Crossing is a joint effort of scientists from quantum physics, cryptography, system security, and software engineering who collaborate in three interconnected project areas. In Project E5, we're doing research on privacy-aware distributed computation.

Edge computing models take advantage of the processing power of decentralized cloud infrastructures, such as geo-distributed data centers, as well as smaller devices such as mobiles and IoT devices. Edge computing enables the distribution of data processing, which maximizes efficiency, optimizes resource

usage, improves cost-effectiveness, and achieves the latency requirements that IoT applications require. However, in such a setting, data and computations may be transferred and processed in variable, potentially untrusted domains which compromise their security.

The result is either lack of privacy or significant inefficiency, as the privacy-performance trade-off is not considered during deployment. This is where Project E5 will advance the design of edge computing systems.

More specifically, we focus on scenarios where a client wants to query and aggregate data from several data sources located on separate hosts that may contain sensitive information. The client relies on distributed computation resources, including cloud nodes and other edge devices,

which may or may not have the rights to access all such data sources. To enable secure and efficient processing of sensitive data, the main goal of Project E5 is to design an edge computing system that automatically distributes subcomputations, such as query language operators, among processing units

of the system in order to optimize performance, while also protecting the processed data from unauthorized access. For data protection, we will consider various cryptography-based techniques in conjunction with privacy-aware placement of data and computations on processing units.

This will also include research results on privacy-preserving computations achieved in other crossing projects like E4 and S6. As privacy-preserving techniques come with different performance security trade-offs, we will model their properties and make the mechanisms interoperable, composable, and uniformly accessible.

To be able to label data with privacy requirements, we will design a specification language and a type system and will develop methods to formally reason about the specification language and the security properties it enforces via cryptographic methods. We will also develop an engine that automatically determines how to distribute encrypted data and computations among trusted and untrusted servers.

With our research, we aim to design edge computing systems that are not only capable of meeting the challenges of an interconnected world, but also enable security by design.

To learn more about privacy-aware distributed computation, please visit our website.