 Financial fraud remains a continuous source of loss around the world. We're talking loss of trust, loss of customers, and loss of reputation. In fact, the Nielsen report estimates that credit card losses topped $24.71 billion in 2016. Although there are plenty of good fraud solutions out there, these centralized solutions have limited information on the users and can't cover all situations and scenarios. There's no personalized solution that takes advantage of customer data to target new threats that arise as fraudsters become more sophisticated. For example, when a fraudster uses a skimmed credit card, centralized solutions don't take into account the owner's physical location, spending habits, or typical purchase patterns. To help you combat this problem, researchers at IBM have developed a solution known as Defend. Defend, short for dispersed financial fraud enhanced detection, sits on top of classical fraud detection solutions. It takes your existing fraud detection to the next level with personalized protection that is built into your client's smartphone devices. Most people keep their mobile device within reach, either on them or very close by. We're using this fact to profile users and build user behavior analytics that help verify the authenticity of financial transactions. If your customer likes to travel and shop in new places, classic fraud detection solutions won't raise an alert if they buy a new coat in Mexico, but if their phone's location shows they're still in New York, Defend will raise an alert to verify the transaction. We examine the calls, phone location, browsing history, installed apps, logs, and more to understand what's normal and what might be suspicious. Your bank doesn't know where you are every minute of the day, but your phone usually does. Unlike classical centralized fraud detection solutions that verify transactions only at the bank or parent organization level, Defend uses a dispersed solution. This makes it easier for us to collect data while complying with regulations. It also makes the system more resilient and harder to bypass. In short, we compliment your centralized information and analysis by using the data available on the user's device. All the computation is done locally to protect customer privacy. No personal data leaves the device. Defend maintains user privacy, automatically enforces regulations, and can help you audit privacy policies at multiple levels. Your customers can specify what data of theirs can be used and how. For example, only when there's advanced suspicion of fraud, only for a limited amount of time, and so forth. Let's take a look at how it works. Jake steals credit card information for a living. Sometimes he sells the information to others for about $12 a card, and sometimes he likes to use it himself for a bit of extra cash, or some online shopping. Last week, when Robert withdrew money from the ATM, Jake had set up a secret camera to capture pin codes and a fake skimmer to get card info. Normally, this would give Jake at least several cash withdrawals and a few nice purchases. But the first time Jake tried to use Robert's card to take up money at another ATM, Defend's sophisticated algorithms cross-checked the ATM location, time of day, amount, and keystroke entry with Robert's normal behavior. Because the ATM Jake tried was in a different neighborhood and took place at 11 a.m., a time when Robert was at work, the system immediately raised an alert, and the transaction was blocked. This is Wendy, who's currently in Montreal for a conference of heart specialists. Last night, when she tried to use her credit card to order pizza to her hotel, the credit card company flagged it as suspicious. They assumed that Wendy doesn't like pizza, since the pizza place does not appear in her call history. Beyond that, Wendy lives in New York, and that's usually where she orders her food. With Defend algorithms in place, the order was cross-checked with Wendy's call history and her current location. The transaction was approved and everything went through smoothly. In short, together with Defend, you can protect customers by using data from the smartphone to understand what's normal and what's suspicious financial activity, detecting suspicious behavior that is not in sync with the user's profile without contacting the user, creating a dispersed fraud protection solution that enhances existing systems and reducing loss of trust due to financial fraud. If you'd like more information on Defend or would like to participate in our pilot, contact us at IBM Research Haifa.