Privacy-preserving tax fraud detection in the cloud with realistic data volumes Version 1 . 1

Tax fraud detection is a suitable use-case for secure multi-party computation to be deployed in the cloud, allowing governments to detect tax fraud by analysing companies’ business transactions, while enterprises maintain control of their private data. This report describes an efficient prototype application that analyzes tax declarations in a privacypreserving way using the Sharemind © secure computation platform developed by Cybernetica. The prototype has been deployed and benchmarked in the Amazon EC2 cloud with realistic data volumes – the size of Estonia’s economy. The benchmarks show unprecedented results in costefficiency of processing large amounts of data with secure computation techniques. In the cloud, we are now able to securely process 100 million business transactions in a matter of hours with less than $100.