COVID-19
Fighting fraud in the time of COVID-19
By Uri Arad, Identiq
We’re all having to get used to a lot of changes during the current crisis, some more difficult than others. But there’s one group which excels at adapting. Fraudsters are alive to the possibilities for crime opened up by the crisis, and your regular statistical models won’t be able to protect you the way they usually do.
Watch Out for AI: Your Silver Bullet is Boomeranging Back to Hit You Many fraud teams have become reliant on machine learning and Artificial Intelligence in recent years, and it’s easy to see why. This technology has tremendous potential when it comes to analysing normal patterns and predicting whether transactions/sign-ups are legitimate on that basis. Many companies have benefited hugely from adopting this technology.
The trouble at the moment is that all such statistical models rely on the present being much like the past, within reasonable variation and seasonal fluctuation. And right now, that’s just not the case. Nothing now is the way it usually is.
The Times They Are A’Changin’ - But It’ll Take Your System Time to Catch Up Some businesses are seeing far more demand than usual (home entertainment, delivery businesses, groceries, remote working, virtual education and so on) while others are experiencing dramatically reduced traffic (travel, vacation rentals, events planning, ticketing etc.).
Moreover the user behaviours have changed. The items that are “hot” this month are likely to be completely different to those last month. They may well be mundane (or bizarre) goods which have never been particularly popular before.
www.directcommercemagazine.com | Direct Commerce
People are online more than usual, but at different times of day. Work purchases are being made from different IPs. Children are being given more license to purchase or drive purchases, in many families. Transaction velocities and amounts will be all over the place.
Statistical models struggle with this sort of change. Even with your manual involvement in retraining and adding rules, your system won’t adapt overnight. This results in many false declines, as well as an opportunity for fraud to hide in the crowd.
It’s hard at a time like this when there are so many competing priorities, but to fight fraud effectively your team needs to give the system more oversight than usual, with more random sampling and manual reviews. Focus on identities rather than the usual statistical patterns, for a while.
Regarding your statistical models, remember to isolate this data set from the normal ones, so it doesn’t confuse future results once things return to a state similar to the past, or when you analyse current trends.
Pro Tip: Use this opportunity to build a system that’s designed for reacting to crises. It’s valuable to have a model ready to go for times of huge change - done right, this could be a true business asset for the future.
Stay safe, stay healthy, and stay alert. And stay in touch with other fraud prevention teams, especially in your space. They’ll be seeing similar things, and sharing best practices is the fastest and best way to guard against your shared enemies. Now is a time to stick together - virtually.
17
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42
