250,000 per hour, 4,200 per minute or 70 per second: this is the average number of transactions such as payments and transfers that PostFinance has to check and analyse to detect fraud. To process such large volumes of data, the financial institution has opted for an application based on stream processing – a technology that is used in the big data environment, for instance. A single CPU is limited by technical restrictions and hence computer performance, meaning that it cannot simply be made to work faster or to transport unlimited quantities of data at will. Consequently, other solutions are needed. “With our operational intelligence streaming application, on the other hand, we can monitor all transactions and events within milliseconds, virtually in real time,” explains Bruno Müller, Monitoring Compliance, Risk and Fraud Solution Team Leader at PostFinance. “And from a technological point of view, it would be capable of handling ten times as many transactions.”
Big data: combating fraud with stream processing
Operational intelligence is a PostFinance application that checks transactions to detect fraud in real time. It is based on stream processing – part of the fascinating IT field of big data technology.
Major challenge, small steps
Although the overall task is a huge one, each of the individual steps that makes up stream processing is tiny. “To process such large amounts of data, we work with microservices or microbatches which connect and disconnect from one another while also using parallelization to cope with the high volumes,” explains Ralf Doerendahl, Application Manager of the operational intelligence streaming app at PostFinance. And that’s what makes the work so interesting: the developers write microservices that evaluate and process data. The fact that it is not a matter of building a small app, but of developing a central solution for one of Switzerland’s leading financial institutions makes the task even more interesting for the IT specialists. “We’re talking about major computing stuff that’s being offered here,” says Doerendahl. In particular because we are also working with state-of-the-art technologies, adds Matej Ilicic, IT architect at PostFinance. These include Apache Kafka and Apache Spark as the basic technologies for streaming, as well as artificial intelligence, machine learning and cybernetics.
Constant expansion of the streaming application
The streaming application is constantly being expanded due to permanent new technical requirements. “The range of services is continuously growing as part of digitization. Since each product is subject to different, and partly new security requirements that need to be met, we keep having to make the necessary adjustments by integrating new channels and services,” explains Bruno Müller. “And it’s also important to keep pace with the times in general, because attackers are thinking up new strategies to outsmart us all the time,” adds Ralf Doerendahl. In this respect, the further development of the streaming application is an ongoing project.