# Why Your Friction-Based Defenses Are Stopping More Users Than Attackers

> CAPTCHAs, OTPs and repeated checks can cost customers time while attackers outsource the work. Learn how to measure the impact and reduce abuse with less friction.

Published: 2026-11-13  
URL: https://www.kairal.io/blog/why-friction-based-defenses-stop-more-users-than-attackers/

### Context

A customer reaches checkout, gets a puzzle wrong and tries again. Then a verification code arrives late. The purchase was almost complete, but now leaving feels easier.

An attacker can approach those same checks with a budget, automation and a supplier. If a service can handle the challenge cheaply enough, the attacker pays and continues.

A high challenge count can hide lost customers and ongoing abuse. What matters is how many customers finish what they came to do and how much harmful activity gets through.

In an [August 2023 USENIX study](https://www.usenix.org/conference/usenixsecurity23/presentation/searles), researchers studied 1,400 participants solving 14,000 CAPTCHAs. They compared how long different challenges took, how users experienced them and when people gave up.

### TL;DR

- Attackers can outsource repetitive checks through solver and account services.
- Completing a challenge or verifying an email address does not establish honest intent.
- Repeated friction can affect purchases, activation and retention while abuse continues.
- Keep useful security controls, assess the action being attempted, and measure customer outcomes alongside confirmed abuse.

### Why attackers can absorb friction

A genuine user usually wants one outcome: buy a ticket, contact a seller, read an article or try a product. Every extra step delays that outcome.

![Visual challenge presented to a user](https://www.kairal.io/blog/_img/4170b689-e5f6-811c-841b-e575378114fe/8a079f6f-9d42-494b-8e01-5418be827075/visual-challenge-redacted.png)

An attacker may be running many attempts. They can accept failed accounts and paid verification if the activity remains profitable. This is particularly damaging when your business pays to acquire a customer who then gives up at a security check.

The pricing advertisement below lists charges per thousand solves. One row advertises €1.4 per thousand, equivalent to €0.0014 each.

![Anonymized solver pricing advertisement](https://www.kairal.io/blog/_img/4170b689-e5f6-811c-841b-e575378114fe/64165cb4-d2f9-4ddd-a1f9-5bdb773d759a/solver-pricing-redacted.png)

Raising the difficulty can still help if it materially reduces successful abuse or makes the activity uneconomic. The business question is what the extra check achieves compared with the customers it interrupts.

Our [classifieds analysis](https://www.kairal.io/blog/why-classifieds-users-are-easy-targets-for-scammers/) showed the same economics across scraping, free and paid solvers, and account-generation offers. These components can support abuse in other industries too. Their availability reduces how much an attacker must build personally.

### How attackers get through friction-based checks

#### They obtain the data and accounts they need

Scraping collects information automatically. An attacker may use it to identify people to contact, products to target or content to copy.

Accounts are also sold as a service. The advertisements in our classifieds article offered account generation, including registration without a one-time password, or OTP.

For a business, the useful question is what access an account receives after registration. Can a newly verified profile immediately send many messages, claim valuable credits or reserve scarce inventory?

Request logs can show registration volume and timing. To understand the damage, you also need records of the actions those accounts take.

#### They outsource the check

A solver is software or a service intended to complete a challenge for an automated workflow. Some services use human workers; others advertise automated solving.

The workflow below presents challenge handling as a routine part of an existing task. That is the commercial proposition: automation encounters a check, obtains a result and continues.

![Anonymized solver workflow](https://www.kairal.io/blog/_img/4170b689-e5f6-811c-841b-e575378114fe/5329e15e-e86e-4b1a-bacd-9b8686961779/solver-workflow-redacted.png)

Solver offers also appear in community marketplaces. This advertisement offers a solver API for enterprise sites, with tests available. Attackers can shop for a service that fits the checks they face.

![Redacted enterprise solver advertisement](https://www.kairal.io/blog/_img/4170b689-e5f6-811c-841b-e575378114fe/3ca7735e-1980-446b-a5ee-fcbe15ad6d41/enterprise-solver-ad-redacted.png)

Our explanation of [how CAPTCHA solvers work](https://www.kairal.io/blog/how-captcha-solvers-work/) covers the market in more detail. For a business decision, the important distinction is between passing a check and being allowed to perform a particular action.

Passing a check should not give an account unlimited access to messages, credits or purchases.

#### They adapt when a component stops working

The menu below advertises support for several challenge products.

![Multi-product solver menu](https://www.kairal.io/blog/_img/4170b689-e5f6-811c-841b-e575378114fe/184ef029-6eb7-41af-b515-eea3b0b3a3a1/Screenshot_2026-11-13_at_13.14.58.png)

Retooling means changing the tools, accounts or tactics used after a defense interrupts an attack. An attacker might replace banned accounts, change a supplier or spread activity across more sessions.

Changing your challenge can therefore impose a cost without ending the activity. Check whether the same abusive sequence returns through different accounts after enforcement.

#### They use legitimate features for harmful purposes

The final action can look ordinary when examined alone.

A person can also perform these actions. A check that distinguishes a human from a script cannot establish whether that person's intended use is legitimate.

### Why traditional defenses miss coordinated abuse

CAPTCHAs can slow automated attempts, especially simple scripts. Solvers and human assistance limit what challenge completion proves. In its [credential-stuffing guidance](https://cheatsheetseries.owasp.org/cheatsheets/Credential_Stuffing_Prevention_Cheat_Sheet.html), OWASP recommends monitoring solve rates for both customer impact and signs of automated solving.

Multi-factor authentication helps protect accounts from stolen-password attacks. Email or phone verification confirms access to that contact method. Attackers can still use accounts they control to send scams, claim rewards or abuse free trials.

A web application firewall, or WAF, filters web requests and can stop many technical attacks. Abuse can use correctly formed requests to permitted features. It needs controls that understand the action and its business rules.

Rate limits constrain volume, but a limit applied separately to each account may miss a coordinated group. Making that limit stricter for everyone can also affect legitimate customers who share a network or perform frequent valid actions.

Static rules help with known patterns, such as a confirmed scam destination. Their value falls when attackers change the details. Keep investigating the behavior behind the rule.

The customer impact also extends beyond inconvenience. [W3C's December 2021 draft note on CAPTCHA accessibility](https://www.w3.org/TR/2021/DNOTE-turingtest-20211216/) explains how interactive challenges can exclude people with disabilities. An alternative path needs to be usable in practice, including on mobile devices and with assistive technology.

### How to stop coordinated abuse before friction costs you customers

#### Measure successful journeys and successful abuse

Track what happens after a check. A completed challenge followed by a fraudulent action is not a successful business outcome.

Check how many real customers finish what they came to do, how long it takes, and how often they face another challenge or ask for help. For confirmed attacks, track what damage happens before you stop them and whether they return through other accounts. Compare the same type of activity on both sides, such as checkout attempts.

Imagine a security check stops 800 out of 1,000 fraudulent checkout attempts. That sounds good: it blocks 80% of attacks. But if it also causes 2% of 100,000 genuine checkout attempts to be abandoned, that means 2,000 lost checkouts. In this example, more genuine checkouts are lost than attacks are stopped.

Compare the cost of lost customers with the fraud prevented and the cost of running the protection.

Use completed purchases, support cases and confirmed fraud reports to check your results. Compare similar customer groups over the same period.

#### Put checks around the action that creates harm

Review risk when someone attempts a sensitive action, even if registration happened yesterday.

For a marketplace, that might be revealing contact details or starting many conversations. For SaaS, it might be granting another allocation of free credits. For ticketing, it might be reserving inventory across related accounts.

Enforce the business rule on the server across web, mobile and API routes. Showing a verification screen is insufficient if another route still permits the protected action.

#### Connect activity across visits and accounts

Record the account, session, timestamp, action, decision and eventual outcome. These records help your team reconstruct what happened and identify repeated sequences.

Connect session and device data with account and payment records so your team can follow repeated activity across visits.

Look at several signals together, such as repeated actions, device links and account history.

When your team confirms abuse, carry those findings into later decisions. This reduces the chance that every replacement account starts a fresh investigation.

#### Make extra verification answer a specific question

Ask what each step establishes. An OTP can confirm access to a channel. Additional authentication can help establish that the account owner is present. A puzzle can add cost to automation.

Keep strong authentication for sensitive access. Use additional checks where the evidence and potential harm justify them, with clear explanations and a route to resolve mistakes.

Remove redundant steps through controlled testing, while monitoring confirmed abuse and customer completion. Include product and support teams: they see failures that never become security tickets.

#### Check whether attackers return

After a rule change, check whether attackers change their timing or spread activity across more accounts.

Look for renewed claims, reservations or outreach from related accounts. Measure how much harmful activity succeeds before detection, and how often legitimate restrictions are reversed.

### How Kairal can help

[Kairal Bot Protection](https://www.kairal.io/product/bot-protection/) helps stop automated login attempts, fake registrations, unwanted scraping and checkout abuse without asking customers to solve CAPTCHA puzzles. It combines signals from the browser, device, network and activity across visits and accounts to identify suspicious automation.

Your team can set protection separately for login, signup, checkout and other important pages. Automatic rules and invisible browser checks assess suspicious traffic, while the console shows related activity, the evidence behind a decision and the action taken. That gives your team context when investigating an attack or a wrongly blocked customer.

[Kairal Business Defense](https://www.kairal.io/product/business-defense/), currently in early access, adds research and testing around how your business can be abused. We review the agreed website and API flows to find where an attacker could collect data, create accounts, reserve stock or repeat payment attempts, and explain what that could cost your business.

We also follow attacker communities for changes in bypass tools and techniques. Through agreed, controlled tests, we check which attempts your defenses stop and where gaps remain. The findings guide protection for the affected actions. Preview lets your team review which requests would be blocked before enforcement begins, with evidence and prioritized changes to work through.
