← Back to Blog

BotShield for Fraud Prevention: How Proof-of-Work Stops Fraud Before It Starts

CAPTCHA asks "are you human?" That was never the right question. The right question is: did this request cost something? Proof-of-work answers that question for every request, every time, without tracking a single user.

CAPTCHA Was Never a Fraud Tool

CAPTCHA asks one question: are you human? That is a different question from is this fraudulent? A human can commit fraud. A bot can be legitimate. The distinction between human and bot is not the distinction between safe and dangerous.

A human with a stolen credit card passes reCAPTCHA every time. A human clicking through referral links for fake signups passes reCAPTCHA every time. A human farm in Southeast Asia solving CAPTCHAs at $2 per thousand passes reCAPTCHA every time. The system was designed to detect bots, not fraud. It does what it was designed to do. The problem is that we have been using it for something it was never designed to do.

The real question that stops fraud is: did this request cost something?

Proof-of-work answers that question by requiring every request to carry a computational cost. Not a behavioral signal. Not a cookie. Not a mouse movement pattern. A mathematical cost that is trivial for a single request and economically devastating at scale. BotShield is not just a CAPTCHA replacement. It is computational fraud prevention.

10 Fraud Vectors BotShield Prevents

1 Account Creation Fraud

Bots mass-create fake accounts to exploit free tiers, send spam, launder money, or build synthetic identities for downstream fraud. A single operator can create 10,000 accounts in minutes against an unprotected signup form.

BotShield makes each signup carry a real CPU cost. At difficulty 16, every account creation requires ~65,000 SHA-256 hashes. Creating 10,000 fake accounts means solving 650,000,000 hashes — real compute, real electricity, real time. The economics of mass account creation collapse.

Use cases: SaaS free tiers, SMS platforms like Textedly, referral programs, any signup flow

2 Credential Stuffing

Attackers test stolen username/password pairs from data breaches against your login page. They have millions of credentials. They need to test them fast. Rate limiting helps but punishes legitimate users behind shared IPs.

BotShield forces a proof-of-work challenge before each login attempt. Testing 100,000 credentials at difficulty 20 (auto-escalated after the first 50 attempts per minute) requires solving 100,000 challenges at ~1,000,000 hashes each. That is 100 billion SHA-256 operations. The credential stuffing attack becomes economically unviable before it finds a single valid pair.

Use cases: any login page, banks, e-commerce, enterprise portals

3 Card Testing / BIN Attacks

Fraudsters test stolen card numbers with small transactions to identify which cards are still active. They need speed. A typical card testing bot validates hundreds of cards per minute against a payment endpoint.

BotShield on payment forms makes rapid-fire card testing impossible. Each payment submission requires a valid proof-of-work token. Difficulty auto-scales at 50+ attempts per minute, escalating solve times from seconds to minutes. The testing window collapses before the fraudster can validate a meaningful batch.

Use cases: Stripe checkout pages, e-commerce payment forms, donation pages

4 SMS Toll Fraud (SMS Pumping)

Bots trigger OTP sends to premium-rate numbers. Every SMS you send costs you money and generates revenue for the attacker. A single bot can trigger thousands of OTP sends per hour, costing you hundreds or thousands of dollars in Twilio or SNS fees.

BotShield before the "send code" endpoint means each OTP request costs computation first. The bot must solve a proof-of-work challenge before your system sends a single SMS. At scale, the compute cost exceeds the toll fraud revenue. The attack becomes unprofitable.

Use cases: OTP flows, Twilio customers, any SMS verification endpoint

5 Promo / Coupon Abuse

Bots exploit signup bonuses, referral codes, and promotional credits by creating thousands of fake accounts. Each account redeems the promotion. The cost to the company scales linearly with the number of fake accounts.

BotShield applies a double tax on abuse: proof-of-work per account creation plus proof-of-work per redemption. A legitimate user solves two challenges and never notices. A bot farm creating 5,000 accounts to redeem $10 bonuses must solve 10,000 challenges. The compute cost makes the $50,000 in fraudulent redemptions economically irrational.

Use cases: DoorDash, Uber, any referral program, signup bonus flows

6 Ticket / Inventory Scalping

Bots buy limited-availability items faster than humans can click. Concert tickets, sneaker drops, GPU launches. The bot completes checkout in milliseconds while humans are still loading the page.

BotShield's 1-3 second solve time is invisible for a single purchase. But a bot attempting 500 simultaneous purchases must solve 500 challenges in parallel. That requires 500 CPU threads running at full capacity. The bot's speed advantage evaporates because each checkout carries a non-negotiable computational delay.

Use cases: ticket sales, sneaker drops, limited-edition releases, GPU launches

7 Scraping / Price Intelligence

Competitors scrape your pricing, product catalog, and inventory data. They hit every page, every hour, building a mirror of your business intelligence. Rate limiting blocks them temporarily. They rotate IPs and continue.

BotShield makes large-scale scraping economically impractical. Each page load requires a valid proof-of-work token. Scraping 100,000 pages means solving 100,000 challenges. IP rotation does not help because the cost is computational, not network-based. The scraper needs CPUs, not proxies.

Use cases: airlines, hotels, real estate listings, e-commerce catalogs

8 Review / Rating Manipulation

Bot farms post fake reviews to inflate or destroy product ratings. They operate at scale across multiple accounts, each posting reviews that look organic individually but form a coordinated campaign collectively.

BotShield on review submission endpoints forces proof-of-work per review. A legitimate customer submitting one review solves one challenge invisibly. A bot farm posting 1,000 fake reviews solves 1,000 challenges. The time cost alone makes coordinated manipulation campaigns impractical within the attack windows that fake review operations depend on.

Use cases: Amazon, Yelp, Google Reviews, app stores, any review platform

9 Ad Fraud / Click Fraud

Fake clicks drain ad budgets. Bots click your PPC ads, you pay per click, the attacker (often a competitor or a fraud ring) profits from wasted spend. Click fraud costs advertisers an estimated $100 billion annually.

BotShield on landing pages verifies real computation before counting a visit. The bot must solve a proof-of-work challenge on the landing page before the analytics pixel fires. Fake clicks that do not render the page and solve the challenge are not counted. Your ad budget goes to real visitors.

Use cases: any PPC advertiser, affiliate networks, display ad campaigns

10 API Abuse / Resource Exhaustion

Bots hammer public APIs to scrape data, abuse free tiers, or perform denial-of-service attacks. Rate limiting by API key is trivially bypassed with key rotation. Rate limiting by IP is bypassed with proxy rotation.

BotShield's createProtectedFetch() wrapper means every API call carries a valid proof-of-work token. The cost is per-request, not per-IP or per-key. A legitimate developer making 100 API calls per day solves 100 challenges silently. A bot making 100,000 calls per day solves 100,000 challenges. The compute cost is the rate limit.

Use cases: any public API, AI/ML endpoints, data feeds, webhook receivers

Why This Is Different from reCAPTCHA

reCAPTCHA answers "is this a human?" by analyzing behavior: mouse movements, browsing patterns, cookies, device fingerprints. BotShield answers "did this request cost something?" by requiring computation. The difference is fundamental.

reCAPTCHA BotShield
Question asked Are you human? Did this cost something?
Human farms Pass every time Still costs compute per request
Data collected Behavioral fingerprints, cookies, IP Zero. No personal data.
GDPR liability Requires consent banner + DPA None. No data to regulate.
Defense model Binary (human/bot) Economic (cheap at 1, devastating at 10K)
Bypass method CAPTCHA farms ($2/1000 solves) No bypass. Math is math.
Accessibility Hostile to screen readers Fully accessible. No UI.

reCAPTCHA can be fooled by human farms. BotShield cannot. The computation is mathematical, not behavioral. You cannot outsource a SHA-256 hash to a cheaper human. The hash costs the same regardless of who initiates it.

reCAPTCHA collects behavioral data that creates GDPR liability. BotShield collects zero data. The challenge is a random nonce. The solution is a number. The session token is a hash. There is no personal data at any point in the protocol.

reCAPTCHA is binary: human or bot. BotShield is economic: cheap for one request, expensive at scale. This is a stronger defensive model because it addresses the attacker's economics directly rather than attempting to classify their identity.

Post-Quantum Fraud Evidence

Every BotShield challenge is signed with CRYSTALS-Dilithium (ML-DSA, NIST FIPS 204) -- a post-quantum digital signature algorithm. The session token is not just an access credential. It is cryptographic proof that a specific device performed real computation at a specific time.

Each BotShield token is admissible evidence: "This signup was performed by a device that solved a SHA-256 challenge at difficulty 16, verified by a Dilithium signature, at 2026-03-21T14:32:07Z." No other CAPTCHA alternative produces cryptographic proof of device computation.

This matters for three specific legal scenarios:

The Economics of Computational Fraud

The fundamental asymmetry of proof-of-work is what makes it unbeatable as a fraud prevention mechanism. The attacker's cost grows exponentially. The defender's cost stays constant.

Actor Difficulty Hashes per Challenge Time per Challenge
Normal user 16 bits ~65,000 1-3 seconds (invisible)
Bot at 50 req/min 20 bits ~1,000,000 15-45 seconds each
Bot farm at 200+ req/min 24 bits ~16,000,000 Minutes each
Defender (verification) any 1 Microseconds (always)

A normal user solves one challenge at difficulty 16. It takes 2 seconds and happens in the background. They never see it. A bot attempting 50 requests per minute gets auto-escalated to difficulty 20. Each challenge now requires a million hashes and takes 15-45 seconds. A bot farm at 200+ requests per minute faces difficulty 24: 16 million hashes per challenge, minutes per solve.

Meanwhile, the defender verifies every solution with exactly one SHA-256 hash. One hash. Microseconds. Regardless of the difficulty level.

Attacker cost: O(2d) — exponential in difficulty
Defender cost: O(1) — constant, always

This is the fundamental asymmetry that makes
proof-of-work unbeatable at scale.

Every 4-bit increase in difficulty makes the attack 16 times more expensive while the defense cost does not change. There is no other fraud prevention mechanism with this property. Rate limiting is linear. CAPTCHAs are binary. Behavioral analysis is probabilistic. Proof-of-work is exponential.

Pricing

BotShield is free to start and scales with your traffic.

Free
$0
2,500 challenges/mo
Pro
$49/mo
100K challenges/mo
Business
$249/mo
1M challenges/mo
Enterprise
$1,499/mo
25M challenges/mo
Scale
Custom

Every tier includes adaptive difficulty, Dilithium-signed challenges, session tokens, and server-side verification. No feature gating. The only variable is volume. Full pricing details.

Stop Fraud Before It Starts

One script tag. Zero tracking. Exponential cost to attackers. Constant cost to you.

Get Started with BotShield →

← Back to Blog   ·   BotShield Product Page →   ·   GDPR Bot Protection →