Descripción del proyecto
Over the past few decades, our use of networks such as the Internet underwent a complete transformation: the data that we send over the network is now routinely used in computations by third parties. This creates a considerable threat to our privacy, as we witness countless data breaches and companies selling our private data.
The aim of secure computation is to solve this problem. It studies the design of methods for computing on sensitive data without compromising confidentiality. Secure computation has been firmly established as one of the most promising approaches to reconcile the use of data-driven algorithms with the necessity of protecting our privacy. It has spurred a considerable research effort in the past decade, which has witnessed the emergence of new models, such as the models of secure computation with silent preprocessing (introduced by the PI), and of non-interactive secure computation.
While these silent and non-interactive features are fundamental for the deployment of secure computation over large networks, current protocols suffer from important downsides which make them entirely impractical in real-world scenarios:
- They scale poorly with the number of participants and cannot be used over mid- to large-size networks.
- They either require many rounds of interaction (hence are impractical over high-latency networks) or trade non-interactivity for huge computational costs.
- They induce a large communication overhead compared to executing the computation on the clear data.
- They are often based on exotic cryptographic assumptions, whose security is not yet well-understood.
The aim of the project OBELiSC is to explore the current limits of silent and non-interactive protocols, and to introduce new methods for overcoming the remaining barriers. The long-term vision of the project is to enable the deployment of a large-scale network that further protects our sensitive data whenever it is used in distributed computation.