Challenges of log and pattern analysis
Nov 30th, 2011 by Jesper Kråkhede
Several of you are surely using different payment services on internet where you could register your credit card and then use the service instead of putting your credit card at risk. It is also possible to accept payments using those sites but sometime there are mistakes made and the account is suspended. Such incidents are constantly ongoing and there are loads of websites like [yourfavouritpaymentproviderehere]sucks.com etc.
So, why are there incidents? First of all, when conducting log and pattern analysis, there is a tremendous amount of data that needs to be analyzed. You probably start off with a working theory or a best practice and go from there. This method is quite often useful but contrary to formal science you seldom challenge your theory but instead build your whole case on it. If it is flawed you won´t find it until mistakes are made and then it could be too costly to change the analysis, hence the payment provider problems.
To mitigate this you should always make sure to either include a process where you challenge your assumptions or better yet have someone challenge you to make sure that flaws are found early. Our brain is a wonderful tool for pattern recognition but it always makes a best guess based on available data.