A flaw in Node.js's permission model allows Unix Domain Socket (UDS) connections to bypass network restrictions when --permission is enabled. Even without --allow-net, attacker-controlled inputs (such as URLs or socketPath options) can connect to arbitrary local sockets via net, tls, or undici/fetch. This breaks the intended security boundary of the permission model and enables access to privileged local services, potentially leading to privilege escalation, data exposure, or local code execution.

In the moment of this vulnerability, network permissions (--allow-net) are still in the experimental phase.

libuv before 0.10.34 does not properly drop group privileges, which allows context-dependent attackers to gain privileges via unspecified vectors.

Node.js could allow a remote attacker to traverse directories on the system. By monkey-patching Buffer internals, namely, Buffer.prototype.utf8Write, an attacker could send a specially crafted URL request containing "dot dot" sequences (/../) to read arbitrary files on the system.

A command inject vulnerability allows an attacker to perform command injection on Windows applications that indirectly depend on the CreateProcess function when the specific conditions are satisfied.

Impact

Undici allows duplicate HTTP Content-Length headers when they are provided in an array with case-variant names (e.g., Content-Length and content-length). This produces malformed HTTP/1.1 requests with multiple conflicting Content-Length values on the wire.

Who is impacted: - Applications using undici.request(), undici.Client, or similar low-level APIs with headers passed as flat arrays - Applications that accept user-controlled header names without case-normalization

Potential consequences: - Denial of Service: Strict HTTP parsers (proxies, servers) will reject requests with duplicate Content-Length headers (400 Bad Request) - HTTP Request Smuggling: In deployments where an intermediary and backend interpret duplicate headers inconsistently (e.g., one uses the first value, the other uses the last), this can enable request smuggling attacks leading to ACL bypass, cache poisoning, or credential hijacking

Patches

Patched in the undici version v7.24.0 and v6.24.0. Users should upgrade to this version or later.

Workarounds

If upgrading is not immediately possible:

1. Validate header names: Ensure no duplicate Content-Length headers (case-insensitive) are present before passing headers to undici 2. Use object format: Pass headers as a plain object ({ 'content-length': '123' }) rather than an array, which naturally deduplicates by key 3. Sanitize user input: If headers originate from user input, normalize header names to lowercase and reject duplicates

A flaw in Node.js TLS hostname handling can cause Embedded-nul hostnames can lead to silent authority rebinding due to c-string truncation in resolver bindings.

A Server-Side Request Forgery (SSRF) vulnerability was found in undici, a HTTP/1.1 client for Node.js. An attacker can manipulate the server-side application to make requests to an unintended location when they use the 'path/pathname' option in 'undici.request'.

A flaw was found in Node.js. These vulnerabilities include remote code execution, Cross-site scripting (XSS), application crashes due to missing input validation of hostnames returned by Domain Name Servers in the Node.js DNS library, which can lead to the output of wrong hostnames (leading to Domain hijacking) and injection vulnerabilities in applications using the library.

A flaw was found in Node.js, where it is vulnerable to a use-after-free attack. This flaw allows an attacker to exploit the memory corruption, which causes a change in the process behavior. The highest threat from this vulnerability is to confidentiality and integrity.

Node.js could allow a remote attacker to bypass security restrictions, caused by a path traversal bypass using non-Buffer Uint8Array objects. By sending a specially crafted request, an attacker could exploit this vulnerability to bypass the experimental permission model.

An old inffast.c optimization turns out to not be optimal anymore with modern compilers, and furthermore was not compliant with the C standard, for which decrementing a pointer before its allocated memory is undefined.

External References:

https://wiki.mozilla.org/images/0/09/Zlib-report.pdf https://docs.google.com/document/d/10i1KZS5so8xDqH2rplRa2xet0tyTvvJlLbQQmZIUIKE/edit#heading=h.t13tvnx4loq7

Upstream patch:

https://github.com/madler/zlib/commit/9aaec95e82117c1cb0f9624264c3618fc380cecb

CVE assignment:

http://seclists.org/oss-sec/2016/q4/602

An upgrade to zlib 1.2.11 to fix a number of low severity CVEs that were present in zlib 1.2.8.

Backport fix for CVE-2015-6764, a bug in v8's JSON.stringify() that can result in out-of-bounds reads for arrays.

An overflow can occur in MDC2Update() under certain circumstances resulting in an out of bounds (OOB) error. This attack is impractical on most platforms due to the size of data required to trigger the OOB error.

Node.js is impacted by this flaw but due to the impracticalities of exploiting it and the very low usage of of MDC-2, it is very low severity for Node.js users.

A security vulnerability has been discovered in the c-ares library that is bundled with all versions of Node.js. Due to the difficulty of triggering and making use of this vulnerability we currently consider this a low-severity security flaw for Node.js users.

More information at https://c-ares.haxx.se/adv20160929.html

A buffer overrun can be triggered in X.509 certificate verification, specifically in name constraint checking. Note that this occurs after certificate chain signature verification and requires either a CA to have signed the malicious certificate or for the application to continue certificate verification despite failure to construct a path to a trusted issuer. An attacker can craft a malicious email address to overflow four attacker-controlled bytes on the stack. This buffer overflow could result in a crash (causing a denial of service) or potentially remote code execution.

Many platforms implement stack overflow protections which would mitigate against the risk of remote code execution. The risk may be further mitigated based on stack layout for any given platform/compiler.

Pre-announcements of CVE-2022-3602 described this issue as CRITICAL. Further analysis based on some of the mitigating factors described above have led this to be downgraded to HIGH. Users are still encouraged to upgrade to a new version as soon as possible.

In a TLS client, this can be triggered by connecting to a malicious server. In a TLS server, this can be triggered if the server requests client authentication and a malicious client connects.

HackerOne: CVE-2023-32002 Node.js Module.load() policy Remote Code Execution Vulnerability

Affected Node.js versions can be exploited to perform HTTP desync attacks and deliver malicious payloads to unsuspecting users. The payloads can be crafted by an attacker to hijack user sessions, poison cookies, perform clickjacking, and a multitude of other attacks depending on the architecture of the underlying system.

Downloads & release details

Node.js v10.19.0 (LTS) - https://nodejs.org/en/blog/release/v10.19.0/ Node.js v12.15.0 (LTS) - https://nodejs.org/en/blog/release/v12.15.0/ Node.js v13.8.0 (LTS) - https://nodejs.org/en/blog/release/v13.8.0/

HTTP header values do not have trailing OWS trimmed

Summary

Out-of-bounds Write vulnerability in the bundled Brotli decoder in yjh0502 erl-brotli allows an attacker who supplies a crafted Brotli-compressed stream to brotli:decode/1 to corrupt decoder state memory, disclose memory contents, crash the Erlang VM, or potentially execute code.

erl-brotli compiles its own copy of the Google Brotli C sources under csrc into a NIF, and brotlidecoderdecompressstream in csrc/brotlinif.c hands the input to BrotliDecoderDecompressStream. In csrc/dec/decode.c, DecodeBlockTypeAndLength returns false for a block switch in a category with at most one block type, and the non-streaming callers DecodeLiteralBlockSwitch, DecodeCommandBlockSwitch and DecodeDistanceBlockSwitch discard that result, so decoding continues with inconsistent state. The decoder runs inside a NIF, so memory corruption or a crash takes down the whole VM rather than the calling process. Because the package carries its own copy of the library, the Brotli version installed on the host does not matter.

This issue affects brotli: from 0.3.0 before 0.3.4.

Impact

An attacker who can get an application to decode a Brotli stream they control, for example an HTTP client that decompresses Content-Encoding: br responses with this package, can crash the Erlang VM and potentially run code inside it.

A flaw in Node.js’s Permissions model allows attackers to bypass --allow-fs-read and --allow-fs-write restrictions using crafted relative symlink paths. By chaining directories and symlinks, a script granted access only to the current directory can escape the allowed path and read sensitive files. This breaks the expected isolation guarantees and enables arbitrary file read/write, leading to potential system compromise.

Summary

Two issues in undici's cache interceptor, both fixed by the same patch on lib/util/cache.js:

1. Shared-cache disclosure: Responses with malformed qualified Cache-Control: private directives such as private="" or private="," can be incorrectly stored in the default shared cache, then served to a later caller with the same cache key. 2. Parse-time crash: Mixed unqualified-and-qualified private directives in the same header (such as public, max-age=60, private, private="hdr") cause an uncaught TypeError in the cache-control parser, terminating the request.

Impact

Shared-cache disclosure

Applications using interceptors.cache() in shared mode may cache a user-specific response and serve it to a later caller with the same cache key. This can disclose private response bodies and headers, including Set-Cookie.

Required conditions:

- the cache interceptor is enabled in shared mode, including the default configuration; - an upstream returns a malformed directive such as Cache-Control: public, max-age=300, private=""; - another request later matches the same cache key, without a separating Vary header.

Parse-time crash

Applications using interceptors.cache() against an upstream that returns a Cache-Control header combining unqualified private with qualified private="..." see an uncaught TypeError: output.private.concat is not a function during response handling. The request rejects; depending on the consumer's error handling, the process may exit.

Details

private="" is parsed as { private: [''] }. The shared-cache guard only rejects private === true, so the response can be stored. When served from cache, the previous user's body and headers may be returned to a different user.

For the crash variant, an unqualified private directive sets output.private = true, then a subsequent qualified private="hdr" directive attempts output.private.concat(['hdr']), which throws because boolean has no concat method.

The patch routes the qualified-directive path through a shared helper that normalizes empty-after-trim arrays to true and preserves existing true values, closing both vectors.

Patches

Upgrade to undici 7.29.0 or 8.9.0. Both releases fix the qualified private directive handling that caused the shared-cache storage and the parser crash.

Workarounds

Until patched, avoid shared interceptors.cache() for user-specific responses, use type: 'private', or disable caching for affected origins.

Credit

Disclosure variant reported by @h0rk1p via HackerOne report #3817497.

Impact

undici's BalancedPool passes its constructor options through a JSON-based deep clone (JSON.parse(JSON.stringify(...))) before forwarding them to each per-upstream Pool. JSON cannot represent functions, so a caller-supplied connect or tls option containing a checkServerIdentity callback (or a custom connector function) is silently dropped before it reaches the TLS layer. As a result, a TLS peer whose certificate a custom checkServerIdentity was written to reject, but which passes Node's default hostname and chain checks, is silently accepted when the request is made through BalancedPool. Client, Pool, Agent, and RoundRobinPool destructure connect/tls before the clone and are not affected. Only applications that use BalancedPool with a function-valued connect/tls option (such as a custom checkServerIdentity or connector) are affected.

Patches

Upgrade to 7.29.1 or 8.10.2. BalancedPool now preserves the connect and tls options outside the JSON clone, so custom TLS verification callbacks are forwarded to each upstream unchanged.

Workarounds

Use Client, Pool, or Agent instead of BalancedPool for connections that rely on a custom checkServerIdentity or connector, until upgraded.

Impact

When an originless Agent composes the redirect and cache interceptors as compose(redirect(), cache()) with a shared cache (type: 'shared'), the cache builds its key from the original request URL before the redirect handler changes the origin. The response produced after a redirect is then stored under the original key, even when the redirect crossed to a different origin.

An attacker who controls the target of a redirect from a trusted URL can have their response cached under that trusted URL. A later caller requesting the trusted URL receives the attacker's body from the shared cache without either origin being contacted, allowing persistent cache poisoning. The same key confusion also discloses private data: when an authenticated request returns a private, no-store redirect to a user-specific resource, the final response is stored under the original URL and can be served to a later unauthenticated caller.

The cache interceptor was introduced in undici 7.0.0, and all 7.x and 8.x releases before the patched versions are affected. Applications are affected only when an originless Agent, a shared MemoryCacheStore or SqliteCacheStore, and the composition compose(redirect(), cache()) are all used. Origin-bound Client and Pool dispatchers, and the reverse composition compose(cache(), redirect()), are not affected.

Patches

Upgrade to 7.30.1 or 8.12.2. Before storing a response, undici now verifies that the origin that produced it still matches the cache key, and preserves each redirect hop's cache-control policy, so cross-origin redirect responses and responses behind a no-store or private hop are no longer cached under the original key.

Workarounds

Compose the interceptors as compose(cache(), redirect()), use per-origin dispatchers and cache stores, or disable shared caching on redirect-capable requests.

Impact

undici's Agent passes its constructor options through a JSON-based deep clone (JSON.parse(JSON.stringify(...))) before forwarding them to each per-origin Pool. JSON cannot represent functions, so a caller-supplied tls option containing a checkServerIdentity callback is silently dropped before it reaches the TLS layer. As a result, a TLS peer whose certificate a custom checkServerIdentity was written to reject, but which passes Node's default hostname and chain checks, is silently accepted when the request is made through Agent. Because Agent is the default dispatcher behind request, fetch, and setGlobalDispatcher, this is the path most applications use. EnvHttpProxyAgent is also affected on its direct and NOPROXY routes, which delegate to an internal Agent.

Only applications that pass a function-valued tls option (such as a custom checkServerIdentity for certificate or public-key pinning) are affected. Passing the same options through connect is not affected. This is the Agent counterpart of the BalancedPool certificate validation bypass fixed in CVE-2026-84961.

Patches

Upgrade to undici 6.29.1, 7.30.1, or 8.12.2. Agent now preserves the tls option outside the JSON clone, so custom TLS verification callbacks are forwarded to each origin unchanged.

Workarounds

Pass the TLS options through connect instead of tls. For proxied EnvHttpProxyAgent destinations, use requestTls / proxyTls.

A vulnerability was found in NodeJS due to weak randomness in the WebCrypto keygen within the SecretKeyGenTraits::DoKeyGen() in src/crypto/cryptokeygen.cc. Node.js made calls to EntropySource() in SecretKeyGenTraits::DoKeyGen(). However, it does not check the return value and assumes the EntropySource() always succeeds, but it can and sometimes will fail. This flaw allows a remote attacker to decrypt sensitive information.

Node.js was affected by OpenSSL vulnerability CVE-2017-3737 in regards to the use of SSLread() due to TLS handshake failure. The result was that an active network attacker could send application data to Node.js using the TLS or HTTP2 modules in a way that bypassed TLS authentication and encryption.

Node.js could allow a remote attacker to bypass security restrictions, caused by improper path traversal sequence sanitization. By using a path traversal attack, an attacker could exploit this vulnerability leading to filesystem permission model bypass.

A buffer overflow flaw was found in the V8 component of the Chromium browser.

Upstream bug(s):

https://code.google.com/p/chromium/issues/detail?id=606115

External References:

http://googlechromereleases.blogspot.com/2016/05/stable-channel-update.html

Impact

When using Socks5ProxyAgent, undici reuses a single connection pool across different origins without verifying that the pool's origin matches the requested origin. All requests are dispatched through the pool connected to the first origin, regardless of the intended destination.

This causes cross-origin request routing: credentials and request data intended for origin B are sent to origin A, responses from the wrong origin are trusted, and HTTPS requests may be silently downgraded to HTTP.

Impacted users are applications that use Socks5ProxyAgent (directly or via setGlobalDispatcher) and make requests to more than one origin.

This was introduced in undici 7.23.0 via #4385 and affects all versions through 8.1.0.

Patches

Upgrade to undici v7.28.0 or v8.2.0

Workarounds

Use a separate Socks5ProxyAgent instance per origin, or avoid using Socks5ProxyAgent with multiple origins.