"Futureproofing Computer Security"
Dr. Lisa Kohl is an expert cryptographer at Centrum Wiskunde & Informatica (CWI), the Netherlands' national mathematics and computer science research institute. She and her colleagues are looking for new ways to keep people safe online. Kohl's research revolves around secure multiparty computation in which several parties want to share data without revealing personal information. A hospital, for example, needs to protect patient data while also sharing information with researchers for a study. A bank may want to use its data to detect money-laundering activities. If the data was shared in an insecure manner, each party would analyze it on their own computer or network. The computation's complexity would be determined by the size of the data input. However, extra computing resources are required for secure computation, and there is a significant increase in communication, according to Kohl. The input and complexity of the program to be computed can also have an impact, which kills secure computation in many applications. Kohl's research focuses on how to achieve secure computation while remaining highly communication efficient. Kohl and her colleagues have recently had some promising results as they have been developing a 'preprocessing phase,' in which some of the complex analysis required for secure computation is performed offline ahead of time. The computation can then be carried out quickly when data is exchanged in the online phase. Previously, a significant disadvantage of this preprocessing phase was that it required a significant amount of computational time and storage, as part of the process involved generating many random multiplications. These random multiplications prevent adversaries from learning the private data of the other parties. The team has now developed 'pseudorandom correlation generators,' greatly improving the efficiency of this process and alleviating the bottleneck. Depending on the complexity of the program, multiple gigabytes of storage may be required, but with pseudorandom correlation generators, this can be compressed by a factor of 1,000 or more. Therefore, less communication and storage are required to set it up. As network speed is frequently a bottleneck, this results in significant speed-ups in practice. This article continues to discuss the team's advancement in secure computation.