Microsoft has developed a solution using artificial intelligence (AI) that can identify fraudulent mobile banking transactions in less than two seconds.
Developed for a banking client using Azure technology, Microsoft has now open sourced the solution to benefit its clients.
Rima Semaan, AI Lead Gulf at Microsoft says the solution is a perfect example of how AI can be mobilised to empower organisations to transform their business.
“At Microsoft, we have been trying to transform critical businesses processes using AI. Banking along with government, manufacturing and retail are the key sectors that can optimise the use of AI for their benefit,” says Semaan. She was speaking during the AI Week Middle East, that was held in Dubai.
In a detailed guide which Microsoft has published explaining the process of the new model, Kate Baroni, the author notes that the financial industry is expecting losses due to mobile bank frauds to increase by almost 100 per cent year-over-year.
According to Baroni, the power of AI becomes significant as today’s existing models work on rule-based engines and do not adapt quickly to new or evolving types of attacks. The current system is ineffective as it does not provide real-time detection, and fraud is detected only after financial loss occurs.
“Rules are hard coded into business logic. Curating the rules, incorporating new data sources, or adding new fraud patterns usually means application changes that impact a business process. Propagating changes throughout a business process can be cumbersome and expensive,” writes Baroni.
The model managed to identify the behaviour pattern after carefully analysing various instances of fraud, the patterns its perpetrators followed and different ways it was committed. The data attributes were mapped to the messages collected in an effort to profile account behaviour that were most relevant for identifying fraud.
The time frame of less than two seconds, the author argues should be the maximum amount of time between when a mobile banking activity gets forwarded for processing and when it needs to be assessed for fraud.
“Latency and response times are critical in a fraud detection solution and the infrastructure to support it must be fast and scalable,” he argues adding that the use of AI models, has the potential to dramatically improve fraud detection rates and detection times, and more banks are using them in combination with other approaches to reduce losses.