👹 Monster333

See past every gate. Zero false positives.

A deep HTTP access-control bypass, WAF-evasion and recon engine for authorized bug bounty hunting and penetration testing. Point it at a URL that returns 401/403 and Monster333 runs 318 verified request mutations across 14 technique families — then filters every result through a six-gate false-positive elimination pipeline so a "CONFIRMED" finding actually means something.

Built by Md Tanjimul Islam Sifat · Community: SftSec Tim — https://discord.gg/XAu8PdDeeT

---

⚡ Quick start

# no install needed — one dependency (requests)
python3 monster333.py demo          # boot a local vulnerable lab and self-scan it

python3 monster333.py scan -u https://target.example/admin
python3 monster333.py recon -u https://target.example
python3 monster333.py analyze -u https://target.example/admin
python3 monster333.py doctor         # environment self-test

Or install properly:

pipx install .          # or: pip install .
monster333 scan -u https://target.example/admin

demo is the fastest way to trust the tool — it plants 9 classic vulnerabilities in a local lab (header trust, path mutation, Unicode, UA allowlist, h2c, CL.TE desync, cache deception…) and proves Monster333 finds all of them with zero false positives, offline, before you ever touch a real target.

---

🎯 What it does

Modern "403 Forbidden" is an opinion expressed by one node in a proxy chain — the edge proxy, the WAF, and the origin frequently parse the same bytes differently. Monster333 systematically exploits those seams:

Command	Purpose
scan	318 verified mutations across 14 families, FP-filtered, double-verified
recon	robots/sitemap harvesting, sensitive-file exposure, CORS, headers, cookies, methods, rate-limiter, TLS SANs, tech hints
analyze	WAF/CDN fingerprint (17 vendors) + recommended technique focus
demo	Local vulnerable lab + full self-scan + proof artifacts
doctor	Environment self-test

Technique families (-t)

Family	What it throws at the target
ip	78 IP/identity spoof headers (X-Forwarded-For, CF-Connecting-IP, True-Client-IP, integer/hex IPv4 forms, Forwarded…)
host	Host header + URL-override headers (X-Original-URL, X-Rewrite-URL, X-HTTP-DestinationURL…)
path	54 path mutations (/x/, //x, /./x, /x..;/, /x%20, /x%00, /x#.php…)
unicode	Fullwidth solidus/full stop, overlong UTF-8, double-encoding (%EF%BC%8F, %252f, %C0%AF…)
protocol	Method override, content-type/accept mutations, connection semantics
service_mesh	Envoy/Istio/Kong/AWS trust headers (X-Envoy-Internal, XFCC SPIFFE, trace headers…)
trust	Magic auth/debug headers (X-No-Auth, X-Role, X-Debug, X-WAF-Bypass…)
useragent	17 crawler/browser impersonations (Googlebot, Bingbot, mobile…)
methods	23 HTTP verbs incl. PURGE, PROPFIND, DEBUG, SYNC, CHECKIN…
wcd	25 web-cache-deception suffixes (;.js, /logo.png, %0a.js…)
reflection	Unkeyed-header reflection probes (cache-poisoning vector)
h2c	Raw-socket Upgrade: h2c probe (101 detection)
smuggling	Self-contained CL.TE / TE.CL / TE.TE desync probes (CWE-444)

Combine freely: -t ip,path,unicode or -t all (default).

---

🔬 The zero-false-positive pipeline

Most 403-bypass tools spray mutations and report every 200 — then you waste an hour on noise. Monster333 runs every candidate through six gates before it may print CONFIRMED:

GATE 0  preflight            open/404/429/5xx baselines are refused outright —
                             only 401/403/405-gated paths are probed
GATE 1  baseline divergence  candidate must differ from the blocked response
GATE 2  random-path control   a garbage path must not produce the same response
GATE 3  junk-header control   X-Monster-Control-* headers must not "unlock" anything
GATE 4  soft-block scanner   200-with-WAF-block-page → SOFT_BLOCK, never CONFIRMED
GATE 5  double verification  every candidate re-sent; only stable 2xx survives —
                             3xx redirects are demoted to informational

Everything else is visibly bucketed as NOISE / FLAKY / SOFT_BLOCK / REDIRECT — nothing is hidden, nothing is guessed. A confirmation means: this exact request, reproduced twice, opens a 2xx door that the baseline and all controls keep closed.

---

🛡️ Scope & safety (by architecture, not by docs)

• Explicit-target allowlist only. The scope engine fails closed — no host outside your -u/-U list is ever contacted. No DNS pivots, no auto-discovery, no redirect-following out of scope.
• Request budget + hard ceiling (--budget, default hard cap 5000) — the tool stops cleanly rather than hammering.
• Rate limiting by default (--delay, --concurrency), 429-aware.
• Nothing destructive — all payloads are request mutations; no payload deletes, corrupts, or floods.
• Clean Ctrl+C — first press finishes in-flight probes and writes partial artifacts; second force-quits.

📦 Output

Every run writes a timestamped directory under monster333-runs/:

monster333-runs/20260912-094142-http---127.0.0.1-18996-admin-settings/
├── results.json    # full machine-readable results (stable schema)
├── report.md       # human report — every confirmed bypass + curl repro
└── report.html     # self-contained dark-mode HTML report

Every finding carries: technique, exact mutation, status, verification reason, and a copy-paste curl reproduction command.

---

🧰 Options

monster333 scan -u URL [options]
  -t, --technique FAMILIES   comma/repeat list (default: all)
  -m, --method METHOD        probe method (default GET)
  -U, --url-file FILE        bulk targets
  --delay SEC                per-request delay (default 0.05)
  --concurrency N            parallel workers (default 4)
  --timeout SEC              HTTP timeout (default 10)
  --budget N                 max requests this run
  --proxy URL                e.g. http://127.0.0.1:8080 (Burp)
  --verify-passes N          resend count to confirm (default 2)
  --json                     machine-readable stdout
  --quiet / --verbose / --debug / --no-color

Examples:

# focused scan through Burp, slow and polite
monster333 scan -u https://api.target.example/v1/admin \
  -t ip,service_mesh,unicode --proxy http://127.0.0.1:8080 --delay 0.5

# bulk + JSON pipeline
monster333 scan -U blocked-urls.txt --json --quiet | jq '.results'

# recon one host
monster333 recon -u https://target.example --verbose

---

🧪 Testing & development

make test     # 57 tests: unit + preflight gates + lab + integration + CLI (offline, 127.0.0.1 only)
make lab      # boot the vulnerable demo lab standalone
make lint     # byte-compile check

The test suite boots real lab instances and proves: every planted vuln is detected, control headers never unlock anything, open/404/429/5xx baselines produce zero claims, budgets stop cleanly, scope denials raise, reports are valid JSON.

---

❌ What Monster333 does NOT do

• No subdomain enumeration, port scanning, or exploit detonation.
• No TCP-level fragmentation (MSS clamping/segment overlap) — those need raw kernel sockets and belong in lab-only tooling.
• No JA4/TLS-fingerprint impersonation — use curl-impersonate/curl_cffi alongside if you need it.
• No submission of reports anywhere. Evidence stays on your disk.

⚠️ Legal

Authorized testing only. You are responsible for having written permission for every target you scan. Unauthorized use is illegal under CFAA (US), Computer Misuse Act (UK), ICT Act (BD) and equivalent laws worldwide.

---

📜 Changelog & license

• CHANGELOG.md
• MIT License — © 2026 Md Tanjimul Islam Sifat

Join the community: SftSec Tim — https://discord.gg/XAu8PdDeeT