Skip to content
Tutorials Center
My learning

Topics

Lesson 19 of 38 ยท Repeater and Intruder

Safe, Rate-Limited Testing

Keep automated testing safe: resource pools, throttling, request budgets, lockout awareness, stop conditions and monitoring, plus how defenders rate-limit abuse.

Automation is where well-meaning testers cause outages. A few hundred requests a second can slow a small site to a crawl, fill logs, trip alarms, lock real users out of accounts and, in the worst case, take the service down. Professional testing is controlled: you decide how many requests, how fast, and when to stop, before you press start.

This lesson shows the controls Burp gives you and the habits that go with them. It also looks at the same subject from the defender's side: how rate limiting and lockouts work, so you can test that they work.

Why volume is a risk

  • Availability. Fragile servers (old hardware, shared hosting, test environments) can fall over.
  • Side effects. Each request can create a record, send an email or an SMS, or charge a fee.
  • Account lockouts. Repeated failures lock accounts, which may belong to real people.
  • Detection and response. Security teams may block your IP, or start an incident response because your traffic looks like an attack.
  • Cost. Metered APIs and cloud services bill per request.

The agreement should state acceptable rates and windows; if it does not, ask.

Controls in Burp

Resource pools (Professional)

A resource pool limits concurrency and adds delays for a task. Settings include maximum concurrent requests, delay between requests (fixed or random), and an increasing delay. In Intruder (Professional) you assign the attack to a pool; setting "1 concurrent request, 500 ms delay" caps you at about two requests a second regardless of what the attack contains.

In Community, Intruder is already throttled by design, which is a built-in safety feature. Do not try to defeat it on systems that are not yours; for a lab, simply accept the pace or use a faster tool you control.

Request budget

Before starting, work out the number of requests:

  • sniper: positions x payloads
  • pitchfork: length of the shortest list
  • cluster bomb: product of the list lengths

Compare the total with your budget. If a plan says 2,000 requests per hour and your attack is 50,000, change the plan.

Stop conditions

Decide what makes you pause immediately:

  • error rate climbing (5xx responses),
  • response times growing,
  • unexpected 429 Too Many Requests or block pages,
  • a request that creates real data,
  • a message from the client.

Watch the attack's results as it runs, and be ready to Pause or Cancel.

Prefer targeted tests

Test with the smallest set that answers the question. To show that an endpoint lacks rate limiting you do not need 100,000 requests; 30 requests in 10 seconds with no 429 is enough evidence for the report.

How defenders rate-limit

Understanding the controls helps you test them, and describe fixes.

Control How it works What a tester checks
Fixed window N requests per minute per key Does the limit exist on login, reset and API?
Token bucket Tokens refill at a steady rate; bursts allowed Does bursting still stay under the cap?
Per-account lockout Lock after N failures Is the message generic? Can it be abused to lock others out?
Per-IP throttle Slow down one address Can headers like X-Forwarded-For spoof it?
CAPTCHA / step-up Challenge after suspicious activity Is it enforced on every route, including APIs?
Alerting Notify staff of spikes Did anyone notice the test?

A finding is rarely "I could send many requests". It is "the password-reset endpoint has no limit, so codes can be enumerated" or "login lockout is keyed on the client-supplied header, so it can be bypassed".

Try it: a token bucket rate limiter

A stand-in implementation of a token bucket on a simulated clock (no sleeping, so it runs instantly). It shows what a defender's limit looks like and why a steady slow tester stays under it.

Python Stand-in: a token bucket limiter

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The burst gets five requests through and then is blocked (X). The steady and slow patterns are never blocked, which shows both why throttled testing is gentle and why an attacker who stays under the limit may go unnoticed unless the defender also detects patterns over time.

Try it: a request budget and a stop condition

A second stand-in plans an attack's size, throttles it with a fixed delay on a simulated clock, and stops automatically if the error rate rises, the habit to build into every run.

Python Stand-in: budget, delay and automatic stop

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

A pre-flight checklist

  1. Written permission covers automated testing and the rate.
  2. Scope contains only the intended hosts.
  3. Request count is calculated and within the budget.
  4. A delay or resource pool is set.
  5. Test accounts are used; real users cannot be locked out or contacted.
  6. Stop conditions and a contact are known.
  7. Someone on the client side knows when the test runs, if the agreement requires it.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. Which control limits how fast Burp Professional sends requests for a task?

    Resource pools cap concurrency and add delays so the attack stays within an agreed rate.
  2. 2. You run a cluster bomb with lists of 200 and 150 items. How many requests is that?

    Cluster bomb multiplies the list lengths: 200 x 150 = 30,000.
  3. 3. Which is the best reason to stop an automated run immediately?

    Rising 5xx errors or unintended side effects mean your test is harming the target. Pause, assess and talk to the owner.
  4. 4. What evidence is usually enough to report a missing login rate limit?

    A small, controlled sample proves the control is missing without causing harm.

Summary

  • Volume is the main risk of automation: outages, side effects, lockouts, detection and cost.
  • Use resource pools or delays, calculate the request budget first, and define stop conditions.
  • Test with the smallest set that proves the point; a modest burst is enough evidence.
  • Know the defender's controls: fixed windows, token buckets, lockouts, per-IP limits, CAPTCHA and alerting.
  • Run a written pre-flight checklist for every automated test.

Next: Continue with the rest of the Burp Suite track