The vulnerability identified as CVE-2026-28348 affects the lxml_html_clean project, a Python library used for HTML content cleaning, which is derived from lxml.html.clean. The core of the issue lies in the _has_sneaky_javascript() method, which attempts to detect and filter out dangerous CSS constructs such as @import and expression() to prevent malicious CSS from executing. However, prior to version 0.4.4, this method improperly strips backslashes before performing these checks. This flawed logic allows attackers to use CSS Unicode escape sequences to obfuscate these dangerous keywords, effectively bypassing the filters. As a result, malicious CSS can be injected that loads external stylesheets or executes JavaScript via CSS expressions, leading to cross-site scripting (XSS) vulnerabilities, particularly in older browsers that support these CSS features. The vulnerability impacts the confidentiality and integrity of affected applications by enabling unauthorized script execution and data leakage. The issue does not affect availability and requires user interaction but no privileges or authentication. The vulnerability has been addressed in version 0.4.4 of lxml_html_clean by correcting the filtering logic to properly handle escape sequences. No known exploits have been reported in the wild as of the publication date.

This vulnerability can lead to cross-site scripting (XSS) attacks, allowing attackers to execute arbitrary scripts in the context of affected web applications. This compromises user confidentiality by potentially exposing sensitive data such as cookies, session tokens, or personal information. Integrity is also at risk as attackers could manipulate the content displayed to users or perform actions on their behalf. Although the vulnerability does not impact system availability, the trustworthiness of affected applications is undermined. Organizations using vulnerable versions of lxml_html_clean in their web applications or services that process HTML content are at risk, especially if they serve users with older browsers susceptible to CSS expression exploitation. The lack of authentication requirements and the possibility of remote exploitation increase the threat level. However, the need for user interaction and the medium CVSS score suggest a moderate risk overall. Failure to patch could lead to targeted attacks, data breaches, and reputational damage.

To mitigate this vulnerability, organizations should upgrade all instances of lxml_html_clean to version 0.4.4 or later, where the issue has been fixed. Additionally, developers should review and enhance input validation and sanitization routines to ensure that CSS content is properly escaped and filtered, especially focusing on handling Unicode escape sequences and other obfuscation techniques. Employ Content Security Policy (CSP) headers to restrict the loading of external stylesheets and the execution of inline scripts, reducing the impact of potential XSS attacks. Regularly audit and update third-party libraries to incorporate security patches promptly. For legacy systems that cannot upgrade immediately, consider implementing additional server-side filtering or proxy-based sanitization to detect and block malicious CSS payloads. Educate developers and security teams about the risks of improper encoding and escaping in HTML and CSS processing to prevent similar issues in future development.