Appium is a cross-platform automation framework for all kinds of apps, built on top of the W3C WebDriver protocol. Prior to 7.2.6, @appium/support ZIP extraction accepts an archive entry that creates a symbolic link to a location outside the extraction directory and then accepts a later regular-file entry whose legitimate name resolves through that link. The lexical containment check examines entry names without validating the symlink target or the real destination, so an attacker-controlled archive can write files to arbitrary locations writable by the Appium process. This is distinct from the earlier parent-directory entry-name traversal fixed in 7.0.6 because the bypass uses no parent-directory or absolute entry name. Depending on the writable destination, the arbitrary file write can overwrite configuration, startup scripts, or executables and lead to code execution. This issue is fixed in version 7.2.6.
Summary
@appium/support contains a ZIP extraction implementation (extractAllTo() via ZipExtractor.extract()) with a path traversal (Zip Slip) check that is non-functional. The check at line 88 of packages/support/lib/zip.js creates an Error object but never throws it, allowing malicious ZIP entries with ../ path components to write files outside the intended destination directory. This affects all JS-based extractions (the default code path), not only those using the fileNamesEncoding option.
Severity
Medium (CVSS 3.1: 6.5)
CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:H/A:N
- Attack Vector: Network — malicious ZIP files can be supplied over the network (e.g., app packages via URL) - Attack Complexity: Low — no special conditions required beyond providing a crafted ZIP - Privileges Required: None — no authentication needed to supply a malicious archive - User Interaction: Required — a user or automation system must initiate extraction of the attacker's archive - Scope: Unchanged — impact stays within the file system permissions of the Appium process - Confidentiality Impact: None — the vulnerability enables file writes, not reads - Integrity Impact: High — arbitrary file write to any location writable by the process - Availability Impact: None — no direct availability impact
Affected Component
- packages/support/lib/zip.js — ZipExtractor.extract() (line 88) and ZipExtractor.extractEntry() (lines 111-145)
CWE
- CWE-22: Improper Limitation of a Pathname to a Restricted Directory ('Path Traversal')
Description
Missing throw renders Zip Slip protection non-functional
The ZipExtractor.extract() method contains a path traversal check intended to prevent Zip Slip attacks. However, the check creates an Error object as a bare expression without the throw keyword, making it a no-op:
javascript // packages/support/lib/zip.js, lines 80-93 const destDir = path.dirname(path.join(dir, fileName)); try { await fs.mkdir(destDir, {recursive: true});
const canonicalDestDir = await fs.realpath(destDir); const relativeDestDir = path.relative(dir, canonicalDestDir);
if (relativeDestDir.split(path.sep).includes('..')) { new Error( // <-- BUG: missing throw Out of bound path "${canonicalDestDir}" found while processing file ${fileName} ); }
await this.extractEntry(entry); // extraction proceeds unconditionally
The presence of a well-formatted error message and surrounding try/catch block (lines 95-99) strongly suggests the throw keyword was accidentally omitted.
yauzl does not provide its own traversal protection
The upstream yauzl library explicitly does not offer path traversal protection regardless of the decodeStrings setting. This means the vulnerability affects all JS-based extractions through ZipExtractor, not only those where fileNamesEncoding is set. The fileNamesEncoding option bypasses yauzl's string decoding (decodeStrings: false), but even with decodeStrings: true, yauzl passes through ../ path components without rejection.
Unprotected write sinks
The extractEntry method writes to attacker-controlled paths with no additional validation:
javascript // packages/support/lib/zip.js, lines 111-145 const fileName = this.extractFileName(entry); const dest = path.join(dir, fileName); // resolves ../pwned.txt outside dir // ... await fs.symlink(link, dest); // symlink creation (line 143) await pipeline(readStream, fs.createWriteStream(dest, {mode: procMode})); // file write (line 145)
Additionally, extractEntryTo() (line 263) used by readEntries() has no traversal check at all:
javascript const dstPath = path.resolve(destDir, entry.fileName); // no validation
Default code path is vulnerable
The extractAllTo() function uses the JS-based ZipExtractor by default. The system unzip fallback (useSystemUnzip: true) must be explicitly enabled and only provides protection if the system binary succeeds:
javascript // packages/support/lib/zip.js, lines 203-210 if (opts.useSystemUnzip) { try { await extractWithSystemUnzip(zipFilePath, dir); return; } catch (err) { log.warn('unzip failed; falling back to JS: %s', err.stderr || err.message); // Falls through to the vulnerable JS implementation } }
Proof of Concept
bash 1) Install deps for the support package cd packages/support npm install --omit=dev --ignore-scripts --no-audit --no-fund --workspaces=false
2) Create a malicious ZIP containing a traversal entry export WORK=/tmp/appiumzipslippoc rm -rf "$WORK" && mkdir -p "$WORK/dest" python3 - <<'PY' import zipfile, os work = os.environ['WORK'] zippath = os.path.join(work, 'evil.zip') with zipfile.ZipFile(zippath, 'w') as z: z.writestr('../pwned.txt', 'ZIPSLIPMARKER') print('created', zippath) PY
3) Extract with the JS implementation (default path, no fileNamesEncoding needed) node --experimental-default-type=module --experimental-specifier-resolution=node - <<'NODE' import path from 'node:path'; import fs from 'node:fs/promises'; import { extractAllTo } from './lib/zip.js';
const work = process.env.WORK; const zipPath = path.join(work, 'evil.zip'); const dest = path.join(work, 'dest');
await extractAllTo(zipPath, dest, { useSystemUnzip: false });
const outside = path.join(work, 'pwned.txt'); console.log('outside exists?', await fs.stat(outside).then(() => true, () => false)); console.log('outside content:', (await fs.readFile(outside, 'utf8')).trim()); NODE Expected output: outside exists? true outside content: ZIPSLIPMARKER
Impact
- Arbitrary file write: An attacker can write files to any location writable by the Appium process, outside the intended extraction directory. - Arbitrary symlink creation: Malicious ZIP entries with symlink attributes can create symlinks pointing to arbitrary targets, enabling further attacks on subsequent file operations. - Potential code execution: By overwriting scripts, configuration files, nodemodules contents, cron jobs, shell profiles, or other executable artifacts, arbitrary file write can chain into remote code execution. - Affects all JS-based extractions: The default code path (without useSystemUnzip: true) is vulnerable regardless of whether fileNamesEncoding is set.
Recommended Remediation
Option 1: Add the missing throw keyword (preferred — minimal fix)
javascript // packages/support/lib/zip.js, line 88 if (relativeDestDir.split(path.sep).includes('..')) { throw new Error( // Add throw Out of bound path "${canonicalDestDir}" found while processing file ${fileName} ); }
This is the lowest-risk fix: it restores the clearly intended behavior of the existing check. The try/catch block at lines 95-99 will catch the error, set canceled = true, close the zip, and reject the promise — exactly the designed error-handling flow.
Option 2: Add traversal protection to extractEntryTo as well
The extractEntryTo function (line 262) also lacks a traversal check. For defense-in-depth, add validation there too:
javascript async function extractEntryTo(zipFile, entry, destDir) { const dstPath = path.resolve(destDir, entry.fileName); const canonicalDest = path.resolve(dstPath); const canonicalDestDir = path.resolve(destDir); if (!canonicalDest.startsWith(canonicalDestDir + path.sep) && canonicalDest !== canonicalDestDir) { throw new Error( Out of bound path "${canonicalDest}" found while processing file ${entry.fileName} ); } // ... rest of function }
Credit
This vulnerability was discovered and reported by bugbunny.ai.