Nethone Guard - Know Your Users to resolve fraud

Nethone Guard is our proprietary machine-learning based product designed to protect the entire user journey, from onboarding to post-payment, helping ecommerce merchants and financial institutions to holistically understand their end-users — also referred to as Know Your Users™ (KYU). We address various challenges, including account takeover, chargebacks, friendly fraud, promo & policy abuse, and CNP fraud. We extract exhaustive user profiling data - unique hardware, software, device, network, and behavioral data - from any device, while the user interacts with the browser or the mobile app of our clients via a javascript code embedded inside of the website or via an SDK embedded in the mobile app respectively. In order to extract more hidden knowledge that can reveal the true intentions of a particular user, we combine 5,000+ in-depth data attributes and enhance it with external data sources, all via a single API.

Our machine learning-based decision engine finds patterns that point out to fraud and then reveals this to our client as a recommendation. We turn raw data into special features the models can use to come up with a recommendation or an assessment of fraud risk. The decision engine delivers three recommendations for our client to decide: accept, review, or refuse. The whole process takes place in real time, passively in the background, with no harm to UX.

Depending on the tiered offer, businesses dealing with large volumes of transactions usually have a data scientist assigned to collect the business insights from our customers and transform them into a logic in our system to deliver the best performance.

By developing a hybrid rule-based, machine learning powered anti-fraud solution, we provide fraud detection and prevention, blocking all risky users without friction to the good ones. We bring out to the industry a modularized approach that offers end-to-end protection at every step of the user journey: user acquisition - login and registration - transaction attempt - post-payment dispute. For each stage, we have specific challenges that we address and tailor-made solutions as follows: Nethone KYC, Nethone PSD2, and Nethone Alerts. In addition, our competitive edge comes from the following reasons: We use advanced monitoring and blocking techniques to protect websites and mobile apps from over 100 types of fraud risks. With our access to first-hand Darknet intelligence, we can proactively warn our clients if they are on fraudsters’ radar and if any fraud is coming their way. We have a fully customizable and modular interface, so according to our clients’ budgets and plans, they pay only for what suits their needs. Our decision engine is fully transparent, explaining the logic behind the recommendations, and our clients can also control the ML models and adapt the rules to fit their requirements.

1. Machine learning models and data hub enhancements

• We have built a model dedicated to Brazilian clients expected to catch 20-25% of fraudulent transactions with a 1% rejection rate.

• To improve the quality of the recommendations our clients receive, we enabled them to send their historical data to us using our API. Doing so allows us to draw insights from our clients’ particular pool of customers and combine it with our own intelligence, providing them with high-precision advice.

• We have expanded and strengthened our anti-fraud capabilities by incorporating additional information about email addresses to more effectively identify and prevent fraudulent activities.

2. Nethone KYC - risk-based approach

• Our profiling solution, combined with additional data from the customer (such as information provided in a registration form or details about the transaction), can be used to assess the risk level of a particular user before requiring them to undergo KYC verification.

• Depending on the risk level, the user will be directed to a specific KYC process, which can help to save money on KYC checks and improve the user experience during registration.

3. Nethone PSD2

• We come with the opportunity of bypassing any friction that may come with PSD2 SCA, including soft or hard declines while keeping fraud prevention at the same high standards for your business and consumers.

• We perform real-time checks addressed by rules and powered by our ability to identify device and network anomalies, as well as risk analysis of the risk-based factors, which can be performed either by using rules or custom machine learning models according to our customers’ tier offering or traffic volume.

• It’s important to note that this new product has a wide spread scope, as it can be used to identify transactions with the lowest risk in geographies where PSD2 is not in effect.

Case study Azul:

The problem Azul Linhas Aéreas is a Brazilian airline with the largest air network in Brazil, serving more than 100 national destinations, in addition to operating selected international routes to the United States and Portugal. The airline sought to enhance its existing anti-fraud capabilities, seeking an option that would work effectively and automatically, cutting operational costs by reducing time-consuming manual reviews. Additionally, they wanted an advanced solution that would identify and prevent all types of fraud affecting their business. There was a strong emphasis on providing a frictionless customer experience, from start to finish of the customer journey.

The solution Upon successful implementation of Nethone’s advanced fraud detection and prevention solution with the help of our customer-centric approach, the customer quickly discovered their payments process was much safer than before. They were able to identify new types of fraud affecting their business - something that was not possible with their previous rules-based setup.

The customer was impressed with the capabilities of behavioral biometrics, and digital fingerprinting, backed up by machine learning (ML) models to distinguish between good and bad customers and fraud actors. During a specific fraud attack episode, Nethone’s ML models and signals (triggers that indicate suspicious behavior indicative of a high probability of fraud) detected 89% of fraudulent transactions, compared to 16% detected by the rules-based system, through discrepancies between user account purchase history and use of multiple email accounts. Such detailed analysis is only possible with an advanced fraud solution.

The benefits of the customer’s improved anti-fraud setup have allowed the company to eliminate unnecessary revenue losses through a reduction in chargeback rates and false positives. Operational costs are down thanks to significantly reducing time-consuming manual reviews.

Results: 95.3% blocked account takeovers (ATO) fraud with passive behavioral biometrics 66% drop in chargeback rate for a travel company entering a new market +10p.p. higher fraud detection precision in a digital banking mobile app 60% drop in manual review rate for a luxury eCommerce player

Testimonial: “When we started the cooperation with Nethone, we aimed to improve the automatic authentication of good customers, improving our customers’ experience and consequently lowering the cost of anti-fraud with manual analysis. In a short time, we were able to obtain encouraging results.” - Felipe Maia, Coordenador Prevenção a Fraude, Azul

https://nethone.com/