GHSA-QPX9-HPMF-5GMW

Vulnerability from github – Published: 2026-03-03 17:46 – Updated: 2026-05-05 22:01
VLAI
Summary
Underscore has unlimited recursion in _.flatten and _.isEqual, potential for DoS attack
Details

Impact

In simple words, some programs that use _.flatten or _.isEqual could be made to crash. Someone who wants to do harm may be able to do this on purpose. This can only be done if the program has special properties. It only works in Underscore versions up to 1.13.7. A more detailed explanation follows.

In affected versions of Underscore, the _.flatten and _.isEqual functions use recursion without a depth limit. Under very specific conditions, detailed below, an attacker could exploit this in a Denial of Service (DoS) attack by triggering a stack overflow.

A proof of concept (PoC) for this type of attack with _.isEqual:

const _ = require('underscore');

// build JSON string for nested object ~4500 levels deep
// (for this to be an attack, the JSON would have to come from
// a request or other untrusted input)
let json = '';
for (let i = 0; i < 4500; i++) json += '{"n":';
json += '"x"';
for (let i = 0; i < 4500; i++) json += '}';

// construct two distinct objects with equal shape from the above JSON
const a = JSON.parse(json);
const b = JSON.parse(json);

_.isEqual(a, b); // RangeError: Maximum call stack size exceeded

A proof of concept (PoC) for this type of attack with _.flatten:

const _ = require('underscore');

// build nested array ~4500 levels deep
// (like with _.isEqual, this nested array would have to be sourced
// from an untrusted external source for it to be an attack)
let nested = [];
for (let i = 0; i < 4500; i++) nested = [nested];

_.flatten(nested); // RangeError: Maximum call stack size exceeded

An application that crashes because of this can be restarted, so the bug is most relevant to applications for which continued operation is important, such as server applications. Furthermore, an application is only vulnerable to this type of attack if ALL of the following conditions are met:

  • Untrusted input must be used to create a recursive datastructure, for example using JSON.parse, with no enforced depth limit.
  • The datastructure thus created must be passed to _.flatten or _.isEqual.
  • In the case of _.flatten, the vulnerability can only be exploited if it is possible for a remote client to prepare a datastructure that consists of arrays at all levels AND if no finite depth limit is passed as the second argument to _.flatten.
  • In the case of _.isEqual, the vulnerability can only be exploited if there exists a code path in which two distinct datastructures that were submitted by the same remote client are compared using _.isEqual. For example, if a client submits data that are stored in a database, and the same client can later submit another datastructure that is then compared to the data that were saved in the database previously, OR if a client submits a single request, but its data are parsed twice, creating two non-identical but equivalent datastructures that are then compared.
  • Exceptions originating from the call to _.flatten or _.isEqual, as a result of a stack overflow, are not being caught.

All versions of Underscore up to and including 1.13.7 are affected by this weakness.

Patches

The problem has been patched in version 1.13.8. Upgrading to 1.13.8 or later completely prevents exploitation.

Note: historically, there have been breaking changes in minor releases of Underscore, especially between versions 1.6 and 1.9. However, upgrading from version 1.9 or later to any later 1.x version should be feasible with little or no effort for all users.

Workarounds

A workaround that works for both functions is to enforce a depth limit on the datastructure that is created from untrusted input. A limit of 1000 levels should prevent attacks from being successful on most systems. In systems with highly constrained hardware, we recommend lower limits, for example 100 levels.

Another possible workaround that only works for _.flatten, is to pass a second argument that limits the flattening depth to 1000 or less.

References

  • https://github.com/jashkenas/underscore/issues/3011
  • https://underscorejs.org/#1.13.8
  • https://underscorejs.org/#flatten
  • https://underscorejs.org/#isEqual
Show details on source website

{
  "affected": [
    {
      "database_specific": {
        "last_known_affected_version_range": "\u003c= 1.13.7"
      },
      "package": {
        "ecosystem": "npm",
        "name": "underscore"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "1.13.8"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-27601"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-674",
      "CWE-770"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-03-03T17:46:06Z",
    "nvd_published_at": "2026-03-03T23:15:55Z",
    "severity": "HIGH"
  },
  "details": "### Impact\n\nIn simple words, some programs that use `_.flatten` or `_.isEqual` could be made to crash. Someone who wants to do harm may be able to do this on purpose. This can only be done if the program has special properties. It only works in Underscore versions up to 1.13.7. A more detailed explanation follows.\n\nIn affected versions of Underscore, the `_.flatten` and `_.isEqual` functions use recursion without a depth limit. Under very specific conditions, detailed below, an attacker could exploit this in a Denial of Service (DoS) attack by triggering a stack overflow.\n\nA proof of concept (PoC) for this type of attack with `_.isEqual`:\n\n```js\nconst _ = require(\u0027underscore\u0027);\n\n// build JSON string for nested object ~4500 levels deep\n// (for this to be an attack, the JSON would have to come from\n// a request or other untrusted input)\nlet json = \u0027\u0027;\nfor (let i = 0; i \u003c 4500; i++) json += \u0027{\"n\":\u0027;\njson += \u0027\"x\"\u0027;\nfor (let i = 0; i \u003c 4500; i++) json += \u0027}\u0027;\n\n// construct two distinct objects with equal shape from the above JSON\nconst a = JSON.parse(json);\nconst b = JSON.parse(json);\n\n_.isEqual(a, b); // RangeError: Maximum call stack size exceeded\n```\n\nA proof of concept (PoC) for this type of attack with `_.flatten`:\n\n```js\nconst _ = require(\u0027underscore\u0027);\n\n// build nested array ~4500 levels deep\n// (like with _.isEqual, this nested array would have to be sourced\n// from an untrusted external source for it to be an attack)\nlet nested = [];\nfor (let i = 0; i \u003c 4500; i++) nested = [nested];\n\n_.flatten(nested); // RangeError: Maximum call stack size exceeded\n```\n\nAn application that crashes because of this can be restarted, so the bug is most relevant to applications for which continued operation is important, such as server applications. Furthermore, an application is only vulnerable to this type of attack if ALL of the following conditions are met:\n\n- Untrusted input must be used to create a recursive datastructure, for example using `JSON.parse`, with no enforced depth limit.\n- The datastructure thus created must be passed to `_.flatten` or `_.isEqual`.\n- In the case of `_.flatten`, the vulnerability can only be exploited if it is possible for a remote client to prepare a datastructure that consists of arrays at all levels AND if no finite depth limit is passed as the second argument to `_.flatten`.\n- In the case of `_.isEqual`, the vulnerability can only be exploited if there exists a code path in which two distinct datastructures that were submitted by the same remote client are compared using `_.isEqual`. For example, if a client submits data that are stored in a database, and the same client can later submit another datastructure that is then compared to the data that were saved in the database previously, OR if a client submits a single request, but its data are parsed twice, creating two non-identical but equivalent datastructures that are then compared.\n- Exceptions originating from the call to `_.flatten` or `_.isEqual`, as a result of a stack overflow, are not being caught.\n\nAll versions of Underscore up to and including 1.13.7 are affected by this weakness.\n\n### Patches\n\nThe problem has been patched in version 1.13.8. Upgrading to 1.13.8 or later completely prevents exploitation.\n\n**Note:** historically, there have been breaking changes in minor releases of Underscore, especially between versions 1.6 and 1.9. However, upgrading from version 1.9 or later to any later 1.x version should be feasible with little or no effort for all users.\n\n### Workarounds\n\nA workaround that works for both functions is to enforce a depth limit on the datastructure that is created from untrusted input. A limit of 1000 levels should prevent attacks from being successful on most systems. In systems with highly constrained hardware, we recommend lower limits, for example 100 levels.\n\nAnother possible workaround that only works for `_.flatten`, is to pass a second argument that limits the flattening depth to 1000 or less.\n\n### References\n\n- https://github.com/jashkenas/underscore/issues/3011\n- https://underscorejs.org/#1.13.8\n- https://underscorejs.org/#flatten\n- https://underscorejs.org/#isEqual",
  "id": "GHSA-qpx9-hpmf-5gmw",
  "modified": "2026-05-05T22:01:45Z",
  "published": "2026-03-03T17:46:06Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/jashkenas/underscore/security/advisories/GHSA-qpx9-hpmf-5gmw"
    },
    {
      "type": "ADVISORY",
      "url": "https://nvd.nist.gov/vuln/detail/CVE-2026-27601"
    },
    {
      "type": "WEB",
      "url": "https://github.com/jashkenas/underscore/issues/3011"
    },
    {
      "type": "WEB",
      "url": "https://github.com/jashkenas/underscore/commit/411e222eb0ca5d570cc4f6315c02c05b830ed2b4"
    },
    {
      "type": "WEB",
      "url": "https://github.com/jashkenas/underscore/commit/a6e23ae9647461ec33ad9f92a2ecfc220eea0a84"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/jashkenas/underscore"
    },
    {
      "type": "WEB",
      "url": "https://underscorejs.org/#1.13.8"
    },
    {
      "type": "WEB",
      "url": "https://underscorejs.org/#flatten"
    },
    {
      "type": "WEB",
      "url": "https://underscorejs.org/#isEqual"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:N/I:N/A:H",
      "type": "CVSS_V3"
    },
    {
      "score": "CVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N",
      "type": "CVSS_V4"
    }
  ],
  "summary": "Underscore has unlimited recursion in _.flatten and _.isEqual, potential for DoS attack"
}



Log in or create an account to share your comment.




Tags
Taxonomy of the tags.


Loading…

Loading…

Loading…

Drag nodes to arrange them. Double-click a related vulnerability to open it, or a source to visit its link. The Physics button on the left tunes the layout and the simulation; the legend at the bottom left hides or shows nodes by kind and by sighting type.

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.
Vulnerability timeline

Sightings and recorded dates, grouped by day · UTC

JSON
How durations are calculated

Following the VTEM framework (Black Hat USA 2015), time to patch starts when the vendor or a coordinator was informed, total time to patch at discovery, and disclosure to patch at public disclosure; vendor response time runs from the report to its acknowledgement. Time to exploit and time unpatched start at the earliest discovery or report, falling back to public disclosure. Unpatched time ends at the first recorded patch, or the calculation time when no patch is recorded. Exploitation to patch shows the interval between those two events; a published proof of concept is a public exploit, not exploitation.

Each result shows its source dates. A negative interval means the end event came first; when date-only inputs leave the order open, the result says so. Date-only inputs are approximate. The earliest evidence wins: KEV report dates and sightings are labelled as such. Records describing another vulnerability are listed but never calculated from, and a PoC alone does not establish exploitation. Missing patch evidence does not prove that no patch exists. Filters below change the event list only.

Events included in the timeline

Loading timeline…


    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.