacker-crafted URL.
Attack Vector
Exploitation requires user interaction over the network. An attacker generates an authentication request against the vulnerable APISIX deployment using the attacker's own identity provider credentials. The attacker then delivers the resulting callback URL to a victim through phishing, a malicious page, or any clickable channel. When the victim clicks the link while browsing the protected site, their browser completes the attacker-initiated authentication flow. The gateway binds the victim's session to the attacker's identity, and the victim unknowingly operates inside the attacker's account for the remainder of that session.
No exploit code is required beyond the crafted link. See the Apache Security Mailing List Post and the OpenWall OSS Security Announcement for the project's technical description.
Detection Methods for CVE-2026-94220
Indicators of Compromise
- Authentication callbacks to feishu-auth or dingtalk-auth plugin routes with a Referer header pointing to an external, untrusted domain.
- Session identities on protected routes that do not match the browser's previously established identity cookies or local storage markers.
- Account-binding events, uploads, or form submissions associated with user accounts that no legitimate operator recognizes.
Detection Strategies
- Inspect APISIX access logs for authentication callback requests to the feishu-auth and dingtalk-auth endpoints that lack a prior login-initiation request from the same client IP and user agent.
- Correlate identity-provider login events with APISIX session establishment events to flag mismatches where the authenticated subject changes mid-session.
- Review downstream application audit logs for account-binding or profile-modification actions performed immediately after an authentication callback.
Monitoring Recommendations
- Forward APISIX gateway logs and identity-provider logs to a centralized analytics platform to enable cross-source correlation.
- Alert on repeated authentication callbacks delivered from identical URLs to multiple distinct user sessions, which indicates a shared attacker-crafted link.
- Track the rate of new third-party account bindings per user and alert on sudden spikes.
How to Mitigate CVE-2026-94220
Immediate Actions Required
- Upgrade Apache APISIX to version 3.19.0, which contains the fix for CVE-2026-94220.
- Inventory all APISIX deployments and confirm which gateways have the feishu-auth or dingtalk-auth plugins enabled.
- Audit recent authentication callback traffic and account-binding activity for signs of exploitation.
Patch Information
The Apache APISIX project fixed this vulnerability in version 3.19.0. Operators running any release from 3.17.0 through 3.18.0 should upgrade. Refer to the Apache Security Mailing List Post for the vendor advisory.
Workarounds
- Disable the feishu-auth and dingtalk-auth plugins until the gateway is upgraded to 3.19.0.
- Restrict access to the authentication callback routes with network ACLs or an upstream Web Application Firewall rule that enforces a trusted Referer or Origin header.
- Require users to re-authenticate before performing sensitive actions such as account bindings or uploads.
# Example: disable the affected plugins in the APISIX configuration
# /usr/local/apisix/conf/config.yaml
plugins:
# - feishu-auth # disabled until upgrade to 3.19.0
# - dingtalk-auth # disabled until upgrade to 3.19.0
Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.
Sources: National Vulnerability Database, CISA KEV, Exploit-DB

