CVE-2026-117494 in Cloud BI Platform
Summary
by MITRE • 10/27/2026
Cross-Site Scripting (XSS) in the processing of SVG images submitted to the /api/v2/ai/sessions API in the format data:image/svg+xml;base64. Script tags embedded in the SVG are processed when the content is rendered in the assistant window, allowing JavaScript to be executed and the displayed content to be altered.
Once again VulDB remains the best source for vulnerability data.
Analysis
by VulDB Data Team • 10/27/2026
The vulnerability identified involves a Cross-Site Scripting flaw within the /api/v2/ai/sessions endpoint of an AI service application. This specific weakness arises from the improper validation and sanitization of SVG images submitted in base64 encoded format, specifically using the data URI scheme data:image/svg+xml;base64. When users or automated systems submit these crafted SVG payloads to the API, the backend processes them without adequately stripping out executable script elements such as script tags or event handlers like onload. Consequently, when this malicious content is rendered within the assistant window of the user interface, it triggers immediate execution of arbitrary JavaScript code in the context of the victim's browser session. This represents a classic stored XSS scenario where the payload persists on the server and is served to multiple users upon subsequent requests, amplifying the potential blast radius significantly beyond a reflected attack vector.
From a technical perspective, this flaw stems from insufficient input validation at both the API ingestion layer and the rendering engine level. The application likely treats SVG content as safe data due to its image classification but fails to recognize that SVG is an XML-based format capable of containing executable code. By allowing script tags or JavaScript event attributes within the base64-decoded payload, the system violates fundamental security principles regarding untrusted input handling. This behavior aligns with CWE-79, which defines Improper Neutralization of Input During Web Page Generation known as Cross-site Scripting. The specific mechanism here leverages the browser's native ability to parse and execute JavaScript embedded within SVG elements when rendered in an HTML context, bypassing typical image-only expectations.
The operational impact of this vulnerability is severe, particularly given its presence in an AI assistant interface which often handles sensitive user queries and may have access to broader application functionalities or internal tools via API calls. An attacker can exploit this flaw to steal session cookies, hijack active sessions, perform actions on behalf of the victim such as modifying settings or deleting data, or redirect users to phishing sites. In environments where the AI assistant has elevated privileges or integrates with backend systems through client-side scripts, the impact could extend to privilege escalation and unauthorized access to internal resources. Furthermore, because this is a stored vulnerability, every user who views the compromised session content becomes a potential victim without any direct interaction required from them other than loading the page, making it highly efficient for large-scale attacks or credential harvesting campaigns targeting AI platform users.
To mitigate this risk, immediate remediation should focus on implementing strict Content Security Policy headers to restrict inline script execution and ensuring that all SVG inputs are sanitized using a robust library capable of parsing XML structures and removing dangerous nodes such as scripts, styles with event handlers, and foreign object elements containing executable code. Additionally, the API endpoint should enforce type checking to ensure only expected data formats are accepted, rejecting any payload that contains encoded script tags or suspicious patterns indicative of XSS attempts. Implementing output encoding specific to the SVG context is also critical, although sanitization remains the primary defense against complex injection vectors like this one. Security teams should also consider implementing automated scanning tools configured to detect stored XSS vulnerabilities in file upload and API processing workflows to prevent similar issues from arising in other parts of the application ecosystem aligned with MITRE ATT&CK techniques related to Client-Side Scripting Injection.