HKUDS OpenHarness CVE-2026-6819

Requires the OpenHarness channel layer to be network-accessible to the attacker (confirmed by CVSS AV:N). User interaction is required (CVSS UI:P), though the exact nature of this interaction is not specified in the description - could involve processing attacker-supplied data through the channel, viewing a malicious message, or triggering automated channel activity. No authentication or special privileges required (PR:N) from the attacker's perspective. Exploitation is limited to deployments where the channel layer endpoints (/plugin install, /plugin enable, /plugin disable, /reload-plugins) are remotely accessible, which is the default configuration in pre-v0.1.7 versions. Systems with network segmentation blocking external access to the channel layer, or those running behind restrictive firewalls, have reduced exposure. The vulnerability affects default configurations without additional hardening.

Real-world risk is high despite requiring user interaction (UI:P in CVSS vector). The vulnerability enables remote unauthenticated attackers to install malicious plugins, which represents a complete system compromise vector with high confidentiality, integrity, and availability impact (VC:H/VI:H/VA:H). While CVSS 8.7 accurately reflects severity, the UI:P requirement suggests the attack requires some form of user action, possibly interaction with a malicious message or request through the channel layer. No EPSS score provided in input data - exploitation probability assessment limited. Not present in CISA KEV, indicating no confirmed active exploitation campaigns. VulnCheck advisory publication and detailed GitHub references (PR, commit, release tag) provide clear remediation path, reducing window of risk for defenders. The privilege escalation tag and plugin installation capability make this highly automatable once initial access is achieved. Primary risk: plugin-based frameworks often grant plugins full application privileges, so arbitrary plugin installation equals arbitrary code execution. Organizations using OpenHarness in production environments with exposed channel layers face immediate risk.

An attacker identifies an internet-exposed OpenHarness instance with accessible channel layer endpoints. They craft a malicious request to /plugin install pointing to an attacker-controlled repository hosting a backdoor plugin. After the user performs some interaction with the channel (trigger varies by deployment - could be processing a message, viewing a dashboard, or automated channel activity), the attacker enables the malicious plugin via /plugin enable and activates it with /reload-plugins. The plugin executes with OpenHarness privileges, establishing persistent access and enabling data exfiltration or lateral movement. GitHub references provide implementation details that would assist exploit development.