Telephone fraud detection

Recently I received an email from a fellow student about a project in the media I was heavily involved. During this project my team and I worked for a small telecommunication company (www.toplink.de) in cooperation with the Hochschule Darmstadt to develop a solution to stop telephone fraud. Telecommunication fraud has become a serious problem for all providers and their customers and produce high costs. There are different types of fraud, e.g. compromised customer phone boxes to do calls to special telephone numbers. The problem with this is the customer has to pay for it even if it is not his fault and the provider needs to keep his customers happy. So there has to be a solution to detect those anomalies in calling behavior to react in a short-time and prevent further damage. The software we developed analyses the CDR¹⁾, which have to be logged by every provider. They contain alot of information about the caller and callee, e.g. duration, type of call termination, route, and a lot more. These are the data which were discussed to be saved by the german data retention law over a long time.

My part in this project was to write a parser to read the CDR into our system, so I wrote a framework to read and provide a plugin interface to analyse the data objects. Every CDR entry is transformed into a object for easier data handling during programming. With the plugin interface you can create a filter chain, where you can feed the result of one filter into another to work with. The main challange is how to analyse customers without storing too much information and without invading their privacy. With the capability to chain filters you can minimize the needed data to detect frauds. After I finished the framework my fellow students were able to write filters in Python. The simple plugin API and the object oriented aspect of the CDR I introduced, a easy and fast way to implement all kind of filter techniques is provided.

Press releases:

Update:

And again a publication in the media, this time it is availble as scanned page from a newspaper called “Wirtschaftsecho”²⁾ Scanned page from Wirtschaftsecho (german).

¹⁾ Call detail records http://en.wikipedia.org/wiki/Call_detail_record

²⁾ Thanks to Torsten Wiens providing me a digital copy

This blog post was created on 2011-10-16 at 16:44 and last modified on 2011-10-17 at 18:08 by Julian Knauer.