PayPal uses a home-grown artificial intelligence engine to detect suspicious activity and, more importantly, to separate false alarms from true fraud, built with open-source tools. Fraud detection is one of the immediate paybacks of machine learning (ML) technology, because it addresses an urgent problem that would be impractical to solve if machine learning didn't exist.

PayPal is a pioneer in using ML techniques for risk management. PayPal uses three types of machine learning algorithms: linear, neural network, and deep learning. Experience has shown PayPal that in many cases the most effective approach is to use all three at once. PayPal uses multiple ML techniques, from linear predictions to deep learning because, according to the Data Science team at PayPal, although linear techniques might be outdated there may be some tasks at which the linear algorithms work better than the more complex deep learning techniques.

So applying all three at same time has significantly improved the accuracy of the fraud detection system at PayPal. They believe with increasing data, these techniques will be inevitable in the security measures in the future.

References: https://developer.paypal.com/docs/classic/fmf/integrationguide/FMFIntro/