CVE-2026-62663

{"id": "CVE-2026-62663", "cveTags": [], "metrics": {"ssvcV203": [{"source": "134c704f-9b21-4f2e-91b3-4a467353bcc0", "ssvcData": {"id": "CVE-2026-62663", "role": "CISA Coordinator", "options": [{"exploitation": "poc"}, {"automatable": "yes"}, {"technicalImpact": "partial"}], "version": "2.0.3", "timestamp": "2026-07-30T17:38:03.008311Z"}}], "cvssMetricV31": [{"type": "Secondary", "source": "security-advisories@github.com", "cvssData": {"scope": "UNCHANGED", "version": "3.1", "baseScore": 7.5, "attackVector": "NETWORK", "baseSeverity": "HIGH", "vectorString": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:N/A:N", "integrityImpact": "NONE", "userInteraction": "NONE", "attackComplexity": "LOW", "availabilityImpact": "NONE", "privilegesRequired": "NONE", "confidentialityImpact": "HIGH"}, "impactScore": 3.6, "exploitabilityScore": 3.9}]}, "affected": [{"source": "security-advisories@github.com", "affectedData": [{"vendor": "masci", "product": "banks", "versions": [{"status": "affected", "version": "< 2.4.4"}]}]}], "published": "2026-07-30T17:16:33.900", "references": [{"url": "https://github.com/masci/banks/security/advisories/GHSA-98rr-gvc9-3cjh", "source": "security-advisories@github.com"}, {"url": "https://github.com/masci/banks/security/advisories/GHSA-98rr-gvc9-3cjh", "source": "134c704f-9b21-4f2e-91b3-4a467353bcc0"}], "vulnStatus": "Deferred", "weaknesses": [{"type": "Secondary", "source": "security-advisories@github.com", "description": [{"lang": "en", "value": "CWE-22"}]}], "descriptions": [{"lang": "en", "value": "Banks generates meaningful LLM prompts using a simple template language. In versions prior to 2.4.4, all four media filters (image, audio, video, document) in banks accept untrusted user input as file paths via Path(value) and pass them directly to open(file_path, \"rb\") without any path sanitization, canonicalization, or directory restriction. An attacker who controls template variables passed to a banks Prompt can use path traversal (../) to read arbitrary files accessible to the Python process\u2014including .env files, SSH keys, cloud credentials, source code, /etc/passwd, and /etc/shadow\u2014with the content returned base64-encoded in the rendered prompt output, making exfiltration trivial. This is particularly dangerous for applications that use banks to process user-provided template variables before sending prompts to an LLM. This issue has been fixed in version 2.4.4."}, {"lang": "es", "value": "Banks genera prompts significativos para LLM usando un lenguaje de plantilla simple. En versiones anteriores a la 2.4.4, los cuatro filtros de medios (imagen, audio, video, documento) en banks aceptan entrada de usuario no confiable como rutas de archivo a trav\u00e9s de Path(value) y las pasan directamente a open(file_path, \u201crb\u201d) sin ninguna sanitizaci\u00f3n de ruta, canonicalizaci\u00f3n o restricci\u00f3n de directorio. Un atacante que controla las variables de plantilla pasadas a un Prompt de banks puede usar salto de ruta (.. / ) para leer archivos arbitrarios accesibles al proceso de Python -incluyendo archivos .env, claves SSH, credenciales de la nube, c\u00f3digo fuente, /etc/p4ssd y /etc/sha4dow- con el contenido devuelto codificado en base64 en la salida del prompt renderizado, haciendo que la exfiltraci\u00f3n sea trivial. Esto es particularmente peligroso para aplicaciones que usan banks para procesar variables de plantilla proporcionadas por el usuario antes de enviar prompts a un LLM. Este problema ha sido solucionado en la versi\u00f3n 2.4.4."}], "lastModified": "2026-10-23T22:10:00.160", "sourceIdentifier": "security-advisories@github.com"}