System Security Lab headed by Prof. Ahmad-Reza Sadeghi and Cryptography and Privacy Engineering Group (ENCRYPTO) headed by Prof. Thomas Schneider have been selected in an international competition to join the Private AI Collaborative Research Institute launched by Intel, in collaboration with Avast and Borsetta.
In this competition top universities were called on to submit research proposals for the Private AI Collaborative Research Institute with the goal to advance and develop technologies in privacy and trust for decentralized AI. In total 9 institute-supported research teams worldwide have been selected for their innovative research ideas.
“Artificial intelligence is currently a true goldmine for cybersecurity research”, said Prof. Ahmad-Reza Sadeghi, head of the System Security Lab and spokesperson of the Cybersecurity profile area (CYSEC) at TU Darmstadt. The focus of his research project is on trustworthy Federated Machine Learning to build efficient and secure AI algorithms and systems both in software and hardware.
“Today, enormous amount of data is generated, collected, and processed by intelligent algorithms and computer systems. We urgently need novel methods of applied cryptography for building privacy-preserving AI systems that guarantee protection of sensitive data”, explains Prof. Thomas Schneider, head of the Cryptography and Privacy Engineering Group (ENCRYPTO). His research project focuses on secure multi-party computation and applications of hardware-accelerated cryptography in decentralized AI.
Under the umbrella of the Private AI Collaborative Research Institute, the partners from academia and industry aim to tackle challenges emerging from the extension of AI into almost all areas of life and industries.
Private AI Collaborative Research Institute
The Private AI Collaborative Research Institute is a joint effort of multiple corporations: it was originally established by Intel, which then expanded the collaborative potential of the institute by inviting Avast, a global leader in digital security and privacy products and Borsetta, a start-up developing edge-computing solutions, to join forces. The goal of the Private AI Collaborative Research Institute is to fund the development of fundamental technologies that are instrumental in strengthening the security and trustworthiness of decentralized artificial intelligence (AI).
The research projects in detail
Engineering Private AI Systems (EPAI)
Principal investigator: Thomas Schneider
For EPAI, the ENCRYPTO group at TU Darmstadt will develop basic technologies to build private AI systems, investigate their orchestration strategies to optimize efficiency and costs on a given network and compute infrastructure, and systematically validate these to allow for automatically selecting the most efficient solution for a specific usage scenario. As underlying technologies, the ENCRYPTO group will mix different building blocks from cryptography and hardware, including secure multi-party computation, hardware acceleration, and trusted execution environments.
Decentralized Trustworthy Federated Learning (TRUFFLE)
Principal investigator: Ahmad-Reza Sadeghi
For TRUFFLE, TU Darmstadt will design a framework for FL that provides comprehensive security and privacy. The design will be resilient against various attacks, such as poisoning, and will incorporate privacy-enhancing technologies based on decentralized aggregators and advanced crypto-based primitives to address privacy requirements of FL. TRUFFLE will incorporate hardware-assisted security and trusted execution environments of varying capabilities.