IBM watsonx.data Intelligence EUVD-2025-210381

Exploitation requires the attacker to hold a valid low-privilege authenticated session on the IBM watsonx.data Intelligence web interface (confirmed by CVSS PR:L). A second precondition is that a victim user must actively navigate to and render the page containing the injected HTML content (CVSS UI:R). Affected versions are explicitly bounded to 5.2.0, 5.2.1, 5.2.2, and 5.3.0 - instances running versions outside this range or running the patched build are not affected. The unchanged scope (CVSS S:U) limits impact to the watsonx.data web origin and prevents the injected code from escaping the browser sandbox to affect the underlying host. No special non-default configuration is required for the vulnerability to be present beyond deploying an affected version with the web interface accessible.

The CVSS 5.7 Medium score is shaped by the PR:L and UI:R constraints, which materially reduce exploitability despite the AV:N/AC:L base. A real-world attacker needs two preconditions: a valid authenticated session and a second user who browses to the injected content. The C:H rating is realistic - stored HTML injection can exfiltrate session cookies or redirect victims to credential-harvesting pages. No EPSS data was provided in the input, so probabilistic exploitation likelihood cannot be quantified. The absence of a CISA KEV listing and no identified public POC code both lower operational urgency compared to the raw CVSS alone. Risk is real but primarily concerns insider threats or low-privilege users who can inject content into shared views seen by higher-privileged users such as administrators. CVSS vector and description are consistent - no conflicts detected.

An attacker with a low-privilege authenticated account submits a crafted payload containing malicious HTML or script tags into a watsonx.data Intelligence input field - such as a shared dashboard name, comment, or label - that is rendered in other users' browsers. When an administrator or higher-privileged user later views the affected page, the injected markup executes in their browser session, potentially exfiltrating their session cookie to an attacker-controlled endpoint. No public proof-of-concept code has been identified at time of analysis.