MAL-2026-17230
Vulnerability from ossf_malicious_packages
Published
2026-09-28 19:36
Modified
2026-09-28 19:36
Summary
Malicious code in llm-nebula (npm)
Details

-= Per source details. Do not edit below this line.=-

Source: amazon-inspector (0eee293a9973027154eea71635c14e8a4cb2ad8a1e4f80a65399ad874afcb669)

Package presents itself as an LLM SDK (nebula.js exposes a small client stub) but declares preinstall: node preinstall.cjs in package.json, and preinstall.cjs is a ~232KB single-line obfuscated blob that runs automatically on npm install. The script wraps its body in a Function(...) constructor and uses a custom PRNG-based string decoder (TEA/xorshift-style constants 0x9e3779b9 / 0x243f6a88 / 0x6a09e667) over a packed printable-ASCII string plus a hex-int array to reconstruct characters via String.fromCharCode, driving a control-flow-flattened switch dispatcher. Node builtins and method names are resolved dynamically at runtime (e.g. Ooa8zU["Bd5Wj1"]("fs")) so module ids, filesystem paths, network destinations, and command strings are not visible without executing the decoder. This obfuscation-plus-lifecycle-hook composition is the canonical install-time dropper / RCE shape: the benign-looking library entry serves as a cover story while the hidden preinstall payload runs on any consumer's machine at install time with the user's privileges, capable of arbitrary filesystem, process, and network operations.

CWE
  • CWE-506 - The product contains code that appears to be malicious in nature.

{
  "affected": [
    {
      "database_specific": {
        "cwes": [
          {
            "cweId": "CWE-506",
            "description": "The product contains code that appears to be malicious in nature.",
            "name": "Embedded Malicious Code"
          }
        ],
        "indicators": {
          "evidence_files": [
            {
              "path": "preinstall.cjs",
              "sha256": "5c8e0e83983ba18c900dc18b67c6e6aa1eff939c0a107f2cbc4d4967f392ebaa",
              "tlsh": "07342ac8920e2e45e6228f1b40ddfc8efd20545345ba4efafd66520ad9670e293fbd11"
            }
          ],
          "package_integrity": [
            {
              "filename": "llm-nebula-1.0.0.tgz",
              "hashes": {
                "sha1": "b88f909b3890d518bd049b975757380ae59242f7",
                "sha512_sri": "sha512-/937+hoW4m97kLwTsLfJDFKo/RNcsNFyFh2LnnId7vAG5mQEgC7fAjO2eUc41WmLMN16HYnO8MfwCMLXkotNdg=="
              }
            }
          ]
        }
      },
      "package": {
        "ecosystem": "npm",
        "name": "llm-nebula"
      },
      "versions": [
        "1.0.0"
      ]
    }
  ],
  "credits": [
    {
      "contact": [
        "inspector-research@amazon.com"
      ],
      "name": "Amazon Inspector",
      "type": "FINDER"
    }
  ],
  "database_specific": {
    "malicious-packages-origins": [
      {
        "id": "IN-MAL-2026-020610",
        "import_time": "2026-09-28T19:40:23.367064504Z",
        "modified_time": "2026-09-28T19:36:11Z",
        "sha256": "0eee293a9973027154eea71635c14e8a4cb2ad8a1e4f80a65399ad874afcb669",
        "source": "amazon-inspector",
        "versions": [
          "1.0.0"
        ]
      }
    ]
  },
  "details": "\n---\n_-= Per source details. Do not edit below this line.=-_\n\n## Source: amazon-inspector (0eee293a9973027154eea71635c14e8a4cb2ad8a1e4f80a65399ad874afcb669)\nPackage presents itself as an LLM SDK (nebula.js exposes a small client stub) but declares `preinstall: node preinstall.cjs` in package.json, and preinstall.cjs is a ~232KB single-line obfuscated blob that runs automatically on `npm install`. The script wraps its body in a `Function(...)` constructor and uses a custom PRNG-based string decoder (TEA/xorshift-style constants 0x9e3779b9 / 0x243f6a88 / 0x6a09e667) over a packed printable-ASCII string plus a hex-int array to reconstruct characters via `String.fromCharCode`, driving a control-flow-flattened switch dispatcher. Node builtins and method names are resolved dynamically at runtime (e.g. `Ooa8zU[\"Bd5Wj1\"](\"fs\")`) so module ids, filesystem paths, network destinations, and command strings are not visible without executing the decoder. This obfuscation-plus-lifecycle-hook composition is the canonical install-time dropper / RCE shape: the benign-looking library entry serves as a cover story while the hidden preinstall payload runs on any consumer\u0027s machine at install time with the user\u0027s privileges, capable of arbitrary filesystem, process, and network operations.\n",
  "id": "MAL-2026-17230",
  "modified": "2026-09-28T19:36:11Z",
  "published": "2026-09-28T19:36:11Z",
  "references": [
    {
      "type": "PACKAGE",
      "url": "https://www.npmjs.com/package/llm-nebula/v/1.0.0"
    }
  ],
  "schema_version": "1.7.4",
  "summary": "Malicious code in llm-nebula (npm)"
}



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…