Innovative CMU Secure Blockchain Initiative Research, in Collaboration with Anaxi Labs, Focuses on Improving the Efficiency of Cryptographic Proof Systems

Tuesday, December 10, 2024 - by Michael Cunningham

In cryptography, proof systems are mathematical frameworks that allow one party (the prover) to convince another party (the verifier) that a given statement (often: this program was executed correctly on a specified input) is true.

Proof systems can give amazingly useful properties like succinctness—meaning that the proof is much easier to check than the statement being proved—and zero knowledge—meaning that the proof reveals no information beyond the truth of the statement itself. Because of this, they play a critical role in many cryptographic protocols, particularly ones that ensure privacy, security, and trust. They are used in a wide variety of enterprise functions, ranging from blockchain technologies and digital identity authentication systems to supply chain management and financial transactions.

Unfortunately, building new applications on top of proof systems is challenging because these systems are both difficult to program and expensive to use. In response, developers have generally taken one of two approaches. One is to carefully craft a CPU emulator on top of the proof system’s low-level computational representation, and then to run the application on that emulated CPU; the other is to directly translate the application in the proof system’s low-level representation. In rough terms, each approach optimizes for one of the main challenges mentioned above: the CPU emulator approach is generally easier to program, whereas the direct translation approach is potentially much less expensive.

A team of Carnegie Mellon University researchers featuring Riad Wahby , assistant professor in the Department of Electrical and Computer Engineering, Kunming Jiang , Ph.D. student in the Computer Science Department, and Fraser Brown, assistant professor in the Software and Societal Systems Department, is overcoming this tradeoff in new research that is supported by Anaxi Labs.

CMU’s research presents an approach that involves compiling high-level software directly to a low-level representation without hand-written protocols and constraints. This approach combines techniques from both CPU emulation and direct compilation, and results in significant efficiency improvements compared to current state-of-the-art systems.

“The idea is to do a careful analysis of the program in order to produce a representation that's broken up into small, indivisible units; then those units become the ‘language’ of the execution,” said Riad Wahby. “This method of breaking the computation into program-specific chunks that take the place of a CPU in an automatic way is a new approach, and we’re excited about it.”

This is the second research project coming out of Anaxi Labs’ sponsorship of the Secure Blockchain Initiative through its partnership with Carnegie Mellon University’s CyLab, the university’s cybersecurity and privacy institute. CyLab and Anaxi Labs are working closely to apply these findings to dramatically speed up the buildout of secure and scalable web3 applications.

“Even the most advanced cryptography products in the industry today use a static, monolithic approach that is often labor-intensive and unauditable, limiting their applicability,” said Kate Shen, cofounder of Anaxi Labs. “This research enabled us to build a performant, language-agnostic but automatic framework. This paves the way for an open, adaptive design paradigm for cryptography with optimizations that can be automated, by combining the best of the latest advancements in proof systems (such as folding and lookup), co-processors and hardware acceleration, maximizing the performance gains of each computational substrate.

“This has significant implications in many important industry applications today that involve massive performance overheads, such as [zero-knowledge] ZK and [Ethereum Virtual Machine] EVM, bringing us one step closer to our vision of cryptographically-secured decentralized consensus with real-time settlement.”

Carnegie Mellon’s CyLab is the university's security and privacy research institute. Its mission is to catalyze, support, promote, and strengthen collaborative security and privacy research and education across departments, disciplines, and geographic boundaries to achieve significant impact on research, education, public policy, and practice.