Introducing the Airwallex Payment Fraud Engine: Stop fraud without killing conversion

As the digital economy expands, protecting online payments has become an even more complex and critical priority. Card‑not‑present transactions, subscription models, digital goods, and cross‑border eCommerce have made it easier for fraudsters to target online payments. At the same time, card scheme programs like Visa VAMP and Mastercard monitoring are tightening their thresholds, and customers expect a fast, low‑friction checkout every time.

Most legacy fraud tools force you into a painful trade‑off:

• Block aggressively and you may end up losing legitimate customers and revenue through false declines.

• Relax controls and you may absorb greater fraud losses, chargebacks, and scheme risk.

The Airwallex Payment Fraud Engine is built to avoid this trade‑off. It combines advanced AI and machine learning, real‑time rules, and deep integration across the payment stack to:

• Stop fraud and coordinated attacks in real time before they hit your P&L.

• Minimise false declines by triggering step-up authentication (e.g. one-time passcodes, banking app challenges) only when needed, so legitimate customers can get through.

• Protect your standing with card networks (like Visa and Mastercard) across markets and card ranges, and stay within their compliance programs.

Our fraud engine prevented 45% more fraud losses in 2025 than 2024, while materially improving payment success and reducing false declines for merchants.

A unified risk and revenue protection stack

The Payment Fraud Engine powers a complete risk and optimisation system across your payments. Instead of being a standalone fraud tool, it sits at the heart of a tightly integrated stack:

• Payment Fraud Engine (from checkout to any disputes): Screens and scores every transaction in real time and returns an allow/challenge/block decision while the payment is being processed. After the payment is processed, outcomes like issuer responses, address checks, security‑code results, and disputes are fed back into the engine to keep improving decisions over time.

• Optimize 360 (payment conversion optimisation): An AI‑powered engine that chooses the best payment path and decides when to retry, when to add extra security checks, and how to format and tokenize payment details so issuers are more likely to approve.

• Airwallex Sentinel (real‑time fraud attack defence): Detects coordinated attacks and abnormal patterns, then automatically deploys targeted strategies. Sentinel shows how much you save based on the fraud cases prevented, so you can see clear ROI during attacks and peak seasons.

• AI Dispute Automation (chargeback defence): An AI‑assisted dispute product that recommends whether to accept or challenge disputes, auto‑generates evidence packs, and automates dispute submissions. If fraud or first‑party misuse still gets through, this helps recover revenue with less manual work.

Because Airwallex controls acquiring, fraud, extra security checks, routing, tokenization, and disputes, all of these components work together:

• Fraud scores and decisions feed directly into routing and authentication.

• Attack detection informs real‑time rules and strategies.

• Dispute outcomes loop back into models to sharpen fraud detection and optimisation.

The result is a single, AI‑powered stack that treats fraud prevention, optimisation, and disputes as one connected problem, instead of separate silos.

Inside the Airwallex Payment Fraud Engine

When a customer tries to pay, the fraud engine decides whether that payment looks safe in real-time. Let’s take a look at what that looks like.

1. Data foundation and risk signals

High‑quality data is the foundation of the engine’s performance. Airwallex consolidates a rich set of signals into a standardised risk data layer to give the engine a clear, reliable view of each transaction. This includes:

• Core payment data and card details

• Device fingerprinting (unique device IDs) and on‑site behaviour signals

• Customer profile and behaviour signals

• Network and card-network data

High-quality data directly improves model accuracy, reduces false declines, and keeps your scheme metrics healthy.

2. Real‑time decisioning pipeline

When a payment reaches Airwallex’s fraud engine, it flows through a clear real-time pipeline designed to assess risk accurately without slowing down checkout:

a. Profile and model selection

Each transaction is mapped to one or more risk profiles and models, such as digital goods, online travel bookings, online‑to‑offline journeys, higher‑risk industry categories, or specific large merchants. This lets the engine tune decisions to the context of each payment, instead of treating everything the same.

b. Feature computation and model scoring

The engine computes thousands of features per transaction, including:

• Long‑window behavioural signals

• Real‑time velocity metrics

• Network and card‑network patterns

These features are combined, scored, then fed into the active fraud model, which creates an optimal risk decision for every payment in real time.

c. Rules and strategies

Scores and features then feed a layered rule engine designed to combine safety, precision, and scheme compliance. This includes:

• Safety‑net rules for card testing, bot traffic, and gross anomalies to stop known attack patterns.

• Segment‑specific rules tuned for different verticals and flows, such as online travel bookings, digital goods, subsequent subscription or recurring payments, and online‑to‑offline journeys.

• Rules aligned with Visa and Mastercard programs to keep fraud and dispute ratios within required thresholds across your card ranges and merchants.

• Merchant‑specific rules and watchlists to handle targeted attacks or bespoke risk appetite.

The engine evaluates all relevant rules and returns the most optimal decision to each payment.

d. Downstream orchestration

The fraud decision doesn’t live in isolation. It’s combined with Optimize 360 to decide:

• Whether to trigger additional security checks (like one‑time passcodes or banking‑app approvals), use exemptions, or keep the flow completely frictionless.

• Which acquiring path to use for the best chance of approval.

• How and when to retry failed payments (if at all).

This approach treats fraud as a first-class optimisation dimension alongside conversion and cost.

3. Merchant controls and transparency

Automation works best when you can see what’s happening, stepping in only when needed. At Airwallex, automation is paired with merchant‑facing controls and explainability:

• Risk Analytics: Dashboards showing fraud and dispute rates, block/verify rates, and payment success trends. These are often benchmarked against similar merchants and tied back to risk engine performance, so you can see what’s working and why.

• Risk Watchlists: You can configure whitelists and blacklists for attributes such as email, card fingerprint, IP address, and device ID. These controls are governed so they don’t undermine overall risk appetite or scheme obligations.

• Structured reason codes: Each decision comes with a simple, two‑level set of reason codes that turns complex internal logic into merchant‑friendly risk decision explanations. This makes it easier to understand declines, respond to questions, and fine-tune strategy over time.

At Airwallex, we’re constantly thinking of what’s next. This is just the beginning.

2025 Performance: What our fraud engine achieved

Across the card payment traffic controlled by our risk engine in 2025 (excluding third‑party or skip‑fraud flows), we saw:

• Higher success rates: Risk‑engine traffic success rates increased by 4.24%, driven by fraud engine and payment optimisation.

• Lower fraud rates: Fraud fell by over 60% on engine‑controlled traffic, even as volumes increased. This kept portfolios safely within scheme thresholds and away from VAMP/monitoring programs.

• Fewer false declines: False decline rates dropped by around half, driven by more precise, model-led decisions, delivering a better user experience and higher payment conversion rates.

• More fraud prevented: The total amount of fraud prevented increased by about 45% year‑on‑year compared to 2024, based on blocked fraud attempts that would otherwise likely have been authorised and later charged back.

These improvements are visible across major segments like digital goods, non‑digital, online travel, and new merchants, with targeted interventions where fraud or false declines spiked.

How Airwallex compares across the industry

Leading solutions like Stripe Radar, Adyen RevenueProtect, FraudNet, Forter, and Signifyd share some common traits:

• AI‑driven scoring for every transaction

• Use of global / network‑level data

• Real‑time decisioning with low latency

• Rule engines and allow/block lists

• Scheme monitoring and dashboards

The Airwallex Payment Fraud Engine offers all of the above, and adds three important differentiators:

1. Deep vertical integration with the payment stack

Airwallex isn't just another fraud plugin sitting on top of a processor. It runs the entire payment stack end to end, from acquiring and fraud checks to security checks and payment routing. That means our fraud engine can:

• Use extra security checks (like 3D Secure) smartly, skipping them, keeping them invisible, or triggering a full challenge (e.g. one-time passcode) when risk justifies it.

• Choose the best path for each payment to maximise approvals, where payments are routed and retried based on how different banks and card ranges behave.

• Send cleaner, richer payment data with secure tokens, making it easier for issuers to approve transactions while helping you keep processing costs under control.

The result is higher approval rates and lower fraud, without added friction or complexity.

2. Full lifecycle coverage from fraud to disputes

Airwallex covers the full payment lifecycle, from checkout to settlement to disputes:

• At checkout: The Payment Fraud Engine and Sentinel work in real time to stop high‑risk payments before they turn into fraud or chargebacks.

• After the payment is approved and captured: We use bank responses and transaction outcomes to keep tuning decisions, so you can reduce fraud without adding friction for legitimate customers.

• If a dispute or chargeback happens: AI Dispute Automation helps you decide which cases to fight, builds the evidence package, and submits it for you.

Together, this blocks fraud early and recovers more revenue later with less manual effort.

3. Proven performance vs dedicated vendors

Internal evaluations and live merchant comparisons have shown that:

• Earlier model generations already matched or outperformed dedicated vendors like Signifyd on key metrics in selected portfolios.

• In live scenarios where both Forter and Airwallex were present, the Airwallex Payment Fraud Engine showed comparable or stronger fraud capture, especially on flows (such as wallets) that weren’t fully integrated with the third‑party vendor.

• False decline rates have been held at around 1.4%, with a roadmap target to go below 1% in 2026, outperforming many reported industry benchmarks (often 2–3%).

With a 2025 fraud rate of around 0.3%, low false declines, and US$12.3 million in fraud prevented, the Airwallex Payment Fraud Engine is an industry-leading product that sits on a single, integrated platform for all your financial needs.

What this means for you

All of this is designed to show up as clear, measurable outcomes for your business:

1. Real‑time protection and attack resilience

While payments are being processed, our fraud engine can immediately block stolen cards, account takeover, card testing, and targeted attacks before they turn into fraud or chargebacks. Sentinel‑style monitoring shows exactly how much fraud and revenue loss has been prevented, especially during attacks and peak seasons.

2. Higher approval rates and better customer experience

Smart blending of models, rules, extra security checks, routing, and retries let more legitimate customers through with fewer unnecessary declines or authentications. This approach continues to evolve through our ongoing experimentation, including trust models, intelligent security checks, and profiles tuned for online‑to‑offline journeys.

3. Lower net fraud and dispute costs

Portfolio‑level fraud rates are kept safely within scheme thresholds. AI Dispute Automation and future chargeback capabilities recover more revenue and cut down manual dispute work.

4. Transparency, control, and partnership

Clear analytics, explainable reason codes, and configurable watchlists make it easy to understand decisions and adjust your strategy. This is supported by risk specialists, data analysts, and AI teams who actively tune strategies and share best practices.

5. Future‑proof, AI‑driven roadmap

Fraud never stands still, and our platform is built to stay ahead, with deep learning fraud models and AI agents built to adapt to new fraud tactics and issuer behaviour. With our AI capabilities, you can  continuously improve performance without constantly rewriting rules.

Bringing it all together

By unifying fraud prevention, optimisation, and disputes in a single AI‑powered stack, Airwallex helps you:

• Maximise revenue by lifting approval rates and reducing unnecessary friction.

• Minimise fraud loss and dispute costs through accurate checkout‑stage controls and AI‑assisted recovery.

• Protect brand and scheme standing with consistently low fraud and dispute rates.

• Deliver a smooth checkout experience that keeps customers coming back.

If you’re looking to upgrade from legacy fraud tools or consolidate multiple point solutions, the Airwallex Payment Fraud Engine offers a modern, integrated solution that’s designed to keep you ahead of both fraudsters and shifting issuer expectations, without forcing you to choose between safety and growth.

About Airwallex

Airwallex is a leading global financial platform for modern businesses. We empower businesses of all sizes to grow beyond borders and conquer new opportunities in a world of endless possibilities. With its proprietary next-generation infrastructure, Airwallex offers a suite of global financial solutions, including payments, treasury, software, and embedded finance.