Businesses have found an ideal way to detect fraudulence and take actions. Auditors are possessing a new defensive tool on this: artificial intelligence.
Association of Certified Fraud Examiners estimates that an organization can lose about 5% of its annual revenue to fraud. This is where AI comes up to anomaly detect frauds.
Marco Schreyer, a researcher associated with Deep Learning Competence Center for the German Research Center for Artificial Intelligence
, is putting his works on accounting fraud, anomaly detection, and unsupervised deep learning. He pointed out that financial fraud can be a reason of reputational setback of a company alongside loss in revenue.
It is the time when companies are on the verge of accelerating digitalization of their businesses. This affects their enterprise resource planning software and lets them keep an audit trail, says Schreyer. Fraudsters, on the other hand, are upgrading themselves too, deviating from ordinary system usage and accounting patterns, creating financial loopholes.
Using trained neural networks, we’re looking at the flow of transactions and how they are captured in systems. - Marco Schreyer, a researcher with the Deep Learning Competence Center for the German Research Center for Artificial Intelligence
DGX-1 system is being used to train autoencoder networks helping the Deep Learning Competence Center for detecting accounting anomalies. It had been tested on two separate real-world accounting datasets. The network was successful in flagging anomalous data entries for audit.
Schreyer foresees more benefit from this kind of network: “The more you can contrast those networks, the better you can track the data. This approach can be used to detect anomalies in different spaces, not just finance.”