Tuesday, March 31, 2009

Here comes the scary part

It is a dark night in Tel Aviv, the kind in which bad things lurk in dark corners. Sitting in a small cafe with the security expert, I hear the wind blowing between the trees. The waitress looks worried as well. A dark night indeed. I look across the table to my partner, a serious person with thick eye glasses that add to his already grim demeanor. Then, to accent his last sentence, he leans towards me - his glasses almost opaque in the dim light - and barely whispers: "but you know, man, you know what the real problem is, right? the real problem is BOTS!".

Oh, is it really?

I didn't go to the MRC conference this year. Somehow, boogieman stories from interested third parties (over early morning session, in Vegas!) sounded less appealing for someone who needed to fly 18 hours for the experience. I did, however, read excerpts and ideas. Boy, I have to admit that the set up was a lot more successful than a Tel-Aviv cafe. Because here's the thing with 3rd party vendors - they are looking to sell, and if you're looking for the real gap in your system (rather than the perceived one), you probably shouldn't be looking at that direction. Let's see what's hot this year: it sppears that Malware and Botnets are attacking everyone, and that Machine ID and phone verification might be the only way of stopping this.

Now's probably the right moment to wonder what's my case. True. Here's my case: buying flashy new technologies when you haven't exploited the old ones is pricy, redundant, plain dumb sometimes. Most of the merchants that will purchase anti-malware and machine id solutions do not, I bet, have a decent user-location system in place, and are instead declining multiple good buyers who live in a set of black-listed locations; most of the merchants that will purchase phone verification products will double their fraud operation costs before they realize that calling alarge percentage of the transaction volume only slows them down instead of bringing that solution to loss mitigation. My case is - proper analysis of what you're dealing with, rather than going with the nifty, trendy new fraud filter, will bring you much higher ROI and a method for solving your own problem. It does, however, require some extra effort that cannot be bought off the shelf: training the right kind of people to do the right kind of work. More on that in future posts.

1 comment:

Stock search said...

I agree - in card fraud have seen many well made detection models totally sub-optimized and lose fraud dollars, only to be replaced by an expensive new model which is sub-optimized in turn.

The real solution was using the old one properly and rep-level training. Make sure you get the basics right, only then get the new nifty technology is a good maxim.