GHSA-rgwj-5xj2-c3m3

ADVISORY - github

Summary

Vulnerability Details

File: lib/compressed_protocol.js Line: 43 (zlib.inflate(body, (err, data) => { ... }) inside handleCompressedPacket)

Root Cause

When a connection is created with compress: true (and the server advertises CLIENT_COMPRESS), every incoming packet is unwrapped by handleCompressedPacket() in lib/compressed_protocol.js, which calls:

zlib.inflate(body, (err, data) => { ... });

No options object (in particular, no maxOutputLength) is passed. Node's zlib convenience methods default maxOutputLength to buffer.kMaxLength, which on this platform is Number.MAX_SAFE_INTEGER — i.e. effectively unbounded until the process runs out of memory. The 3-byte "length of payload before compression" field in the compressed-packet header is read (packet.readInt24()) but is only used to branch on !== 0; it is never used to cap or validate the actual inflate output size, and the real decompressed size is determined purely by the attacker-supplied deflate stream.

Because DEFLATE can reach compression ratios over 1000:1 for crafted repetitive input, an attacker who controls (or MITMs, on a non-TLS connection) the MySQL server endpoint can send a single small compressed packet that expands to gigabytes in the client's memory — a classic decompression-bomb / "zip bomb" applied to MySQL's client-compression protocol.

Attack Scenario

  1. Application connects with mysql2/mysql2/promise using compress: true (a documented option for reducing bandwidth, commonly used for cloud/WAN DB connections).
  2. The connection target is attacker-controlled or attacker-compromised, or an attacker MITMs a non-TLS connection.
  3. Right after authentication succeeds, the malicious endpoint sends one crafted compressed packet whose deflate stream is small on the wire (hundreds of KB) but decompresses to several GB.
  4. zlib.inflate() starts allocating memory for the full decompressed output with no ceiling.
  5. The Node.js process's RSS grows uncontrolled until OOM-kill or crash — no query needs to be issued by the client; the malicious packet alone is enough.

Impact

Denial of Service of the client application (process crash / OOM) — not the database itself. No authentication bypass or data exposure. Requires compress: true plus a malicious/compromised server or MITM position.

Vulnerable Code

function handleCompressedPacket(packet) {
  const connection = this;
  const deflatedLength = packet.readInt24();
  const body = packet.readBuffer();

  if (deflatedLength !== 0) {
    connection.inflateQueue.push((task) => {
      zlib.inflate(body, (err, data) => {
        if (err) {
          connection._handleNetworkError(err);
          return;
        }
        connection._bumpCompressedSequenceId(packet.numPackets);
        connection._inflatedPacketsParser.execute(data);
        task.done();
      });
    });
  } else {
    ...
  }
}

Recommended Fix

const MAX_INFLATED_PACKET_SIZE = 1 * 1024 * 1024 * 1024; // e.g. 1 GiB, ideally configurable

zlib.inflate(body, { maxOutputLength: MAX_INFLATED_PACKET_SIZE }, (err, data) => {
  if (err) {
    connection._handleNetworkError(err);
    return;
  }
  ...
});

maxOutputLength makes zlib.inflate abort with ERR_BUFFER_TOO_LARGE as soon as the decompressed size would exceed the cap, routing into the exact same (already-existing) errconnection._handleNetworkError(err) path, so no new error-handling logic is required.

Verification

Dynamically confirmed on v3.23.0 (HEAD) using a minimal rogue "MySQL server" built on node-mysql2's own server-mode helpers (mysql.createServer, Packets.Handshake, connection.writeOk()). The rogue server completes a real handshake advertising CLIENT_COMPRESS, then writes one raw compressed frame (509,604 bytes on the wire — a zlib deflate of 500 MB of zero bytes, ratio 1028.8:1) directly to the socket. A normal mysql.createConnection({ ..., compress: true }) victim client — which never issues any query — had its RSS grow from 74.3 MB to 1115.0 MB after receiving that single packet, before erroring out with PROTOCOL_UNEXPECTED_PACKET once the client tried to parse the inflated zero-filled buffer as MySQL packets. The memory allocation happens unconditionally before any content validation.

Common Weakness Enumeration (CWE)

ADVISORY - github

Improper Handling of Highly Compressed Data (Data Amplification)


GitHub

CREATED

UPDATED

EXPLOITABILITY SCORE

-

EXPLOITS FOUND
-
COMMON WEAKNESS ENUMERATION (CWE)

CVSS SCORE

N/Amedium
PackageTypeOS NameOS VersionAffected RangesFix Versions
mysql2npm--<=3.23.03.23.1

CVSS:3 Severity and metrics

The CVSS metrics represent different qualitative aspects of a vulnerability that impact the overall score, as defined by the CVSS Specification.

The vulnerable component is bound to the network stack, but the attack is limited at the protocol level to a logically adjacent topology. This can mean an attack must be launched from the same shared physical (e.g., Bluetooth or IEEE 802.11) or logical (e.g., local IP subnet) network, or from within a secure or otherwise limited administrative domain (e.g., MPLS, secure VPN to an administrative network zone). One example of an Adjacent attack would be an ARP (IPv4) or neighbor discovery (IPv6) flood leading to a denial of service on the local LAN segment (e.g., CVE-2013-6014).

A successful attack depends on conditions beyond the attacker's control, requiring investing a measurable amount of effort in research, preparation, or execution against the vulnerable component before a successful attack.

The attacker is unauthorized prior to attack, and therefore does not require any access to settings or files of the vulnerable system to carry out an attack.

The vulnerable system can be exploited without interaction from any user.

An exploited vulnerability can only affect resources managed by the same security authority. In this case, the vulnerable component and the impacted component are either the same, or both are managed by the same security authority.

There is no loss of confidentiality.

There is no loss of trust or accuracy within the impacted component.

There is a total loss of availability, resulting in the attacker being able to fully deny access to resources in the impacted component; this loss is either sustained (while the attacker continues to deliver the attack) or persistent (the condition persists even after the attack has completed). Alternatively, the attacker has the ability to deny some availability, but the loss of availability presents a direct, serious consequence to the impacted component.