The vulnerability in Jupyter nbconvert allows an attacker to read any file on the host system when embedding images as base64 data URIs within a notebook. By setting HTMLExporter.embed_images=True, the markdown renderer processes image references that contain path traversal strings, causing the conversion tool to read the referenced file and encode it for inclusion in the output HTML. This results in exfiltration of sensitive data, an instance of CWE‑22 and CWE‑73.

Pages that use the nbconvert tool from the Jupyter project, specifically versions 6.5 through 7.17.0. The flaw is tied to the configuration option HTMLExporter.embed_images, which is active only when explicitly enabled.

The vulnerability carries a CVSS score of 6.5, indicating moderate severity, and is not flagged in CISA's Known Exploited Vulnerabilities catalog. The EPSS score is unavailable, so the current likelihood of exploitation is unknown, but the flaw requires a crafted notebook that the attacker can have converted on a host with access to the notebook. Because embed_images is not enabled by default, the attack vector relies on the attacker having control of the notebook conversion process.