CVE-2026-73643

js-yaml is a JavaScript YAML parser and dumper. From 5.0.0 until 5.2.2, parsing a small YAML document can take exponential time when an application calls load() or loadAll() on untrusted input. In src/parser/parser.ts, readFlowCollection uses restoreState and calls parseNode a second time when a flow-sequence entry is recognized as a key: value pair. If the key is a nested flow sequence of the same shape, every level is parsed twice, causing O(2^n) work and allowing an input under 200 bytes to keep one CPU busy for minutes, block the Node.js event loop, and stall the process. No anchors, aliases, merges, tags, or nondefault options are required. This issue is fixed in version 5.2.2.

js-yaml is a JavaScript YAML parser and dumper. From 5.0.0 until 5.2.2, parsing a small YAML document can take exponential time when an application calls load() or loadAll() on untrusted input. In src/parser/parser.ts, readFlowCollection uses restoreState and calls parseNode a second time when a flow-sequence entry is recognized as a key: value pair. If the key is a nested flow sequence of the same shape, every level is parsed twice, causing O(2^n) work and allowing an input under 200 bytes to keep one CPU busy for minutes, block the Node.js event loop, and stall the process. No anchors, aliases, merges, tags, or nondefault options are required. This issue is fixed in version 5.2.2.

Summary

Parsing a small YAML document can take exponential time. An application that calls load() or loadAll() on untrusted input can be hung by a payload under 200 bytes.

Details

When an entry in a flow sequence turns out to be a key: value pair, the parser rewinds and parses that entry a second time as the key. If the key is itself a nested flow sequence of the same shape, every level is parsed twice, so the total work is O(2^n) in the nesting depth. The default maxDepth of 100 does not help, because the time is already unmanageable at about 30 to 40 levels.

Root cause, potentially the: readFlowCollection in parser.ts, the restoreState followed by a second parseNode further down.

PoC

const yaml = require('js-yaml')
const n = 30
yaml.load('[ '.repeat(n) + '1' + ' ]: 0'.repeat(n))

With default options: 22 levels takes about 1 second, 26 levels about 17 seconds, 30 levels over 2 minutes. The input stays under 200 bytes and grows linearly with n.

Impact

Denial of service. A single small request can keep one CPU busy for minutes or longer and blocks the Node event loop, so one request can stall the whole process. No anchors, aliases, merges, tags, or non default options are required, and it reproduces on the default schema.

Summary

maxTotalMergeKeys does not count empty mappings. An attacker can repeatedly merge a large sequence of them and consume significant CPU without reaching the configured limit.

Example

arr: &arr [{}, {}, {}, ...] # N empty mappings
targets:
  - <<: *arr                # repeated K times

For every target, the loader iterates all N elements of arr. This results in O(N * K) work while totalMergeKeys remains unchanged.

PoC

import { performance } from 'node:perf_hooks'
import { load, YAML11_SCHEMA } from 'js-yaml'

const n = 20000

const src =
  'arr: &arr [' + '{},'.repeat(n).slice(0, -1) + ']\n' +
  'targets:\n' +
  '  - <<: *arr\n'.repeat(n)

const started = performance.now()

load(src, { schema: YAML11_SCHEMA })

console.log(`${(performance.now() - started).toFixed(1)} ms`)

Observed results:

| N | YAML size | Time | |---:|---:|---:| | 800 | ~13 KB | ~20 ms | | 3200 | ~50 KB | ~180 ms | | 20000 | ~500 KB | ~13 s |

Impact

When merge keys are enabled, an attacker can submit a relatively small YAML document that causes prolonged CPU consumption despite the default maxTotalMergeKeys limit.

Fix

Count every merge source mapping as one budget unit in addition to counting its keys.

js-yaml is a JavaScript YAML parser and dumper. From 5.0.0 until 5.2.2, parsing a small YAML document can take exponential time when an application calls load() or loadAll() on untrusted input. In src/parser/parser.ts, readFlowCollection uses restoreState and calls parseNode a second time when a flow-sequence entry is recognized as a key: value pair. If the key is a nested flow sequence of the same shape, every level is parsed twice, causing O(2^n) work and allowing an input under 200 bytes to keep one CPU busy for minutes, block the Node.js event loop, and stall the process. No anchors, aliases, merges, tags, or nondefault options are required. This issue is fixed in version 5.2.2.

js-yaml is a JavaScript YAML parser and dumper. From 5.0.0 until 5.2.2, parsing a small YAML document can take exponential time when an application calls load() or loadAll() on untrusted input. In src/parser/parser.ts, readFlowCollection uses restoreState and calls parseNode a second time when a flow-sequence entry is recognized as a key: value pair. If the key is a nested flow sequence of the same shape, every level is parsed twice, causing O(2^n) work and allowing an input under 200 bytes to keep one CPU busy for minutes, block the Node.js event loop, and stall the process. No anchors, aliases, merges, tags, or nondefault options are required. This issue is fixed in version 5.2.2.