As estimativas colocam as perdas de comércio eletrônico com fraudes de pagamento online em US$ 41 bilhões globalmente em 2022, com expectativa de crescer para US$ 48 bilhões em 2023. Além disso, relatórios do Consumer Sentinel da FTC sugerem que a Amazon é a favorita dos golpistas. Entre julho de 2020 e junho de 2021, cerca de uma em cada três pessoas que denunciaram um imitador de negócios disse que o golpista alegou ser a Amazon. Cerca de 96.000 pessoas relataram ter sido atacadas e quase 6.000 disseram ter perdido dinheiro. É claro que você deve se manter atento ao operar ou fazer compras em empresas de comércio eletrônico, mas as notícias de um serviço de detecção de fraude da Amazon dão esperança às empresas que usam a AWS.
A Amazon está extremamente ciente dessa prevalência de fraude online, então eles agora implementaram um serviço gerenciado de detecção de fraude. Eles implantam IA em seu Amazon Fraud Detector, um modelo de aprendizado de máquina de detecção de fraude personalizado para seus dados em um processo totalmente automatizado.
Guia de detecção de fraude da Amazon (+ 3 verificadores gratuitos ou baratos):
Tipos típicos de fraude no comércio eletrônico
Todo proprietário de loja de comércio eletrônico precisa se proteger contra fraudes. Mas o que queremos dizer com fraude no comércio eletrônico? A fraude de comércio eletrônico pode ser abrangente e inclui todos os tipos de ações maliciosas projetadas para explorar sua loja online. Exemplos comuns de fraude em comércio eletrônico variam desde o uso de números de cartão de crédito roubados ou hackeados até fraudes de devolução e tentativas de aquisição de conta.
Algumas das tentativas mais comuns de fraude de comércio eletrônico na Amazon (algumas visando compradores, outras vendedores) incluem:
While Amazon Fraud Detector targets stores that use Amazon’s AWS cloud platform, preventing fraud detection is equally important for sellers on the main Amazon site. And be aware that some of your fellow Amazon sellers are indeed fraudsters themselves, which can become problematic when trying to police your brand from illegal sellers.
Sometimes you will come across Amazon sellers offering counterfeit versions of your goods. If you come across these ads, make a couple of test purchases. If they aren’t genuine, you can make a claim with Amazon against the seller.
If you have a distinct brand, it would be worthwhile signing up for Amazon’s Brand Registry. This gives automatic protections and can proactively remove suspected infringing or inaccurate content. Amazon claims to have prevented more than 2.5 million attempts to create new selling accounts from “bad actors” before they listed a single product for sale.
Amazon’s Fraud Detector Service
Amazon Fraud Detector automates the detection of potentially fraudulent activities online. These include unauthorized transactions and the creation of fake accounts. It uses machine learning to analyze your data, in a way that utilizes the experience gained by more than 20 years of fraud detection at Amazon.
Amazon has created its fraud detection service for its AWS customers to identify potentially fraudulent activities and catch more online fraud faster.
Amazon makes its Fraud Detector available to eCommerce businesses that use Amazon S3 and the Amazon Fraud Detector API. This means that it has a broader use case than just reducing fraud in your Amazon store. Any eCommerce store using Amazon’s cloud computing platform has access to Amazon’s fraud detector service.
Benefits of Amazon Fraud Detection
Amazon’s fraud detection uses fully automated machine learning models customized to meet your specific business needs. You can use Amazon Fraud Detector models to identify potential fraud in many types of online transactions, such as new account creations, online payments, and guest checkout.
As it uses an automated model, you don’t have to worry about creating and training a model. Amazon Fraud Detector automatically finds the most suitable fraud detection algorithm for your use case and creates the model.
Amazon Fraud Detector validates your model’s performance and offers visual tools for you to assess the results. It generates fraud predictions for your organization’s business activities.
Typical use cases of Amazon Fraud Detector include:
- New account fraud, within an account sign-up process
- Online identity fraud
- Payment fraud for online orders
- Guest checkout fraud
- Loyalty account protection
- Account takeover detection
- Seller fraud in online marketplaces
How Does Amazon Fraud Detector Work?
Amazon Fraud Detector uses machine learning models trained with historical fraud data that you provide. Amazon considers a model to be an output of machine learning algorithms implemented in code and run on event data you provide. Amazon has developed various model types, i.e. specialized recipes used to build fraud detection models for specific fraud cases. So, in this case, Amazon’s AI will examine the data you give it and then use what it believes to be the best fraud detection model to train a customized model for your circumstances.
You first need to define an event you want to assess for fraud. Once you know your focus, you upload historical data to Amazon Simple Storage Service (Amazon S3) and select a fraud detection model type. The service automatically trains, tests, and deploys a customized fraud detection model based on your unique information.
When ready, Amazon Fraud Detector imports your model to detectors. You can integrate the Amazon Fraud Detector API into your website’s transactional functions, such as account sign-up or order checkout. You can use these to configure decision logic, such as rules, to interpret the model’s score and assign outcomes. Possible outcomes include approving a transaction, reviewing the transaction, or sending the transaction for future investigation.
Whenever someone initiates a transaction through your online store, Amazon Fraud Detector runs its model (based on your customized model) and determines the most suitable outcome for a transaction, based on your experience.
Workflows for Detecting Fraud with Amazon Fraud Detector
Let’s look at how Amazon Fraud Detector works in practice.
Fraud detection is a continuous process. So, although the above graph starts with you setting up your version of Amazon Fraud Detector from scratch, you will repeat much of this process every time somebody tries to transact with you. This is because your model will be continuously learning and refining each time it runs.
You will want to regularly evaluate its performance scores and metrics based on the prediction explanations. This helps develop genuine risk indicators and limit false positives. You can continually add updated data to your model. As more data becomes available, continue retraining your model to increase accuracy.
Accessing Amazon Fraud Detector
Currently, Amazon Fraud Detector only has limited availability. You can use it in the US East (N. Virginia), US East (Ohio), US West (Oregon), Europe (Ireland), Asia Pacific (Singapore), and Asia Pacific (Sydney) AWS Regions. If you live somewhere else you will have to look at one of the other Amazon fraud detection programs (also suitable for other eCommerce platforms) we cover later in this post.
How to Use Amazon Fraud Detector
To be able to use Amazon Fraud Detector you first need an Amazon Web Services (AWS) account (and presumably some form of eCommerce store using AWS). You must set up the permissions that give your AWS account access to all the interfaces, granting Amazon Fraud Detector access to your account to perform tasks on your behalf and to access resources that you own.
Amazon offers a full tutorial to help you learn how you can use Amazon Fraud Detector to build, train, and deploy a fraud detection model.
Creating an Event Dataset
An important stage of setting up Amazon Fraud Detector is creating an event dataset, using the historical fraud data for your company. This means that the fraud detector model is of little use to a new eCommerce business – you won’t have had a chance to build up any data. However, in that case, Amazon will rely on its 20 years’ worth of data in the model, giving new businesses at least a reasonable starting point.
Your dataset is used to create a model to predict fraudulent events. You will need to decide on a goal for creating a particular fraud detection model, and then be clear on how you intend to use the model. From that, you need to decide on the metrics you will use to measure the performance of the model.
Amazon provides an example of a model that evaluates potential account registration fraud. You might set a goal to “auto-approve legitimate registrations, but capture fraudulent registrations for later investigation.” Once you know that, you might decide that the purpose of your model is for real-time fraud detection for each account registration. Potential metrics you could use to determine the effectiveness of your model include capturing X% fraud registrations with a Y% false positives rate and accepting no more than Z% of auto-approved registrations that are fraudulent.
Other eCommerce Fraud Checkers
Amazon Fraud Detector isn’t your only hope to avoid becoming a victim of eCommerce fraud, however. There are several other options for eCommerce retailers, including some operating through Amazon’s marketplace.
Note that some of these fraud checkers are free to install (and use with limited functionality), but you will have to upgrade to a paid plan to gain maximum benefit.
1. SEON Fraud Prevention
SEON offers a fraud detection service you can use to reduce chargebacks and resources lost to fraudulent orders. They give you the tools necessary to investigate false refunds and chargeback requests and to help you meet your card network’s chargeback ratios.
You can detect fake accounts through a complete digital footprint analysis. You’ll be able to flag customers using stolen or fake IDs, ensuring that only real shoppers benefit from your promo codes. Minimize labor-intensive manual reviews with a user-friendly Dashboard for risky transactions. They combine the power of email, phone, IP, device, and behavior analysis with machine learning to help you detect fraudulent orders in real time.
2. Amazon Payment Services: Protect and Protect Plus
Amazon Payment Services offers standard and advanced fraud protection services.
They call their standard service Protect. Amazon recommends you enable this if your eCommerce business is at low risk of fraud, or if your eCommerce technology already includes comprehensive anti-fraud technology. You apply simple rules to guard against fraudulent transactions.
For maximum anti-fraud capabilities, you could consider activating Amazon’s advanced fraud protection service, called Protect Plus. It isn’t free, but it could save you considerable money, making it a very good investment.
Protect Plus uses intelligent, real-time analysis that uses machine learning models to detect fraudulent activity. It works alongside Amazon’s standard fraud management service. Protect Plus provides significantly more capable anti-fraud measures compared to Protect, Amazon Payment Services’ standard, rules-based anti-fraud solution.
3. Red Points’ Amazon Brand Protection Services
You can use Red Points’ Amazon Brand Protection solution to identify trademark, patent, and brand infringement on Amazon 24/7. Scan Amazon for potential online infringements every hour. Image recognition technology scans Amazon listings’ photos to identify trademark, patent, and brand infringement.
It helps you automatically enforce against infringing sellers using your trademark, patent, or copyright on Amazon. Alternatively, you can review detections and decide on cases yourself. Use image cards to validate, pause or discard detected incidents on Amazon.
With the help of machine learning tools, such as keyword generation and image recognition, Red Points’ Amazon brand protection services can process the information retrieved from Amazon to detect fakes, piracy, or sellers online and allow their customers to take enforcement actions.
Wrapping Things Up
If you use AWS as cloud storage for your eCommerce website then you could benefit from using Amazon Fraud Detector, to minimize the chances of you becoming a victim of fraudulent activity and scams. We have suggested a few apps (or Amazon payment settings) you could consider using to assist with this.
If you are a seller on the Amazon Marketplace, however, things are not so simple. To a large extent, you will need to be wary and protect yourself against any suspicious activity. Areas you should always on be the lookout for include:
- Phishing – i.e., someone (including Amazon) trying to acquire personal details from you. Consider using email authentication protocols such as DKIM and SPF.
- Buyers claiming “Failed Delivery” for legitimate deliveries – use track-and-trace postage when sending expensive items
- Golpe de “Substituir e Reembolsar” – os compradores substituem seu novo produto por um antigo ou danificado e, em seguida, enviam-no de volta para um reembolso. É uma boa ideia colocar adesivos invioláveis em seus produtos, para que você saiba se eles foram abertos.
- Golpes de e-mail – só porque você recebeu um e-mail mostrando que fez uma venda, não significa necessariamente que você tenha feito. Verifique novamente sua conta do Seller Central antes de despachar mercadorias.