Project dates: 01. Mar 2022 - 28. Feb 2027
Electronics are part of all aspects of our daily life, at home, at work, while on the road. Smart cities, medical devices, financial transactions, self-driving cars, all rely on electronics. At the same time, these activities generate huge amounts of data, which for privacy and security reasons, are stored encrypted on a possibly untrusted cloud server. Yet, many applications benefit from operations on this data. Statistics or machine learning on medical, financial, automotive or energy data could discover trends or abuse, could be used to monitor pandemics, to tune supply and demand, and much more. Ideally, the calculations on the data should be done without decrypting them first to avoid data leaks.
Computing on encrypted data is the new magic in the field of cryptography: it enables calculations on the encrypted data, while data remains encrypted in the cloud and without the need to decrypt it. The result will only be decrypted by the final recipient. Challenging in these novel mathematical concepts is a gigantic blow-up in the size of ciphertext data, in the amount of calculations on the encrypted data, and in the novel lattice based arithmetic used.
The core challenge addressed in BELFORT is to make this new technology feasible from a hardware perspective and in realistic scenarios, providing the best performance/energy balance, for it to run on cloud servers and edge devices, within reasonable memory, computation and energy budgets. We propose reconfigurable hardware-based acceleration with domain-specific programmable co-processors, which combine energy efficiency with domain specific programmability and with resistance to attacks from classic and quantum computers. We will develop multiple new ideas as well as bring them together in an end-to-end demonstrator.
The PI has the unique expertise and is a pioneer in translating novel cryptographic algorithms into efficient and secure realizations on a wide range of hardware/software platforms.