CVE-2025-36321: IBM watsonx.data Intelligence HTML Injection Vulnerability

Technical summary

This vulnerability is classified as HTML injection (CWE-80), affecting IBM watsonx.data intelligence versions 5.2.0, 5.2.1, 5.2.2, and 5.3.0. The flaw stems from insufficient input sanitization or output encoding in the web application layer. An authenticated attacker can inject arbitrary HTML—including JavaScript—into the application, which is then rendered unsafely to other users. The CVSS 3.1 score of 5.7 (Medium severity) reflects the requirement for prior authentication and user interaction, but acknowledges the high confidentiality impact if successful. The attack does not require elevated privileges or network complexity beyond what is typical for web applications.

Business impact

Organizations running affected IBM watsonx.data intelligence versions face insider or lateral threat scenarios where an authenticated user can compromise other users' sessions or data. In enterprise analytics and AI platforms like watsonx.data, this could lead to unauthorized access to sensitive datasets, analytics queries, or model configurations. The impact is bounded by the fact that exploitation requires authentication and user action, but in busy collaborative environments where users frequently access shared dashboards or reports, the attack surface is substantial. Reputational and compliance risks emerge if user data is exfiltrated through session hijacking.

Affected systems

IBM watsonx.data intelligence is affected in versions 5.2.0, 5.2.1, 5.2.2, and 5.3.0. IBM Software Hub (a related product distribution platform) may also be affected. Organizations should verify their installed versions and check IBM's patch advisories for the full scope of impacted deployments. Versions prior to 5.2.0 and any patches released after 5.3.0 should be assessed separately against vendor guidance.

Exploitability

Exploitation is moderately feasible but requires authenticated access to the application. The attacker must first gain valid credentials—through phishing, credential stuffing, insider access, or compromise of another account—then craft and submit malicious HTML through an input field or parameter. The victim must then view the injected content for the payload to execute. This is not a zero-click or unauthenticated attack, which limits widespread automated exploitation. However, in environments with weak password policies or high user counts, the barrier to obtaining credentials is lower. The attack is deterministic once a victim is directed to the payload.

Remediation

IBM has released patched versions addressing this vulnerability. Organizations must identify all instances of IBM watsonx.data intelligence running versions 5.2.0–5.3.0 and upgrade to the patched release. Additionally, implement input validation and output encoding best practices: sanitize all user-supplied input before storage, apply context-appropriate encoding (HTML entities for HTML context, JavaScript escaping for script context), and consider implementing a Content Security Policy (CSP) to restrict inline script execution. Conduct user awareness training on reporting suspicious injected content.

Patch guidance

Apply the latest patched version of IBM watsonx.data intelligence released by IBM to address CVE-2025-36321. Consult IBM's official security advisory for specific version numbers and download links. Deploy patches in a staging environment first to validate compatibility with your analytics workflows and integrations. Plan patching during a controlled maintenance window to minimize disruption to users relying on watsonx.data for business intelligence. Monitor IBM's Security Advisory pages and your vendor support portal for any interim guidance or additional affected versions.

Detection guidance

Monitor application logs and web access logs for unusual HTML or JavaScript patterns in user-supplied input fields, particularly those that feed into reports, dashboards, or shared content areas. Deploy Web Application Firewall (WAF) rules to detect and block common HTML/JavaScript injection patterns. Implement user and entity behavior analytics (UEBA) to flag authenticated users submitting suspicious payloads. Within IBM watsonx.data, enable audit logging if available to track content creation and modifications. Periodically review stored user-generated content for signs of injected HTML or script tags.

Why prioritize this

Although this vulnerability carries a Medium CVSS score and requires authentication, it should be prioritized based on your user population size and data sensitivity. Environments with many concurrent users, sensitive analytics, or compliance requirements (healthcare, finance) face higher exploitation likelihood. The vulnerability is not yet tracked in CISA's KEV catalog, but this does not diminish urgency—early patching prevents weaponization. Prioritize systems hosting sensitive datasets or accessible to external contractors or partner organizations.

Risk score, explained

The CVSS 3.1 score of 5.7 reflects a Medium severity rating anchored by the requirement for authenticated access (PR:L) and user interaction (UI:R). The high confidentiality impact (C:H) acknowledges that a successful injection can lead to session theft or sensitive data disclosure, offsetting the lower availability and integrity scores. The network-accessible attack vector (AV:N) and low attack complexity (AC:L) indicate that once an attacker has credentials, exploitation is straightforward. The score appropriately balances the insider-threat nature of the attack against the real harm it can inflict in collaborative environments.

Frequently asked questions