Fraud detection accuracy in Stripe with AI
Stripe, one of the world’s leading online payment platforms, has remarkably increased its fraud detection accuracy from 59% to 97% overnight by leveraging artificial intelligence. This achievement is a prominent example of the positive impact of AI applications in fintech and payment security.
At the recent Stripe Sessions event, the company unveiled its new model called the Payments Foundation Model. It is claimed to be the world’s first foundational model in the payment field, designed to improve fraud detection accuracy and enhance security in financial transactions.
Fraud detection accuracy in Stripe with AI
Solving the credit card test challenge with the new model
One common threat in the payment world is credit card testing. In these attacks, attackers use stolen card information to verify their validity for further misuse. Stripe’s new model, focusing on this type of attack, has achieved significantly better performance compared to traditional methods.
Creative use of transformers in payment modeling
Gautam Kedia, Machine Learning Manager at Stripe, explained on LinkedIn that although the company’s previous models were somewhat effective, they required separate design and training for each task. He added, "Inspired by the architecture of large language models (LLMs), we decided to use transformers to build a comprehensive model. Although payment structures are completely different from language, we found that this model can understand hidden relationships between transaction data similar to how words relate in sentences."
This model represents each payment as a dense, multidimensional vector using data from billions of transactions. As a result, similar transactions cluster together in a specific vector space, enabling Stripe to detect complex patterns that were previously overlooked.
Fraud detection accuracy in Stripe with AI
Leap in accuracy and performance in fraud detection
Stripe announced that by using a vector-based classification model derived from their foundational model, they have been able to detect and stop card testing attacks before they fully occur. This action led to an unprecedented increase in fraud detection rates for these attacks, rising from 59% to 97%.
Kedia believes that payment transactions, like words, have semantic dependencies that can only be understood through vector analysis—a capability that manual feature engineering could not provide.
Fraud detection accuracy in Stripe with AI
Recovering billions of dollars from mistakenly rejected transactions.
Following its successes, Stripe has leveraged AI to recover over $6 billion in transactions mistakenly declined by issuing banks. The smart tool, Adaptive Acceptance, has played a key role by detecting patterns of erroneous transaction declines.
The model used in this tool was initially XGBoost, but Stripe replaced it with an enhanced version of tabular transformers called TabTransformer+ to improve performance. This new model increased the accuracy of identifying mistakenly declined transactions by up to 70% and reduced repeated attempts to complete transactions by 35%.
Significant drop in fraud detection accuracy at Stripe using AI
Additionally, Stripe's fraud prevention system, called Radar, has been updated with new features such as automated authentication. This tool can automatically enable two-factor authentication for suspicious transactions. The initial results of implementing this system showed a 30% reduction in fraud rates for reviewed transactions.
Future Outlook: Artificial Intelligence at the Heart of the Payment Industry
The accuracy of fraud detection at Stripe using artificial intelligence shows that the future of the payment industry is unimaginable without AI. Companies like Razorpay are following a similar path, leveraging AI to optimize operations, reduce payment delays, and manage returns.
Ultimately, Stripe’s efforts prove that artificial intelligence can not only enhance security but also transform the customer payment experience. This path can serve as an inspiring model for other fintech companies.
Fraud detection accuracy in Stripe with AI
Source: hooshio
Click the link below to read more news related to artificial intelligence.
Kish iVision Blog Page
