<?xml version="1.0" encoding="UTF-8" standalone="yes"?><oembed><version><![CDATA[1.0]]></version><provider_name><![CDATA[Gigaom]]></provider_name><provider_url><![CDATA[http://gigaom.com]]></provider_url><author_name><![CDATA[Derrick Harris]]></author_name><author_url><![CDATA[http://search.gigaom.com/author/dharrisstructure/]]></author_url><title><![CDATA[How PayPal uses deep learning and detective work to fight fraud]]></title><type><![CDATA[link]]></type><html><![CDATA[<p>Hui Wang has seen the nature of online fraud change a lot in the 11 years she&#8217;s been at PayPal. In fact, <a href="http://www.darkreading.com/vulnerabilities---threats/how-cybercriminals-choose-their-targets-and-tactics/d/d-id/1138970?">a continuous evolution of methods</a> is kind of the nature of cybercrime. As the good guys catch onto one approach, the bad guys try to avoid detection by using another.</p>
<p>Today, said Wang, PayPal&#8217;s senior director of global risk sciences, &#8220;The fraudsters we’re interacting with are&#8230; very unique and very innovative. &#8230;Our fraud problem is a lot more complex than anyone can think of.&#8221;</p>
<p>In deep learning, though, Wang and her team might have found a way to help level the playing field between PayPal and criminals who want exploit the online payment platform.</p>
<p>Deep learning <a href="https://gigaom.com/2015/01/29/new-to-deep-learning-here-are-4-easy-lessons-from-google/">is a somewhat new approach to machine learning and artificial intelligence</a> that has caught fire over the past few years thanks to companies such as <a href="https://search.gigaom.com/company/google/">Google</a>, <a href="https://search.gigaom.com/company/facebook/">Facebook</a>, <a href="https://search.gigaom.com/company/microsoft/">Microsoft</a> and Baidu, and a handful of prominent researchers (some of whom now work for those companies). The field <a href="https://gigaom.com/2014/10/29/it-doesnt-matter-if-deep-learning-mimics-the-brain-or-watson-is-cognitive-it-matters-if-they-work/">draws a lot of comparisons</a> to the workings of the human brain because deep learning systems use artificial neural network algorithms, although <a href="https://gigaom.com/2015/02/14/why-deep-learning-is-at-least-inspired-by-biology-if-not-the-brain/">&#8220;inspired by the brain&#8221; might be a more accurate description</a> than &#8220;modeled after the brain.&#8221;</p>
<div id="attachment_825826" class="wp-caption aligncenter" style="width: 804px"><img  src="https://gigaom2.files.wordpress.com/2014/03/calista-deepface.png?w=804&#038;h=275" alt="How DeepFace sees Calista Flockhart. Source: Facebook" width="804" height="275" class=" aligncenter" /><p class="wp-caption-text">A visual diagram of a deep neural network for facial recognition. Source: Facebook</p></div>
<p>Essentially, the stacks of neural networks that comprise deep learning models are very good at recognizing patterns and features of the data they&#8217;re trained on, which has led to some huge advances in <a href="https://gigaom.com/2015/02/13/microsoft-says-its-new-computer-vision-system-can-outperform-humans/">computer vision</a>, <a href="https://gigaom.com/2014/12/18/baidu-claims-deep-learning-breakthrough-with-deep-speech/">speech recognition</a>, <a href="https://gigaom.com/2013/09/26/how-deep-learning-can-teach-computers-spanish-without-a-tutor/">text analysis</a>, <a href="https://gigaom.com/2014/08/05/how-spotify-is-working-on-deep-learning-to-improve-playlists/">machine listening</a> and <a href="http://googleresearch.blogspot.com/2015/02/from-pixels-to-actions-human-level.html">even video-game playing</a> in the past few years. You can learn more about the field at our <a href="https://events.gigaom.com/structuredata-2015/">Structure Data</a> conference later this month, which includes deep learning and artificial intelligence experts from Facebook, Microsoft, Yahoo, Enlitic and other companies.</p>
<p>It turns out deep learning models are also good at identifying the complex patterns and characteristics of cybercrime and online fraud. Machine-learning-based pattern recognition has long been a major part of fraud detection practices, but Wang said PayPal has seen a &#8220;major leap forward&#8221; in its abilities since it began investigating precursor (what she calls &#8220;non-linear&#8221;) techniques to deep learning several years ago. PayPal has been working with deep learning itself for the past two or three years, she said.</p>
<p>Some of these efforts are already running in production as part of the company&#8217;s anti-fraud systems, often in conjunction with human experts in what Wang describes as a &#8220;detective-like methodology.&#8221; The deep learning algorithms are able to analyze potentially tens of thousands of latent features (time signals, actors and geographic location are some easy examples) that might make up a particular type of fraud, and are even able to detect &#8220;sub modus operandi,&#8221; or different variants of the same scheme, she said.</p>
<div id="attachment_919466" class="wp-caption aligncenter" style="width: 530px"><img  src="https://gigaom2.files.wordpress.com/2015/03/editfilters1.gif?w=530&#038;h=401" alt="Some of PayPal's fraud-management options for developers." width="530" height="401" data-attribution="PayPal" class=" aligncenter" /><p class="wp-caption-text">Some of PayPal&#8217;s fraud-management options for developers.</p></div>
<p>The patterns are much more complex than &#8220;If someone does X, then the result is Y,&#8221; so it takes artificial intelligence to analyze them at a level much deeper than humans can. &#8220;Actually,&#8221; Wang said, &#8220;that&#8217;s the beauty of deep learning.&#8221;</p>
<p>Once the models detect possible fraud, human &#8220;detectives&#8221; can get to work assessing what&#8217;s real, what&#8217;s not and what to do next.</p>
<p>PayPal uses a champions-and-challengers approach to deciding which fraud-detection models to rely on most heavily, and deep learning is very close to becoming the champion. &#8220;We’ve seen roughly a 10 percent delta on top of today’s champion,&#8221; Wang said, which is very significant.</p>
<p>And as the fraudulent behavior on PayPal&#8217;s platform continues to grow more complex, she&#8217;s hopeful deep learning will give her team the ability to adapt to these new patterns faster than before. It&#8217;s possible, for example, that PayPal might some day be able to deploy models that take live data from its system and become smarter, by retraining themselves, in real time.</p>
<p>&#8220;We’re doing that to a certain degree,&#8221; Wang said, &#8220;but I think there’s still more to be done.&#8221;</p>
]]></html><thumbnail_url><![CDATA[https://i2.wp.com/gigaom2.files.wordpress.com/2015/03/shutterstock_76767721.jpg?fit=440%2C330&quality=80&strip=all]]></thumbnail_url><thumbnail_height><![CDATA[293]]></thumbnail_height><thumbnail_width><![CDATA[440]]></thumbnail_width></oembed>